The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch

Modal CEO Erik Bernhardsson predicts inference market growth, discusses GPU supply, and outlines platform strategy.

1h 2m Listen to the original Episode page
AI InferenceGPU Supply ChainVenture CapitalCloud InfrastructureOpen Source LLMsStartup Strategy

Summary

Erik Bernhardsson, CEO of Modal, argues that the inference market will grow exponentially (5x-20x) rather than peaking, driven by bundling inference with post-training and agentic tools. He details Modal’s strategy of aggregating GPU demand to smooth capacity planning, noting a 30-40% month-over-month growth in compute demand and a current supply shortage. Bernhardsson asserts that NVIDIA’s moat lies in its full ecosystem, not just CUDA, and predicts long-term hardware price decreases. He also highlights Modal’s internal adoption of coding agents, spending ~$500k/month, and disputes the 'death of seed investing' by emphasizing capital efficiency and product-led growth.

Episode sections

  1. 0:00 – 8:00

    Founder Background and Company Origins

    Erik Bernhardsson discusses his early start in coding and founding Modal at age 36-37, emphasizing a preference for long-term building over immediate exits.

  2. 8:00 – 10:15

    Fundraising Philosophy and Capital Efficiency

    Bernhardsson explains his strategy of raising small rounds at lower valuations to maintain investor hunger and avoid dilution, noting Modal’s profitability despite increased GPU burn.

  3. 10:15 – 13:45

    Defensibility: Speed, Team, and Brand

    The speakers discuss how early-stage defensibility comes from team quality, speed, and brand personality, citing TypeSafe as a model for community engagement.

  4. 13:45 – 15:30

    The Inference Market as a New Cloud

    Bernhardsson predicts the inference market will match the cloud market in size, arguing that value lies in bundling inference with post-training and agentic tools.

  5. 15:30 – 17:55

    Margin Compression in the AI Stack

    The discussion highlights software margin compression from 90% to 25-40% due to the multi-layered vendor stack, with Modal projecting long-term margins of 40-60%.

  6. 17:55 – 20:38

    GPU Procurement and Aggregation Strategy

    Bernhardsson describes the GPU market as chaotic and explains Modal’s strategy of aggregating demand to smooth out variable workloads without long-term customer locks.

  7. 20:38 – 23:38

    GPU Supply, Demand, and Pricing Dynamics

    Bernhardsson notes 30-40% month-over-month compute demand growth and a current GPU shortage, predicting long-term hardware price decreases as supply catches up.

  8. 23:38 – 25:22

    NVIDIA's Moat and Competitive Landscape

    He argues NVIDIA’s moat is its full ecosystem, including kernel optimizations, and predicts it will continue to dominate despite growth from specialized chip companies.

  9. 25:22 – 27:22

    Neo-Clouds: Friends or Foes?

    Bernhardsson expresses skepticism about neo-clouds moving up the stack to own inference, citing the difficulty of replicating Modal’s infrastructure engineering.

  10. 27:22 – 30:08

    Open Source vs. Frontier Models

    He argues open-source models will persist due to customization needs, contrasting US and European capabilities in capital accumulation and talent aggregation.

  11. 30:08 – 37:58

    Modal’s Product Strategy and Growth

    Bernhardsson details Modal’s evolution into a multi-product platform, highlighting the success of 'Sandboxes' and the importance of building foundational building blocks.

  12. 37:58 – 40:14

    Market Outlook and Valuation

    He disputes the view that open-source LLM inference has peaked, predicting exponential growth for both open-source and frontier markets, and notes Modal’s $1 billion revenue milestone.

  13. 40:14 – 41:33

    Modal's Growth Ambitions and Product Adjacencies

    Bernhardsson discusses Modal’s expansion into training, post-training, and data infrastructure, acknowledging challenges in prioritization.

  14. 41:33 – 1:02:09

    The Limits and Scope of Post-Training and RL

    He argues post-training and RL are effective for tasks with finite feedback loops but may face limitations in general software engineering.

Assets discussed

AssetKindStanceContext
NVIDIANVDAequityneutralLike today you have applications, you have models, you have infrastructure, you have cloud, you have NVIDIA.

“Stance” is how the speakers talked about the asset, not a recommendation.

Claims worth verifying

Modal raised $750 million led by Excel and is priced at $15.75 billion.

Unidentified speaker ·0:45

How to verify: Check Modal’s official press releases or SEC filings (if public) or reputable news sources covering the D-round announcement around November 2026.

Modal is profitable on a P&L basis but has negative cash flow due to GPU capacity spending.

Unidentified speaker ·9:54

How to verify: Review Modal’s financial disclosures or investor updates regarding P&L vs. Cash Flow statements.

Modal works with 25 different 'neo clouds' to aggregate compute.

Unidentified speaker ·18:31

How to verify: Check Modal’s technical documentation, partner pages, or engineering blog posts listing cloud providers.

GPU lead times are currently about three to four months.

Unidentified speaker ·20:28

How to verify: Verify with industry reports from NVIDIA, cloud providers, or supply chain analysis firms in late 2026.

Modal controls approximately 0.1% of the world's compute.

Unidentified speaker ·18:36

How to verify: This is an internal estimate; verify by comparing Modal’s reported GPU capacity against industry-wide estimates of total AI compute capacity.

Inference providers generally operate with gross margins in the 25 to 40 percent range.

Unidentified speaker ·15:34

How to verify: Analyze financial reports of public AI infrastructure companies (e.g., CoreWeave, if public, or competitors) or industry benchmarking reports.

Bernhardsson started coding in 1992 at age eight.

Unidentified speaker ·4:50

How to verify: Verify Erik Bernhardsson’s birth year (approx. 1984) and early career history via LinkedIn or public biographies.

Modal built its own file system, container runtime, and scheduler, avoiding Kubernetes and Docker.

Unidentified speaker ·19:00

How to verify: Review Modal’s engineering blog posts or technical whitepapers detailing their infrastructure stack.

TypeSafe launched overnight on Modal.

Unidentified speaker ·19:50

How to verify: Check TypeSafe’s launch announcements or case studies featuring Modal as their infrastructure provider.

Modal’s long-term gross margin target is 40-60%.

Unidentified speaker ·16:30

How to verify: Monitor Modal’s investor presentations or financial guidance for stated margin targets.

Compute demand has been growing at 30-40% month-over-month for the last 12 months.

Unidentified speaker ·20:38

How to verify: Check Modal’s public financial disclosures or earnings reports for revenue growth rates and customer acquisition metrics during the 12 months prior to November 2026.

Modal has reached $1 billion in revenue.

Unidentified speaker ·38:38

How to verify: Verify via Modal’s official press releases, funding announcements, or public financial filings if available.

Modal’s 'Sandboxes' product has a $300 million run-rate and accounts for about a third of Modal’s revenue.

Unidentified speaker ·34:54

How to verify: Cross-reference with Modal’s internal financial data or public statements regarding product-specific revenue breakdowns.

NVIDIA acquired open-source model companies, specifically mentioning 'First Pool Side' (likely First Mode) and 'Reflection' (likely Refraction or similar).

Unidentified speaker ·29:04

How to verify: Check NVIDIA’s official acquisition announcements and press releases for deals involving open-source model providers.

Modal’s Sandboxes product was launched in 2023 and saw significant success in summer 2025.

Unidentified speaker ·34:33

How to verify: Check Modal’s product launch history, blog posts, or press releases from 2023 and 2025.

Jensen Huang stated that supply will catch up with demand within one to five years, affecting pricing.

Unidentified speaker ·21:48

How to verify: Search for Jensen Huang’s remarks from his London visit in October 2026 (prior to the Nov 2, 2026 episode) via news transcripts or NVIDIA investor presentations.

Modal grew 15x in the last 12 months.

Unidentified speaker ·39:10

How to verify: Verify via Modal’s financial reports or investor updates comparing revenue from Nov 2025 to Nov 2026.

Modal is valued at $15 billion (implied by Aaron Powell’s comment on the price).

Unidentified speaker ·39:33

How to verify: Check recent funding round announcements or valuation reports for Modal from late 2026.

Cerebras stock was down 30% at the time of the interview.

Unidentified speaker ·24:15

How to verify: Check Cerebras’ stock price history on financial platforms like Bloomberg or Yahoo Finance for the date of the interview (Nov 2, 2026).

Modal staffed up a team from 1-2 people to 30 people in 12 months to focus on open-source inference after the release of Llama ('Deep C').

Unidentified speaker ·37:40

How to verify: Verify via LinkedIn employee growth data for Modal or internal company announcements regarding hiring spikes in 2024-2025.

Modal hopes to reach a $100 million run rate in the next six to twelve months for training and post-training services.

Unidentified speaker ·40:19

How to verify: Check Modal's public announcements, press releases, or financial filings for revenue guidance or run rate milestones.

Modal spends approximately $500,000 per month on coding agents.

Unidentified speaker ·44:15

How to verify: Verify against Modal's internal financial disclosures or public statements regarding AI infrastructure costs.

Brandon McCall stated that some companies spend three times their human salaries on inference.

Unidentified speaker ·44:26

How to verify: Search for Brandon McCall's public comments, tweets, or interviews where this specific ratio is cited.

GitHub Copilot (referred to as Claw Code) generates $60 billion in revenue.

Unidentified speaker ·44:59

How to verify: Check Microsoft's earnings reports or official GitHub announcements for Copilot revenue figures.

Mark Benioff stated that Salesforce spends 5% of engineers' salaries, or $300 million, on Claw Code.

Unidentified speaker ·45:05

How to verify: Search for Mark Benioff's public statements, earnings calls, or interviews referencing this specific spend figure.

Flex allows users to float every purchase for 60 days.

Unidentified speaker ·1:00:30

How to verify: Review Flex's official website or product documentation for payment terms and credit facilities.

Do your own research

  • Verify Modal’s revenue figures ($1 billion total, $300 million from Sandboxes) against any available public filings or credible third-party reports, as private company data is often unverified.
  • Check NVIDIA’s latest earnings reports and investor presentations to confirm claims about ecosystem dominance and competitive landscape against specialized chip makers like Cerebras and Etched.
  • Research the current market pricing and lead times for GPUs to verify Bernhardsson’s claims about the 'chaotic' procurement market and 3-4 month lead times.
  • Analyze Anthropic’s IPO prospectus (if filed) or public disclosures to verify the 'Claw Code' (GitHub Copilot) revenue figure of $60 billion cited in the episode.
  • Review open-source LLM performance benchmarks and cost metrics to assess the validity of the claim that open-source inference is not a 'race to the bottom' when bundled with other services.
  • Investigate the financial engineering models of neo-clouds like CoreWeave and Nebius to understand the risks associated with their long-term contract and debt repayment structures.
  • Examine the technical limitations of current reinforcement learning models in general software engineering contexts to evaluate the claim that RL has finite applicability.
Full transcriptexpand
0:00
Unidentified speaker: I think there is an opportunity to create a hundred billion dollar company with a lot of specialized models in inference.
0:05
Unidentified speaker: There's gonna be a pullback.
0:07
Unidentified speaker: I'm okay with that.
0:08
Unidentified speaker: Because I think the long-term opportunity is so fucking big.
0:11
Unidentified speaker: There's just so much demand for it.
0:12
Unidentified speaker: It's like so large.
0:13
Unidentified speaker: I don't want to sell this company because I would just go and start model two.
0:16
Unidentified speaker: We've lost new grads going to, you know, some of the labs for like inside like talking, you know, 20 plus million.
0:22
Unidentified speaker: I mean, these are like exceptional people.
0:24
Unidentified speaker: Exceptional.
0:25
Unidentified speaker: This is 20 VC with me, Harry Stebbings.
0:27
Unidentified speaker: And do you know what I wish I knew 12 months ago that I now know?
0:30
Unidentified speaker: The demand for inferences is infinite.
0:33
Unidentified speaker: And we're seeing some of the biggest companies in the world be built in this space.
0:37
Unidentified speaker: So today I'm thrilled to welcome one of those founders to the hot seat in the form of Eric, the co-founder and CEO of Modal.
0:43
Unidentified speaker: Now, Modal just hit a billionaire.
0:45
Unidentified speaker: They just raised $750 million led by Excel, and they're priced at $15.75 billion today.
0:52
Unidentified speaker: I actually think it's a deal, by the way.
0:54
Unidentified speaker: But today we dig into the inference market, what to come, what no one sees that everyone should see, and so much more.
1:01
Unidentified speaker: This was a phenomenal discussion with Eric.
1:03
Unidentified speaker: But before we dive into the show today, Blitzy, the autonomous software development platform that's built for enterprise scale, is proud to sponsor today's show.
1:12
Unidentified speaker: Enterprise code bases have grown beyond human comprehension.
1:15
Unidentified speaker: Tens of millions of lines, decades of tech debt, the complexity existing tools cannot fathom.
1:21
Unidentified speaker: Blitzy starts by understanding.
1:23
Unidentified speaker: Thousands of agents reverse engineer your code base into a proprietary knowledge graph that maps architecture dependencies and business logic.
1:30
Unidentified speaker: That graph surfaces every CVE and zero-day hiding in your code before anyone hands you a ticket.
1:37
Unidentified speaker: Grounded in that context, Blitzy then executes entire epics autonomously, modernize off legacy systems, ship new features 5x faster, or batch remediate CVEs at scale surfaced by the platform.
1:50
Unidentified speaker: Blitzy not only generates code, it validates and end-to-end tests every single line.
1:56
Unidentified speaker: And that's why Fortune 500 enterprises are shipping enterprise grade code five times faster.
2:02
Unidentified speaker: Go to blitzy.com forward slash 20vC to learn more.
2:07
Unidentified speaker: While Blitzy helps you build faster, Flex helps you run your finances smarter.
2:12
Unidentified speaker: Business owners run their whole financial life on Flex.
2:16
Unidentified speaker: One platform from business revenue to their personal spend.
2:19
Unidentified speaker: Float every purchase for 60 days, tap capital that grows with your revenue and pay vendors in 170 countries across 32 currencies.
2:29
Unidentified speaker: Plus the whole back office, bills, expenses, accounting, all in one place.
2:35
Unidentified speaker: So you spend less time reconciling and more time growing.
2:39
Unidentified speaker: That's why thousands of owners use Flex, named one of Fast Company's most innovative companies of 2026.
2:45
Unidentified speaker: Visit Flex.1, that's F-L-E-X.on e and use the code 20VC.
2:54
Unidentified speaker: While Flex runs your finances, Vanta keeps your compliance covered.
2:58
Unidentified speaker: When you're building a company, you learn that trust is what closes deals.
3:02
Unidentified speaker: You may have the best product, but no buyers will sign these days without proof of your security.
3:06
Unidentified speaker: Here's what happens if you're not prepared.
3:08
Unidentified speaker: A prospect asks for proof of compliance.
3:10
Unidentified speaker: The deal stalls when you scramble, your engineer gets pulled off the roadmap to audit prep.
3:15
Unidentified speaker: Every enterprise conversation turns into this horrible fire drill.
3:19
Unidentified speaker: That's where Vanta comes in.
3:20
Unidentified speaker: Vanta is the leading agentic trust platform that not only gets you compliant fast with frameworks like SOC 2, ISO 27001, HIPAA, and GDPR, but keeps you compliant by continuously monitoring your controls.
3:33
Unidentified speaker: So your deals keep moving and your engineers really keep building.
3:36
Unidentified speaker: Now access the Vanta agent everywhere you work, even if your team lives in Claude or Cursor.
3:42
Unidentified speaker: And that's why Vanta's trusted by more than 16,000 companies like Ramp, Harvey, and Writer.
3:48
Unidentified speaker: So prove you're ready for business and get $1,000 off Vanta.
3:52
Unidentified speaker: When you go to Vanta.com forward slash 20 VC, that's vanta.com slash 20 VC.
3:58
Unidentified speaker: You have now arrived at your destination.
4:01
Unidentified speaker: Eric, listen, dude, it is so great to have you on the show.
4:04
Unidentified speaker: You know, I was sitting downstairs with you chatting before, and I was just like, you can feel when it's gonna be a good show because people are themselves.
4:12
Unidentified speaker: And that was an amazing conversation, and we're gonna have a lot of fun.
4:15
Unidentified speaker: So thank you for joining me.
4:17
Unidentified speaker: It's so exciting to be here.
4:18
Unidentified speaker: I mean, I've listened to so many episodes.
4:19
Unidentified speaker: It's it's kind of surreal to be inside of it right now.
4:22
Unidentified speaker: That's really awesome.
4:23
Unidentified speaker: As I said to you, you you could have texted me when you started the company and maybe I would have been a little bit richer.
4:28
Unidentified speaker: Yeah, but you probably didn't know about me then.
4:30
Unidentified speaker: No, I didn't, but I would have welcomed investing in another suite.
4:34
Unidentified speaker: I wanted to start with, I was talking to Zach Frankel last night, and he said about your childhood and like early indicators of success and like exceptional ability.
4:44
Unidentified speaker: How did you fall in love with technology?
4:46
Unidentified speaker: And for me as an investor, what would have made me excited in seeing your childhood?
4:50
Unidentified speaker: I started coding 1992 or something like that.
4:54
Unidentified speaker: I was eight and just loved it.
4:55
Unidentified speaker: You know, it's like one of those things.
4:57
Unidentified speaker: Like I realized you can control computers.
4:59
Unidentified speaker: I had a Mac Plus, you know, one megabyte of RAM or whatever.
5:02
Unidentified speaker: Um and yeah, started making games and like, I mean, they were terrible, but you know, it's just I think back at like in a way, I'm like so grateful.
5:09
Unidentified speaker: I look back at like my childhood.
5:11
Unidentified speaker: My friends, they were like collecting hockey carts or whatever.
5:14
Unidentified speaker: That was like their hobby.
5:15
Unidentified speaker: My hobby was programming.
5:16
Unidentified speaker: And now I make a lot of money from it.
5:18
Unidentified speaker: Isn't that kind of crazy?
5:19
Unidentified speaker: Like, you know, you know when you think about it, like the fact that I can make a lot of money, like doing what what was my hobby since I was eight?
5:24
Unidentified speaker: It's kind of nuts.
5:25
Unidentified speaker: So I I I'm just so grateful about you know, the fact that I fell in love with technology and coding.
5:29
Unidentified speaker: And then, you know, I started doing programming competitions when I was 16, 17, and so much fun.
5:35
Unidentified speaker: Modal was your first company or is your first company.
5:38
Unidentified speaker: Why did it take so long?
5:40
Unidentified speaker: I think the consistent pattern in my life is like I never had quite enough confidence in myself.
5:46
Unidentified speaker: And a lot of it was that was missing was just the confidence, a little bit the money too.
5:50
Unidentified speaker: But uh so it wasn't until I was like 36, 37 I started modal.
5:54
Unidentified speaker: And I think I don't think I've ever wanted to necessarily start a company.
5:58
Unidentified speaker: It was like one of those things where like I just saw this big opportunity in the market.
6:01
Unidentified speaker: I almost thought to myself, like, why is no one else doing this?
6:04
Unidentified speaker: It's such an obvious idea.
6:05
Unidentified speaker: And that led me to think like maybe I should start a company.
6:08
Unidentified speaker: And then, you know, five years later, here we are.
6:10
Unidentified speaker: It's gone pretty well.
6:11
Unidentified speaker: I think in hindsight, maybe I should have started it much earlier.
6:14
Unidentified speaker: Do you wish you had?
6:15
Unidentified speaker: I look at some of these like Stanford kids and they're like 21, and like I kind of feel like they maybe have too much confidence.
6:21
Unidentified speaker: You know what I mean?
6:22
Unidentified speaker: Like they're like, you know, think they're gonna take over the world and they're 21.
6:24
Unidentified speaker: I think the ideal age is maybe like, I don't know, 32 or something like that.
6:29
Unidentified speaker: You have a little bit more experience, you know how to navigate people, you know how to hire people, you know how to fire people, like all these things.
6:35
Unidentified speaker: I don't know if I would have been a good founder at like in my 20s.
6:37
Unidentified speaker: Before we get into the market and everything beneath that, we were chatting downstairs.
6:41
Unidentified speaker: We often hear that founding companies is like eating glass.
6:44
Unidentified speaker: It's hard, it's painful.
6:46
Unidentified speaker: How do you feel about those?
6:48
Unidentified speaker: Dude, I hate that stuff.
6:49
Unidentified speaker: It's like, I don't get it.
6:50
Unidentified speaker: Like you're eating glass like every day.
6:52
Unidentified speaker: Like, why are you doing it then?
6:53
Unidentified speaker: Like, I don't know.
6:53
Unidentified speaker: You expect some like magic reward, like some big cookie in the end.
6:56
Unidentified speaker: Like, I never understood that mindset.
6:58
Unidentified speaker: I think it's like this mindset that a lot of people have.
7:00
Unidentified speaker: I'm gonna suffer pain and then there's gonna be this big reward in the end.
7:03
Unidentified speaker: I don't know.
7:04
Unidentified speaker: I don't think about it that way.
7:05
Unidentified speaker: I enjoy every day I go to work.
7:06
Unidentified speaker: Of course, there's like hard times.
7:08
Unidentified speaker: Like there's, you know, you have to have hard conversations, maybe fire people, maybe you like do hard things.
7:12
Unidentified speaker: But like in the grand scheme of things, like I mean, such is life, and I enjoy life.
7:15
Unidentified speaker: I I enjoy building a company.
7:17
Unidentified speaker: And I don't I don't think of it as like one day we're gonna flip this company and be I I hope to work on this for 20, 30, 40, 50 years.
7:24
Unidentified speaker: I'm having so much fun.
7:25
Unidentified speaker: I don't want to sell this company because I would just go and start Model 2.
7:28
Unidentified speaker: You really wouldn't.
7:29
Unidentified speaker: No, I don't.
7:29
Unidentified speaker: I don't think money necessarily makes you happy.
7:31
Unidentified speaker: I mean, a little bit.
7:32
Unidentified speaker: Like it's good to have some money, but you know, beyond a point.
7:35
Unidentified speaker: Like, what you know, I like building stuff.
7:37
Unidentified speaker: I love technology, I love coding.
7:38
Unidentified speaker: So much fun.
7:39
Unidentified speaker: You're such a Swede.
7:41
Unidentified speaker: Is that Swedish is that Swedish mentality?
7:43
Unidentified speaker: Maybe it is.
7:43
Unidentified speaker: Yeah.
7:44
Unidentified speaker: Never has an American come on before and said it's not about money.
7:47
Unidentified speaker: I mean, I do care about money a little bit, but maybe you're right.
7:49
Unidentified speaker: Because I do think in Swedish mentality, it's like it's like self-actualization.
7:53
Unidentified speaker: Like doing what you enjoy is like kind of the highest form of prestige in a certain way.
7:57
Unidentified speaker: And it could be music, it could be like poetry, whatever.
7:59
Unidentified speaker: Like it almost doesn't matter.
8:00
Unidentified speaker: Aaron Powell What do you do that you don't enjoy?
8:03
Unidentified speaker: Yeah.
8:04
Unidentified speaker: I kind of hate fundraising, honestly.
8:06
Unidentified speaker: I mean it's fine.
8:06
Unidentified speaker: I don't know.
8:07
Unidentified speaker: Like I enjoy it when it's like good people, but it's like kind of challenging.
8:10
Unidentified speaker: You have to say no to a lot of people.
8:11
Unidentified speaker: I guess that's a good problem to have.
8:12
Unidentified speaker: But you know, in general, it's it's oh you have to say no to a lot of people.
8:16
Unidentified speaker: Oh, Eric.
8:17
Unidentified speaker: I know, I know.
8:18
Unidentified speaker: That's a horrible thing to pull it.
8:19
Unidentified speaker: You can only have 10 million and 50 million.
8:22
Unidentified speaker: Oh you've never done a deck, have you?
8:24
Unidentified speaker: Aaron Ross Powell, I've never made a pitch deck, yeah.
8:26
Unidentified speaker: Every round we've done so far, it's I like to get to know people well in advance of raising money.
8:32
Unidentified speaker: So every round we've done, I mean, most recently we announced a D-round yesterday with Excel.
8:37
Unidentified speaker: I've been speaking to Rich and Gonzalo at Excel for I think at least a year.
8:42
Unidentified speaker: And at some point, you know, we're like, actually, you know what?
8:44
Unidentified speaker: We could take some money.
8:46
Unidentified speaker: Like, what are the terms?
8:47
Unidentified speaker: Like, let's discuss terms.
8:48
Unidentified speaker: And then it happened very quickly because we already had that relationship.
8:50
Unidentified speaker: Aaron Ross Powell How do you think about price optimization?
8:53
Unidentified speaker: Is it a game of getting the highest price?
8:55
Unidentified speaker: No, I don't think so.
8:56
Unidentified speaker: I I think maybe in later stages it becomes a little bit more of a commodity like the money itself.
9:00
Unidentified speaker: But I think in the early days, like I mean for for several reasons.
9:03
Unidentified speaker: Like one is that there's a misaligned incentive.
9:05
Unidentified speaker: Like VCs always want to put in way more money than a company needs.
9:08
Unidentified speaker: And that you know, you just take on a lot of dilution.
9:10
Unidentified speaker: It's not good as a shareholder.
9:11
Unidentified speaker: But the other thing is like you actually want to hold back, I think.
9:14
Unidentified speaker: And and it's often better to raise a lot less money at a lower valuation and then keep the VCs hungry for the next round.
9:20
Unidentified speaker: Aaron Powell Is this not the generation though where actually capital is the moat and you need so much more money than you think?
9:26
Unidentified speaker: You said that VCs want to give you more money than the company needs.
9:29
Unidentified speaker: I thought AI was finally the output exhaust for venture dollars that was needed.
9:34
Unidentified speaker: I think there are times in the market where you are capital constrained.
9:38
Unidentified speaker: We have not been capital constrained for most of the period of time of the company.
9:41
Unidentified speaker: We've been extremely capital efficient.
9:43
Unidentified speaker: We burn only a few million a month.
9:45
Unidentified speaker: I mean, only like you know, it's a lot of money.
9:46
Unidentified speaker: I respect for the money.
9:48
Unidentified speaker: No, I mean up until like last year.
9:50
Unidentified speaker: Now we have to spend a lot of money on capacity.
9:52
Unidentified speaker: Aaron Powell How much do you burn a month now?
9:53
Unidentified speaker: Aaron Powell A lot.
9:54
Unidentified speaker: I mean, we we actually are profitable on a PL basis, but like cash flow is quite negative because we have to spend a lot of money on on getting GPUs and building out capacity.
10:01
Unidentified speaker: And so that is a different period in you know in the company and how we think about capital.
10:06
Unidentified speaker: Capital is I don't want to call it emote, because you know, I feel like people throw it on that word like to mean like things that you know kick moat to me has like a very strict definition.
10:14
Unidentified speaker: But it you know certainly helps.
10:16
Unidentified speaker: Capital helps at this point.
10:17
Unidentified speaker: Aaron Powell What's the moat definition for you?
10:19
Unidentified speaker: I mean mote to me is like you go read like seven powers or like one of those things.
10:23
Unidentified speaker: It's a good book.
10:24
Unidentified speaker: But I think you know, then people read that and they get delusional, like you know, they're like, yeah, we're starting this company, it has all these moats.
10:30
Unidentified speaker: You never have moats in the early days.
10:32
Unidentified speaker: Like, you know, like Google could do anything they wanted to.
10:35
Unidentified speaker: Google could do yeah, but it's also like, you know, like what are the seven?
10:38
Unidentified speaker: I don't even remember.
10:39
Unidentified speaker: It's like cornering resources, like market power, like whatever.
10:42
Unidentified speaker: It's like none of those things happen until you're like 10 years in.
10:45
Unidentified speaker: You know?
10:46
Unidentified speaker: There's an old blog post.
10:47
Unidentified speaker: I I always come back to it by Ben Thompson.
10:48
Unidentified speaker: It talks about Snap and the like the gingerbread man strategy.
10:52
Unidentified speaker: Do you know this thing?
10:52
Unidentified speaker: And it's like the whole point is like Snap doesn't have emote, they just run faster.
10:56
Unidentified speaker: Like they're the gingerbread man.
10:57
Unidentified speaker: And I actually always come back to that.
10:59
Unidentified speaker: I think that's so true.
11:00
Unidentified speaker: Like for almost every startup these days, it's like you just have to like move faster than everyone else.
11:04
Unidentified speaker: I actually think like the three things that matter to say when I invest is like one, team.
11:08
Unidentified speaker: Are they a top one percent team?
11:09
Unidentified speaker: Yeah.
11:10
Unidentified speaker: Two, speed of product decision making very specifically.
11:13
Unidentified speaker: You look at like instinct today, the speed of feature shipping is really very impressive and they're very and then three is brand.
11:20
Unidentified speaker: Yeah.
11:20
Unidentified speaker: And I think those are the three determinants to me of defensibility moat, but it's a kind of package.
11:25
Unidentified speaker: Yeah.
11:26
Unidentified speaker: I mean, I again I don't know if I would, but but I kind of agree with your point.
11:28
Unidentified speaker: Like I think brand, I don't want to call it a moat, but it it's it's actually, I think the value of brand has gone up a lot.
11:34
Unidentified speaker: Can I be a dick?
11:35
Unidentified speaker: Yeah, please.
11:36
Unidentified speaker: Do you think mode will have done well with brand?
11:38
Unidentified speaker: I think pretty well.
11:40
Unidentified speaker: I think we have it's coupled with the developer experience and the product and the website and the visual identity and all these things.
11:46
Unidentified speaker: We just did an event in in San Francisco and I thought a lot about that, sort of how we package it, the event and the visual look and feel.
11:52
Unidentified speaker: But like, do you think you have a bigger brand than base 10 and firewise?
11:55
Unidentified speaker: I think we have that.
11:56
Unidentified speaker: I think we have like really, really good vibes.
11:58
Unidentified speaker: Like I'm I'm termally online.
11:59
Unidentified speaker: Like people tweet like way more about modal.
12:02
Unidentified speaker: I think there's like a little bit more of a community.
12:03
Unidentified speaker: I think there's like much better vibes around us.
12:05
Unidentified speaker: Like, but like I am paranoid about those companies.
12:08
Unidentified speaker: Like, I am paranoid about competition.
12:09
Unidentified speaker: Of course, I track it.
12:10
Unidentified speaker: Like, I quite of course, like I I see everything that gets tweeted and about our competitors.
12:14
Unidentified speaker: And this is important.
12:16
Unidentified speaker: Like everyone wants to be a part of the community.
12:18
Unidentified speaker: Everyone, you know what I think is like people misunderstand is I I love TypeSafe with Jev.
12:22
Unidentified speaker: Yeah.
12:22
Unidentified speaker: Because like everyone wants to clone Jev.
12:24
Unidentified speaker: No one wants to get up on stage and wear a fucking pink fur coat and just like talk about the Jevolution in an unhinged way.
12:30
Unidentified speaker: That is what brand to me is.
12:32
Unidentified speaker: Like it's like that guy, it's like he's playing the game on a whole different level.
12:35
Unidentified speaker: And I think it's just like so like true about like companies today.
12:39
Unidentified speaker: Like you have to think about that.
12:40
Unidentified speaker: What so can you jump out?
12:41
Unidentified speaker: Why?
12:42
Unidentified speaker: Because he's authentic and different?
12:43
Unidentified speaker: I think so.
12:44
Unidentified speaker: I I think you just have to create a personality.
12:46
Unidentified speaker: You kind of have to like figure out like what is the personality of your company and just be consistent and create like almost like a sense of movement, a sense of like community.
12:53
Unidentified speaker: Like this is something cool.
12:55
Unidentified speaker: And people want to use you because you're one of the cool products out there.
12:58
Unidentified speaker: They're like, have you tried modal?
12:59
Unidentified speaker: It's pretty cool.
13:00
Unidentified speaker: Aaron Powell What has modal done then that Base 10 and Firewalls haven't to build that?
13:04
Unidentified speaker: I think we started with the developer experience.
13:06
Unidentified speaker: So that was and we had a very clear sort of PLG product market fit, product-led growth in the sense that people came bottoms up.
13:12
Unidentified speaker: They just sign up on a product.
13:14
Unidentified speaker: Like I think the consumer background of coming from Spotify is very important in how I thought about building what's now a B2B company, selling tech effectively.
13:21
Unidentified speaker: You know, it's like a very boring category, but I think you know, like having personality and having a little bit of an edge is a big part of how you create that sort of movement and that sort of like sense of excitement and the feeling that you're different.
13:35
Unidentified speaker: And I want it to start the moment you go to a website.
13:38
Unidentified speaker: And the moment you try the product and you just feel that it's different.
13:41
Unidentified speaker: The magic moment.
13:42
Unidentified speaker: Aaron Powell Can I ask you?
13:43
Unidentified speaker: We've spoken about the three different providers there.
13:46
Unidentified speaker: Inference as a market has many different speculators commenting on it.
13:50
Unidentified speaker: When you look at the next five years, how does the inference market play out?
13:55
Unidentified speaker: Aaron Powell First of all, I think the inference is going to be an enormously large market.
13:58
Unidentified speaker: It's like the new cloud, basically.
13:59
Unidentified speaker: But I don't think inference in itself is maybe like the only thing you need.
14:03
Unidentified speaker: One of the things that I feel very strongly about is like it has to come with like a lot of other products.
14:07
Unidentified speaker: It has to come with post-training, sandboxes, agentic stuff, all kinds of other types of inference, like low latency inference, batch inference.
14:16
Unidentified speaker: So building that platform.
14:17
Unidentified speaker: By the way, we've been working on this for the last five years.
14:19
Unidentified speaker: I think that to me is you know not a moat, but it's a very hard product to build that infrastructure.
14:25
Unidentified speaker: Why do you need to have the breadth of that product set?
14:27
Unidentified speaker: Otherwise, you're just in a compression layer.
14:30
Unidentified speaker: Trevor Burrus, Jr.: Yeah, because otherwise you're just selling commodity product.
14:32
Unidentified speaker: And like if you're just selling like tokens for dollars, there's no switching cost.
14:36
Unidentified speaker: You know, people can just switch a different provider.
14:38
Unidentified speaker: It's just a race to the bottom.
14:39
Unidentified speaker: But if you couple it with all these other things around it, sandboxes and post-training and reinforcement learning and all these things, like you you can actually create a whole set of tools.
14:48
Unidentified speaker: Like no one wants to build a, you know, buy a single-trick pony.
14:50
Unidentified speaker: People want to buy a portfolio of different things.
14:53
Unidentified speaker: Like they want to buy a whole suite of tools that basically lets them build amazing AI applications on top of it, right?
14:59
Unidentified speaker: And the in the infrastructure underneath is a lot of differentiation, a lot of you know staying power in the markets.
15:05
Unidentified speaker: It's very hard to copy the infrastructure required to build all of these products.
15:09
Unidentified speaker: What does no one see about the inference market, or you don't see commented on much that is very obvious to you?
15:15
Unidentified speaker: Aaron Ross Powell One thing that people don't see about the inference market is I actually think in the long term there is so much gross margin opportunity in this space because you can build all these things around it.
15:23
Unidentified speaker: All the sandboxes, all the agentic stuff, workflows, post-training, reinforcement learning.
15:28
Unidentified speaker: If you do all those things, there's actually so much opportunity to make money.
15:32
Unidentified speaker: Aaron Powell Is it a market that has software margins?
15:34
Unidentified speaker: Because we're looking at margins now in the 25 to 40 percent range.
15:38
Unidentified speaker: That's kind of traditional players.
15:40
Unidentified speaker: I don't know if that's yours, but that's kind of what inference providers are generally.
15:43
Unidentified speaker: Do they go to 70 as you earn more and more of the stack?
15:46
Unidentified speaker: I think basically almost no company today has software margins anymore.
15:50
Unidentified speaker: Like what you used to see in like Silicon Valley in the you know 10 years ago with this like 90% margins plus, I think that's basically gone.
15:57
Unidentified speaker: And the reason is actually almost like topological.
15:59
Unidentified speaker: It's like you have this like stack of vendors, right?
16:01
Unidentified speaker: Like today you have applications, you have models, you have infrastructure, you have cloud, you have NVIDIA.
16:07
Unidentified speaker: And so when a customer pays a dollar at the top of the stack, it has to flow through a lot of different layers.
16:12
Unidentified speaker: And at every layer, there's going to be 20%, 30% gross margins.
16:16
Unidentified speaker: So I think there's a lot of margin opportunity, and there's a lot of opportunity to make a lot of money.
16:20
Unidentified speaker: I mean, the space is so big, but you're also splitting it across a lot of layers in the stack.
16:24
Unidentified speaker: So I mean, you know, like the the 90% softer margin businesses are gone.
16:28
Unidentified speaker: I I think Modal has an opportunity long-term to get to 40%, 50%, 60% margins, but we're never going to get to softer margins.
16:35
Unidentified speaker: Aaron Powell What's a wild prediction you have on the future of the inference market that you think many people would disagree with?
16:40
Unidentified speaker: Aaron Powell I think one prediction about the inference market is going to split into a lot of different things.
16:44
Unidentified speaker: Because the market got so big in the last I don't think anyone expected, me included.
16:48
Unidentified speaker: I don't think anyone expected it to be so big.
16:51
Unidentified speaker: It's like 10x or 100x bigger than I ever thought.
16:53
Unidentified speaker: And that to me means there's going to be so much room for specialization, low latency, high latency, like various types of fine-tuning and post-training.
17:00
Unidentified speaker: I think it's just going to be some wacky new technologies that let you do it with way less resources, new hardware, like all kinds of different things.
17:07
Unidentified speaker: I'm really sorry.
17:07
Unidentified speaker: Does that not go against what you just said though about the need to own it all and have that breadth of product?
17:12
Unidentified speaker: If it's so much larger that you need real specialization.
17:15
Unidentified speaker: Specialization by product.
17:16
Unidentified speaker: I think it's going to fragment into a lot of different products.
17:19
Unidentified speaker: Now, we can build a lot of those products and host it all under modal.
17:22
Unidentified speaker: And that to me is very exciting because a lot of different, you know, there's going to be a lot of different products that companies are going to need.
17:28
Unidentified speaker: For one product, they're going to need low latency inference.
17:30
Unidentified speaker: For one product, it's going to they're going to need batch inference.
17:32
Unidentified speaker: For one product, they're going to need sandboxes, which we have.
17:34
Unidentified speaker: For one product, they're going to do pre-training their own models.
17:37
Unidentified speaker: You know, maybe it's uh you know certain specialized audio models or whatever.
17:41
Unidentified speaker: Like we do a lot of stuff besides LM inference.
17:43
Unidentified speaker: And so that to me creates a lot of opportunity to basically kind of rethink what a cloud looks like in the era of AI and kind of deconstruct the cloud and rebuild it for this new era of software.
17:54
Unidentified speaker: Aaron Powell Earlier you said about building capacity with GPUs.
17:58
Unidentified speaker: What's it like doing that today?
18:00
Unidentified speaker: Is it as crazy as reports suggest?
18:03
Unidentified speaker: Aaron Powell So it is pretty crazy.
18:04
Unidentified speaker: I spent some time in March at GTC and Video Speak conference in San Jose.
18:08
Unidentified speaker: I would describe that as like an open-air drug market, but GPUs instead of drugs.
18:13
Unidentified speaker: You're running around, you know, talking to 500 different cloud vendors.
18:16
Unidentified speaker: You're like, do you have GPUs?
18:18
Unidentified speaker: Like please give it and like you know in the end, you come out, like you do a lot of deals.
18:21
Unidentified speaker: And that's something we've had to develop the muscle.
18:24
Unidentified speaker: I think part of the value prop of modal is like our customers don't have to think about that.
18:28
Unidentified speaker: We do the dirty work of doing all the GPU deals under the hood.
18:31
Unidentified speaker: We work with 25 different neo clouds.
18:34
Unidentified speaker: Right now we control like 0.1% or whatever, I don't even know the number of like the world's compute.
18:39
Unidentified speaker: But like my like crazy like, you know, how I think about the long-term goal is like as we control more and more of the world's compute, you can aggregate that capacity and aggregate demand on the other side and manage that capacity and demand extremely efficiently.
18:52
Unidentified speaker: So that to me is like something we actually think we can do, we can fill a big gap.
18:55
Unidentified speaker: What does that allow you to do being that aggregation layer?
18:58
Unidentified speaker: Well so we built a lot of our infrastructure ourselves.
19:00
Unidentified speaker: Like we we we don't use Kubernetes and Docker and things like that.
19:03
Unidentified speaker: We built our own file system, we built our own container runtime, we built our own scheduler, we built our own basically how we manage the capacity.
19:09
Unidentified speaker: And that that enables us to basically treat like all of the world's compute as one big pool of capacity.
19:14
Unidentified speaker: And so if you think about the fact that there's we we already have many thousand paying customers, and over time it just grows and grows.
19:20
Unidentified speaker: Each of those customers have like very sort of variable requirements or like their their uh their workloads kind of vary over the day and they have maybe bursts, and each individually the like very sort of unpredictable.
19:31
Unidentified speaker: But if you pool all of those customers together, you end up with a very sort of smooth overall demand.
19:37
Unidentified speaker: And that is something we can manage extremely efficiently.
19:39
Unidentified speaker: And so one of the valid props of modal, besides like we we can scale up with you, is also you don't even have to do capacity planning.
19:45
Unidentified speaker: Just come to us, we'll give you 100 GPUs.
19:47
Unidentified speaker: I think it's so amazing going back to TypeSafe, for instance.
19:50
Unidentified speaker: Like they we had talked to them uh, I think a long time ago before they launched, but they launched overnight on Modal, and we we weren't even aware of it.
19:56
Unidentified speaker: They just like scale up and we gave them a ton of GPUs and it just happened automatically.
20:01
Unidentified speaker: So that magic, I think, is something that is people really evaluate with modal.
20:05
Unidentified speaker: It's like we can give them a lot of GPUs on very short notice.
20:07
Unidentified speaker: And we can handle that sort of variable.
20:09
Unidentified speaker: Like if you go to us, you don't have to lock up, you know, like a rectangular block of GPUs over three years, and you have to know the exact number.
20:16
Unidentified speaker: Just pay as you go.
20:17
Unidentified speaker: You scale up and down, you pay per GPU second.
20:19
Unidentified speaker: Aaron Powell How do you do GPU buying in terms of forecasting then?
20:23
Unidentified speaker: Because you don't want to have an insane excess of supply just sitting on your books.
20:27
Unidentified speaker: Totally.
20:28
Unidentified speaker: The GPU lead times right now is about three to four months.
20:31
Unidentified speaker: So we basically do a straight line or exponential line extrapolation and we look at in aggregate across all our customers.
20:36
Unidentified speaker: We have a very diversified set of customers.
20:38
Unidentified speaker: You can actually look at the demand in aggregate pretty well and predict based on monthly growth rates, which has been about 30, 40 percent month to month for the last 12 months.
20:46
Unidentified speaker: And you just project that, you know, three to four months into the future, you end up with like two to three X.
20:50
Unidentified speaker: You go out and you know buy that capacity and make sure to ramp up.
20:53
Unidentified speaker: And sometimes can you just buy that capacity?
20:55
Unidentified speaker: Aaron Powell You can, yeah.
20:56
Unidentified speaker: I mean, I think there's a little bit of like a misconception that like we're out of capacity.
21:00
Unidentified speaker: I mean, and videos like shipping like an insane amount of GPUs like every month.
21:03
Unidentified speaker: I don't know how many hundred thousand black holes are getting installed.
21:06
Unidentified speaker: Prices are up for sure.
21:07
Unidentified speaker: You have to pay more to get the same amount of GPUs.
21:10
Unidentified speaker: And I don't want to say it's like an efficient market, but if you're willing to pay more than other people, you can get a lot of GPUs.
21:15
Unidentified speaker: Aaron Powell Is there a price at which it doesn't make sense for you?
21:18
Unidentified speaker: Of course there is, but like at what point do we don't think so.
21:20
Unidentified speaker: I think in fact like we're actually more long than we're short GPUs, because you know we we can charge a higher price to our customers.
21:25
Unidentified speaker: Like there's effectively a market price and it's like different dynamics because we lock up longer-term reservations and rent them out on slightly shorter duration.
21:32
Unidentified speaker: But you know, if if the underlying price goes up, people are also willing to pay more.
21:36
Unidentified speaker: Aaron Powell What happens to the price over time?
21:38
Unidentified speaker: H-100s were meant to depreciate and go down.
21:42
Unidentified speaker: Yeah.
21:42
Unidentified speaker: I mean and like I mean, actually, like it's funny.
21:44
Unidentified speaker: Well, dude, if you have no idea, what am I meant to have?
21:48
Unidentified speaker: Did you see Jensen yesterday?
21:49
Unidentified speaker: He was in London, actually.
21:50
Unidentified speaker: Someone asked him this question.
21:51
Unidentified speaker: He's like, well, obviously like supply is going to catch up with demand eventually.
21:54
Unidentified speaker: And then we'll see what happens with pricing.
21:56
Unidentified speaker: It might happen in a year, it might happen within five years.
21:58
Unidentified speaker: He said exactly the same thing I always tell people.
22:00
Unidentified speaker: Like hardware prices, like it's like gravity.
22:02
Unidentified speaker: Hardware prices always go down in the very long run.
22:05
Unidentified speaker: Right now we're in a very, you know, there there's a supply shortage.
22:08
Unidentified speaker: And you know, that means that the prices have gone up.
22:10
Unidentified speaker: They may continue to go up for another year, two years, five years.
22:13
Unidentified speaker: I don't know.
22:14
Unidentified speaker: But eventually it will come down.
22:15
Unidentified speaker: Aaron Ross Powell, Jr.
22:15
Unidentified speaker: How early are we?
22:16
Unidentified speaker: Because like you know, I run Instinct for everything.
22:19
Unidentified speaker: I have over a hundred automations on Instinct doing uh things that are just absolutely wild from buying concert tickets to planning shows to helping with like medication for my mother.
22:28
Unidentified speaker: The amount of compute I am using is Javin's Paradox times 100.
22:33
Unidentified speaker: Yeah.
22:33
Unidentified speaker: If that is equivalent for actually many people in society, because it's accessible within WhatsApp and what how early are we though?
22:41
Unidentified speaker: I think we're very early.
22:42
Unidentified speaker: I think you're saying the same thing.
22:43
Unidentified speaker: I mean, you know, Jevin's power, whatever.
22:45
Unidentified speaker: Like I think there's just so much latent demand for these things.
22:48
Unidentified speaker: And you know, I I think, you know, eventually we're gonna start to run up against new capital constraints, right?
22:53
Unidentified speaker: Like there's interest rates are going up, like we just can't build data centers fast enough.
22:57
Unidentified speaker: Maybe that's good.
22:58
Unidentified speaker: I don't know.
22:58
Unidentified speaker: Like I think in in a way, like, you know, it it's a guarantee against like overbuild, which but I also don't know if we can overbuild these things.
23:04
Unidentified speaker: There's just so much demand for it.
23:06
Unidentified speaker: Aaron Powell, I do think you can overbuild in the next five years.
23:08
Unidentified speaker: Probably not.
23:09
Unidentified speaker: No.
23:09
Unidentified speaker: I think I think what could happen is like people realize you know instinct is a good example.
23:13
Unidentified speaker: Like allegedly they don't you know charge or they don't make revenue yet.
23:17
Unidentified speaker: So I think it's a good example.
23:18
Unidentified speaker: But like clearly instinct can make a ton of money in the future.
23:22
Unidentified speaker: But so you know, that's a dynamic in the market.
23:24
Unidentified speaker: There's a lot of companies that are not yet making money yet and they're using a lot of deputes.
23:28
Unidentified speaker: They just have to eventually start to make money.
23:30
Unidentified speaker: But like that that lag could take a while, right?
23:32
Unidentified speaker: And and that we might run up against that.
23:33
Unidentified speaker: Trevor Burrus, Jr.
23:34
Unidentified speaker: You mentioned the the specialization that comes within the inference market, and we we spoke about NVIDIA there.
23:38
Unidentified speaker: We're seeing more and more specialization within the chip market.
23:41
Unidentified speaker: Is NVIDIA's business today stronger than ever or actually weaker than ever as you see the fragmentation of the market into very specialized players from Cerebrus to etched to OLIX to fractile that weren't there before?
23:55
Unidentified speaker: I think you're uh uh trying to phrase it as like one versus the other.
23:59
Unidentified speaker: My view is like I think uh Eric Vischer was actually talking it's funny, like the same thing.
24:03
Unidentified speaker: Like all of these things are gonna grow.
24:05
Unidentified speaker: Like I think NVIDIA is gonna keep growing, they're gonna keep shipping amazing chips.
24:08
Unidentified speaker: I think there's also gonna be like Cerebrus and all these etched and all these other ones like shipping specialized chips, and they're also gonna grow.
24:14
Unidentified speaker: So I think there's gonna be markets for all of it.
24:15
Unidentified speaker: Aaron Powell So why is Cerebrus down 30%?
24:18
Unidentified speaker: I don't know.
24:18
Unidentified speaker: I don't follow really the stock market.
24:20
Unidentified speaker: Aaron Ross Powell You don't follow the stock market.
24:21
Unidentified speaker: No.
24:21
Unidentified speaker: I I'm in like a very passive, you know, buy the index.
24:24
Unidentified speaker: Mostly because of my own time.
24:26
Unidentified speaker: Like I don't feel like I have the energy to follow the stock market.
24:28
Unidentified speaker: Aaron Powell To what extent is CUDA the moat that everyone says it is for NBI?
24:31
Unidentified speaker: I actually I think it's a little overstated.
24:33
Unidentified speaker: I don't think CUDA itself is the moat.
24:35
Unidentified speaker: CUDA is actually kind of annoying to work with.
24:37
Unidentified speaker: It's like a pretty annoying programming model.
24:39
Unidentified speaker: But I do think all the NVIDIA the ecosystem is.
24:41
Unidentified speaker: Like I think you know, all the kernel optimizations and and all the stuff that, you know, every every like Torch always supports NVIDIA much better than anything else.
24:51
Unidentified speaker: And and the open source engines like VLM and STLang because of that, they always support it much better.
24:55
Unidentified speaker: So like everything is NVIDIA first today.
24:58
Unidentified speaker: And that's probably gonna continue.
24:59
Unidentified speaker: It's like funny for the last like 15 years, I've been saying, you know, like next year is the you know, it's the year of non-NVIDIA accelerators.
25:06
Unidentified speaker: It never keeps happening.
25:07
Unidentified speaker: I'm always wrong.
25:07
Unidentified speaker: Nvidia has just shown to be incredibly good at having that staying power and just dominating the market.
25:13
Unidentified speaker: We'll see.
25:14
Unidentified speaker: I mean, there's like training, there's TPUs, there's AMD, there's a lot of other stuff that does look quite good.
25:19
Unidentified speaker: And they're catching up, but you know, I still think they're quite behind.
25:22
Unidentified speaker: Aaron Powell To what extent are the Neo Clouds a friend?
25:25
Unidentified speaker: Because they're partnered with you in a lot of cases, I'm sure.
25:28
Unidentified speaker: As you mentioned, I think 25, you said, or or a foe.
25:31
Unidentified speaker: You know, I had the founder of Nabius on, and he was like, we're coming for the inference market.
25:35
Unidentified speaker: Like we need to own full stack.
25:37
Unidentified speaker: And I I don't mean that against him, I totally understand it.
25:40
Unidentified speaker: But are they friends or are they foes?
25:42
Unidentified speaker: Aaron Powell, I feel like everyone in startup plan is like a friend and a foe at the same time.
25:45
Unidentified speaker: They have this like weird sort of you know relationship with everyone.
25:48
Unidentified speaker: I I think some of the top neo clouds, Nabius included, they are trying to move up to stack.
25:53
Unidentified speaker: I'm personally somewhat skeptical.
25:54
Unidentified speaker: I think it's like very hard to be good at like your core business and also trying to move into a different market at the same time.
26:00
Unidentified speaker: Aaron Powell Why would it be hard for them?
26:01
Unidentified speaker: Like what's the bear case on them being able to move into the inference layer efficiently?
26:05
Unidentified speaker: I mean, I'm biased, but like we spent the last five years building an incredibly complex infrastructure stack.
26:10
Unidentified speaker: I think it's very, very hard to replicate what we built.
26:11
Unidentified speaker: Aaron Powell What's the hardest thing?
26:13
Unidentified speaker: All the code and you know, everything around the file system, the container runtime, the scheduler, like building the system that can you know place like millions of sandboxes every minute, you know, scale up and down many, many thousands of GPUs, you know, power hundreds of thousands of tasks running concurrently.
26:28
Unidentified speaker: These are these are very hard engineering problems.
26:30
Unidentified speaker: I I mean I'm I'm an infra geek.
26:32
Unidentified speaker: Like I love that shit.
26:33
Unidentified speaker: It's so much fun to work on it, but like these are hard problems.
26:37
Unidentified speaker: Is Nebius a better business than Cool Weave?
26:40
Unidentified speaker: I actually don't know.
26:40
Unidentified speaker: Like I don't know.
26:41
Unidentified speaker: I I think they're both like they've executed extremely well, both of them.
26:44
Unidentified speaker: I have a tremendous respect for what we've done.
26:46
Unidentified speaker: I think at the at the end of the day, like the thing that I have the most respect for is actually more on the financial engineering side.
26:51
Unidentified speaker: They're very good at that kind of stuff.
26:52
Unidentified speaker: Aaron Powell Why do you say that?
26:53
Unidentified speaker: Because I think a lot of the constraint for building a neo cloud very quickly is on the financial side.
26:59
Unidentified speaker: You need to be very good at you go lock in like a long-term customer contract, then you go raise a bunch of money against it, then you go buy the chips, then you deliver the chips, and then you pay off the debt like over the next few years.
27:10
Unidentified speaker: Doing that sort of dance with the financials is kind of an interesting game.
27:14
Unidentified speaker: Uh, and Corviv has done it exceptionally well.
27:16
Unidentified speaker: It's a capital game, it's a financial game.
27:19
Unidentified speaker: When you move a layer beneath them, we spoke about Jeb, who obviously worked with modal, we see incredible movement from the open source community, which has been exciting.
27:28
Unidentified speaker: Yeah.
27:28
Unidentified speaker: But then you also see Haiku and Anthropics newest models being incredibly performant and incredibly cheap.
27:34
Unidentified speaker: What's the benefit of open models if frontier models are almost as cheap and as efficient?
27:41
Unidentified speaker: I think open source is always gonna be there's always gonna be great models coming out.
27:45
Unidentified speaker: Let's also remember like the open source, the open source models, they're all moving into licensing models.
27:49
Unidentified speaker: So like they're getting better at capturing the value of those models.
27:52
Unidentified speaker: And I think that means they're gonna have more funding to train better models.
27:55
Unidentified speaker: And I really hope there's gonna be European and US models, because I don't think it's a good situation that we have so many Chinese models.
28:00
Unidentified speaker: I mean, they're they're obviously I have tremendous respect for all the Chinese labs.
28:03
Unidentified speaker: They're extremely good at what they're doing.
28:04
Unidentified speaker: But I think it's good for the world if there's you know more power spread out.
28:07
Unidentified speaker: But anyway, going back to the question, I do think that open source, there's always gonna be people who want to fine-tune, want to have more control over their models and sort of tweak the parameters, like go low latency, high throughput, like whatever it is.
28:20
Unidentified speaker: And and open source is gonna play a big role in that.
28:22
Unidentified speaker: It's like Windows versus Linux.
28:24
Unidentified speaker: I'm sorry, with the greatest of respect.
28:26
Unidentified speaker: Do you think Europe will produce a competitive open source model?
28:30
Unidentified speaker: Yeah.
28:30
Unidentified speaker: I mean, I I'm I love Europe.
28:32
Unidentified speaker: I grew up in Europe.
28:33
Unidentified speaker: I grew up in Sweden, but you know, I I do think that like Europe does struggle with you know accumulating large amounts of capital.
28:41
Unidentified speaker: I think both of those things are things that Europe struggles with.
28:44
Unidentified speaker: And that is something that China has done really well.
28:46
Unidentified speaker: They've accumulated massive amounts of human capital.
28:48
Unidentified speaker: And US obviously has very good, you know, very good at like raising enormous amounts of money and like training very large models.
28:54
Unidentified speaker: I don't know if Europe has the either of those.
28:56
Unidentified speaker: Aaron Powell Does it matter if NVIDIA owns the open source model layer in the US with first pool side and now reflection?
29:04
Unidentified speaker: Aaron Powell I think Nvidia is very smart to focus on the ecosystem, building up an ecosystem.
29:10
Unidentified speaker: And I think that includes like propping up a lot of open source models.
29:13
Unidentified speaker: And I think they probably saw an opportunity to pick up open source models.
29:18
Unidentified speaker: These companies are very resource intensive, right?
29:20
Unidentified speaker: Very capex intensive.
29:22
Unidentified speaker: And I think it's smart of NVIDIA to like buy some of those and sponsor them effectively in order to create like American-built models.
29:29
Unidentified speaker: By the way, I want to take back like one thing I said.
29:32
Unidentified speaker: Europe does have an insane amount of human capital.
29:34
Unidentified speaker: I do think what Europe is very bad at is like aggregating them and putting all of them at the same company.
29:39
Unidentified speaker: That is something I think US is very good at.
29:41
Unidentified speaker: And I think inspiring a mindset of ambition and huge, huge scale.
29:45
Unidentified speaker: I think to the point that you said earlier, which is like, oh, I didn't have the belief in myself.
29:49
Unidentified speaker: I respectfully think if you're in the US, you might have been surrounded by influences that told you to start a company earlier.
29:55
Unidentified speaker: Yeah, yeah, yeah.
29:56
Unidentified speaker: I mean, it's almost like too much on the other side to that point.
29:58
Unidentified speaker: But when you're four, you're encouraged to start, leave school, drop out, kindergarten wasn't worth it.
30:03
Unidentified speaker: Exactly.
30:03
Unidentified speaker: I I totally agree with you.
30:04
Unidentified speaker: I do want to talk about, you know, we said about models and Java in particular.
30:08
Unidentified speaker: Are we seeing like the unbundling of models on a like per function basis where Java is obviously very good for certain tasks?
30:15
Unidentified speaker: I think so.
30:15
Unidentified speaker: And I think that's a function of what I talked about earlier, where like the market for inference is so large.
30:19
Unidentified speaker: And that, you know, when when a market is like you know a billion dollars, like you're not gonna have too many submarkets.
30:24
Unidentified speaker: But when a market is starting to get 10 billion, 100 billion, whatever, right?
30:27
Unidentified speaker: Like you're gonna have a lot of different submarkets.
30:29
Unidentified speaker: And and JEV does classification and things like that, regression type things very, very well.
30:34
Unidentified speaker: And and there's a large enough market for that.
30:35
Unidentified speaker: Okay.
30:36
Unidentified speaker: Welcome.
30:36
Unidentified speaker: It's your first day as a partner at 20 VC Fund.
30:39
Unidentified speaker: And we're in the investment committee.
30:41
Unidentified speaker: And very kindly we have the opportunity to invest.
30:44
Unidentified speaker: We're all in on Jev's new approach.
30:46
Unidentified speaker: What's what's the bull case for Jeb being worth $100 billion?
30:49
Unidentified speaker: I told you before, but like I think the brand and the community and the movement around, you know, this just the sort of everything they're doing is is so incredible.
30:56
Unidentified speaker: And that to me is like a massive, massive asset in this new age.
31:00
Unidentified speaker: That's a hundred billion dollar company.
31:01
Unidentified speaker: I think there is an opportunity to create a hundred billion dollar company with a lot of specialized models in inference.
31:07
Unidentified speaker: You would have done it though at 8.5 billion.
31:09
Unidentified speaker: I would have done it at 8.5 billion.
31:11
Unidentified speaker: I I think there's a pretty big outcome here.
31:13
Unidentified speaker: Could be an exit, could be, you know, like a public company, who knows?
31:17
Unidentified speaker: But it but I do think there's a lot of upside here.
31:19
Unidentified speaker: Dude, I'd love to see him ringing the bell on the New York Stock Exchange.
31:23
Unidentified speaker: You know what I mean?
31:23
Unidentified speaker: Yeah, fantastic guy.
31:24
Unidentified speaker: Love that guy.
31:25
Unidentified speaker: Is he great?
31:26
Unidentified speaker: I mean, you know, his charisma, his like you have to create like a personality like around your brand.
31:31
Unidentified speaker: You have to create this like visual identity.
31:33
Unidentified speaker: It's obviously not the visual identity I would have picked.
31:36
Unidentified speaker: I hear you, but I think then you should be more loud, if I'm just honest, Eric.
31:39
Unidentified speaker: Moodle?
31:39
Unidentified speaker: No, we well, yeah, but you.
31:41
Unidentified speaker: Yeah, I mean that's possible, right?
31:43
Unidentified speaker: Like I don't think there's like a personality-led cult from Modal from your perspective.
31:47
Unidentified speaker: I think that's a fair point of criticism.
31:48
Unidentified speaker: I think I could be out there even more.
31:50
Unidentified speaker: I try to tweet.
31:51
Unidentified speaker: I try to uh blogging, I've been like kind of bad at that recently.
31:53
Unidentified speaker: No one gives a shit about blogging.
31:55
Unidentified speaker: Yeah, maybe right.
31:55
Unidentified speaker: Maybe I just need to like go more unhinged on Twitter.
31:57
Unidentified speaker: It's like fundra-led marketing, basically.
31:59
Unidentified speaker: Like in the early days of Modal, like that was a big part of our growth, right?
32:02
Unidentified speaker: Like me just like tweeting all the time about what we're building.
32:04
Unidentified speaker: And I think that's really important.
32:05
Unidentified speaker: I mean, every every company has to find its own sort of you know go-to-market strategy.
32:09
Unidentified speaker: But in the early days, like we spent a lot of time just talking to people on Twitter, just like hitting people up, tried a product, what do you think?
32:15
Unidentified speaker: Half of them ended up angel investing.
32:17
Unidentified speaker: You know, they're now supporters of modal.
32:19
Unidentified speaker: And and you know, you have to lean into what you're good at.
32:21
Unidentified speaker: And in the early days, I did a lot of shit posting.
32:24
Unidentified speaker: Do you know for every follower that followed me before 50,000, I would send them a personal, like, thank you message.
32:29
Unidentified speaker: That's amazing.
32:30
Unidentified speaker: You typed it out yourself every night from 1 till 2:30 a.m.
32:33
Unidentified speaker: That's amazing.
32:34
Unidentified speaker: And I'd say, like, I see that you're from Sweden.
32:36
Unidentified speaker: I love the Grand Hotel.
32:38
Unidentified speaker: Let me know when you're in London.
32:39
Unidentified speaker: And I would send like 50 to 100 a night.
32:41
Unidentified speaker: That's education.
32:42
Unidentified speaker: That's amazing.
32:42
Unidentified speaker: No, but it was the customer acquisition flywheel.
32:44
Unidentified speaker: Yeah.
32:45
Unidentified speaker: Like they would go into the office the next day and be like, wow, you know, this guy, he'd send me a message and like, you should listen to a show.
32:50
Unidentified speaker: Yeah.
32:50
Unidentified speaker: It's pretty good.
32:51
Unidentified speaker: It's smart.
32:52
Unidentified speaker: I love what you've been doing in create in terms of building up this like media brand.
32:55
Unidentified speaker: It's it's fantastic.
32:56
Unidentified speaker: Neil Meta has a great thing, which is like, how can I 10x the customer experience?
33:00
Unidentified speaker: And actually, like, let's let's do it.
33:02
Unidentified speaker: Fuck it, let's play a game.
33:03
Unidentified speaker: How do I 10x the current modal customer experience?
33:07
Unidentified speaker: I think the biggest thing we need is we need way more products in in modal.
33:11
Unidentified speaker: Like we want to create a basic comprehensive experience of anything you need to do in order to build an AI application should be a product inside Modal.
33:17
Unidentified speaker: The other thing we can do is hiring way more FDEs and way more sales reps.
33:22
Unidentified speaker: I will say we've been behind on that, but we should hire like way more people that reach out to you personally and say, hey, I saw you get started in Modal.
33:29
Unidentified speaker: I saw you're working in the biospace.
33:30
Unidentified speaker: How can we help you?
33:31
Unidentified speaker: Here's some bio examples, you know, protein folding, whatever, right?
33:35
Unidentified speaker: Like we work with a bunch of other similar companies.
33:36
Unidentified speaker: You seem skeptical.
33:38
Unidentified speaker: Well, no, I don't at all.
33:39
Unidentified speaker: I I'm actually just thinking about like I I like the product ambition, but I go, as a European who goes, wow.
33:46
Unidentified speaker: We're in the B2B space.
33:47
Unidentified speaker: So I think it's like, you know, I think But do you worry about being stretched too thin?
33:50
Unidentified speaker: Yeah.
33:51
Unidentified speaker: To your point on like, you know, why can't Nebius do what you do so efficiently?
33:54
Unidentified speaker: Because you're so good and you spend so much time, as you expand into all the specialities, there are players in each of them who've been doing that for years where I'm like, I think that's true.
34:04
Unidentified speaker: I don't think they're very good, a lot of them.
34:06
Unidentified speaker: But but I mean, I I sort of agree with your higher level premise.
34:09
Unidentified speaker: Like, if there's anything that's like super critical and never goes away, it's prioritization and resource allocation, right?
34:14
Unidentified speaker: Like you can't boil the water and try to do everything at the same time.
34:17
Unidentified speaker: You need to be extremely ruthless.
34:18
Unidentified speaker: And having a multi-product strategy, I think it's a very hard balance.
34:21
Unidentified speaker: Like, when do you branch out into a second product or a third product, right?
34:23
Unidentified speaker: Like it took us four years to see success with a second product, which was sandboxes.
34:28
Unidentified speaker: Four years to see success with your second product.
34:31
Unidentified speaker: But you're five years to a billion in revenue.
34:33
Unidentified speaker: Yeah.
34:33
Unidentified speaker: So we we launched sandboxes in 2023.
34:35
Unidentified speaker: Didn't see a lot of success in the early days.
34:37
Unidentified speaker: I I mean there was like some success, you know, it wasn't horrible, but you know, it was a relatively insignificant part of our revenue up until summer of 2025 when models, I think like Sonic 4.5, one of those, like they just got really good at generating code and building whole apps.
34:52
Unidentified speaker: And then it's it just exploded.
34:54
Unidentified speaker: I mean, sandboxes today is a 300 million run-rate business.
34:57
Unidentified speaker: It's about a third of our revenue.
34:58
Unidentified speaker: And so that's been a kind of a crazy journey.
35:00
Unidentified speaker: And now we're thinking about you know, what's the third and the fourth?
35:02
Unidentified speaker: We just launched a training product.
35:03
Unidentified speaker: I'm super excited about that.
35:04
Unidentified speaker: Working on batch products, working on reinforcement learning.
35:07
Unidentified speaker: We're working on all kinds of other stuff.
35:08
Unidentified speaker: And so that evolution into being becoming a platform, it's actually there.
35:12
Unidentified speaker: Like if you go back to the very early decks of Modal, you know, I put together some just for my own thinking.
35:17
Unidentified speaker: It was all there.
35:18
Unidentified speaker: It's like we we had this like market map and all these things, you know, it ended up being very different products than we thought.
35:22
Unidentified speaker: But it's so important for me to create like, you know, to focus on a category and focus on a set of solutions, not just like be a single, single widget vendor.
35:30
Unidentified speaker: Aaron Powell What do you know about the sandboxes market today that you wish you'd known when you started and released the first product?
35:36
Unidentified speaker: It actually is like one of those products.
35:38
Unidentified speaker: I I think one of the benefits of being an infrastructure, but also one of the drawbacks in a weird way is like you kind of just have to trust your gut on like what are like important primitives and just build like amazing building blocks.
35:49
Unidentified speaker: I think we're in the space of like providing building blocks.
35:51
Unidentified speaker: We try to try to look at like what do engineers need and not like listen too carefully because like they'll ask for like all kinds of random stuff, but kind of distill it down into like what are some of the building blocks.
36:01
Unidentified speaker: And we put out a serverless function product in in 2021, it saw success in 2023 with Gen AI inference.
36:07
Unidentified speaker: We put out a sandbox product in 2023 because we saw a lot of people using LMs to generate code.
36:12
Unidentified speaker: And then it saw success in in 2025, like I said.
36:15
Unidentified speaker: So a lot of like how we think about building products has always been kind of like, let's just put out like good building blocks.
36:20
Unidentified speaker: And then it exploded.
36:21
Unidentified speaker: But like to answer your question, it was reinforcement learning, it was vibe coding, background agents, and maybe some other stuff too in the future.
36:28
Unidentified speaker: Those were things we actually didn't anticipate.
36:30
Unidentified speaker: So I think I think I think infrastructure is often like that.
36:33
Unidentified speaker: Like we didn't anticipate stable diffusion when we started building our serverless function product.
36:37
Unidentified speaker: It just happened to be the killer app.
36:38
Unidentified speaker: Aaron Powell What's your takeaway from that?
36:40
Unidentified speaker: The only thing is that matters is you've got to stay in the game long enough for your timing to be right?
36:46
Unidentified speaker: Sometimes I think that's right.
36:48
Unidentified speaker: I mean, we we've definitely been extremely lucky in certain ways with timing.
36:51
Unidentified speaker: Like we put out products like right before we started the company in 2021, and then you know, Stable Diffusion and ChatGPT came out in 2022.
36:56
Unidentified speaker: We had really good timing with that, right?
36:58
Unidentified speaker: It just created so much demand and we're very well positioned.
37:00
Unidentified speaker: But some of it is also like with infrastructure.
37:02
Unidentified speaker: Like again, like I think it's about like building products that you know you anticipate will be useful in the future.
37:08
Unidentified speaker: And from your own experience are things that you always wanted to see.
37:11
Unidentified speaker: I think you have to kind of trust your gut in that.
37:13
Unidentified speaker: You said that about resource allocation and that being very, very important as you think about this breadth of products.
37:19
Unidentified speaker: Where did you allocate resources where, with the benefit of hindsight, you wish you hadn't?
37:23
Unidentified speaker: And how do you reflect on that?
37:24
Unidentified speaker: I think we should have allocated more resources towards LM inference like earlier.
37:27
Unidentified speaker: I think we're a little bit on on the sidelines for a while.
37:31
Unidentified speaker: We didn't realize how big open source inference would be.
37:33
Unidentified speaker: And we thought it would be kind of a race to the bottom, just a commodity space.
37:36
Unidentified speaker: And then Deep C came out, and we're like, fuck, like we gotta move very fast.
37:40
Unidentified speaker: And so we staffed up a team immediately, you know, went from like, you know, one or two people working on this to like 30 people 12 months later.
37:47
Unidentified speaker: And and now we're one of the top three providers.
37:49
Unidentified speaker: Jason Lamkin, a dear friend of mine who I do a show with every Thursday, he always says that now we've reached peak open source LLM inference usage.
37:58
Unidentified speaker: We're gonna see Frontier providers bluntly provide incredibly cheap products that they are now.
38:04
Unidentified speaker: US enterprises don't want to use Chinese models.
38:06
Unidentified speaker: That will continue.
38:07
Unidentified speaker: And so actually, with those two core prevailing wins, Frontier will actually resurge and we will see a reduction/slash plateauing of the OS LLM inference market.
38:17
Unidentified speaker: I think that's a little bit of a fixed pie size mentality.
38:20
Unidentified speaker: Like I I I think, you know, with with all due respect, I do think both spaces will grow 5x, 10x, 20x over the next 12 months and and beyond.
38:27
Unidentified speaker: Like I and so like it's hard to argue like, you know, which one's gonna be bigger than the other, but I think both will grow like in you know, orders of magnitude over the next few years.
38:34
Unidentified speaker: Aaron Powell 5x, 10x, 20x over the next 12 months.
38:37
Unidentified speaker: Yeah, for sure.
38:38
Unidentified speaker: You just hit a billion in revenue.
38:39
Unidentified speaker: Yeah.
38:40
Unidentified speaker: Well done.
38:40
Unidentified speaker: Yeah.
38:41
Unidentified speaker: Does that feel pretty cool?
38:42
Unidentified speaker: Aaron Powell It feels pretty cool.
38:43
Unidentified speaker: It's you know, like I I don't know.
38:45
Unidentified speaker: Like I'm this like you know, Swedish guy that you know grew up in a random For me, it's like these numbers are absurd, but like, yeah, like it feels great.
38:53
Unidentified speaker: It's so much money.
38:55
Unidentified speaker: I wasn't expecting you to be so likeable and cool.
38:59
Unidentified speaker: I don't know.
39:00
Unidentified speaker: Like it's like a lot of zeros.
39:01
Unidentified speaker: Like how many zeros are in a billion?
39:02
Unidentified speaker: It's it's a lot.
39:03
Unidentified speaker: Your wealth manager will tell you.
39:06
Unidentified speaker: When you look at that, where will you be in 12 months?
39:09
Unidentified speaker: I don't know.
39:10
Unidentified speaker: What I can say is we've been growing, we grew 15x in the last 12 months.
39:14
Unidentified speaker: So you can extrapolate that.
39:15
Unidentified speaker: I mean, I don't know.
39:16
Unidentified speaker: Like at some point we start, you know, like markets just get very big and they start to slow down a little bit.
39:20
Unidentified speaker: But um with the growth rate right now, like the growth rates are not slowing down.
39:24
Unidentified speaker: Like we're growing 30, 40 percent every month.
39:26
Unidentified speaker: 4 billion?
39:26
Unidentified speaker: I hope to be way beyond that.
39:28
Unidentified speaker: Aaron Powell With the greatest of respects, 15 billion then feels like quite a cheap price.
39:33
Unidentified speaker: Like if I'm an investor, so this is now my ballpark.
39:35
Unidentified speaker: If you think you're gonna be at 4 billion, say dropping, I don't know, a billion two to the bottom line, then they actually only paid what, 14x end of year, next year bottom line?
39:46
Unidentified speaker: Three X top line.
39:47
Unidentified speaker: Yeah, it's kind of cheap.
39:48
Unidentified speaker: That feels pretty cheap.
39:49
Unidentified speaker: I think it's cheap.
39:50
Unidentified speaker: I mean, like I wouldn't do this if I didn't think we could grow and become like a generational company.
39:53
Unidentified speaker: I think there's so much upside.
39:55
Unidentified speaker: I think this is one of the largest markets opportunities that's ever existed.
39:58
Unidentified speaker: I think the the demand for compute, the demand for AI applications, it's just gonna it's gonna be enormous.
40:03
Unidentified speaker: And they're all gonna need a place to run.
40:05
Unidentified speaker: And I hope a lot of them are gonna run a model.
40:07
Unidentified speaker: Aaron Powell What revenue line do you not have today or is tiny that you think will be very significant in three years' time, five years' time?
40:14
Unidentified speaker: Aaron Powell That is a good question.
40:16
Unidentified speaker: I think training is something we just launched.
40:18
Unidentified speaker: I think that could be very large.
40:19
Unidentified speaker: I hope that could get to a hundred million run rate in the next six to twelve months, post-training similarly.
40:24
Unidentified speaker: Uh working on batch jobs.
40:26
Unidentified speaker: I'm very excited also about a lot of like sort of traditional like data warehousing stuff, databases, workflow orchestration, feature stores, like vector database.
40:34
Unidentified speaker: I mean, there's just so many parts of this.
40:36
Unidentified speaker: And they all kind of fit together very nicely.
40:38
Unidentified speaker: And that that to me is like what I'm so excited about.
40:40
Unidentified speaker: It's like once you start building in this space, there's so many natural adjacencies.
40:43
Unidentified speaker: Like this is like a problem with me.
40:45
Unidentified speaker: It's like I get excited about so I have like no lack of ideas.
40:48
Unidentified speaker: Like at Modo, like we have so many ideas of things we want to go.
40:50
Unidentified speaker: I would love to get into continuous integration.
40:52
Unidentified speaker: That's like a very interesting space.
40:54
Unidentified speaker: Agent harnesses, maybe we should train models.
40:57
Unidentified speaker: Maybe you know, there's just 5,000 directions, you know, and that's like my ADD brain like wants to run in all of those directions.
41:03
Unidentified speaker: You have to be obviously ruthless with prioritization and only take on like one or two things at it at the right time.
41:07
Unidentified speaker: But but there's so many different directions we can go in.
41:10
Unidentified speaker: It's amazing that your wife calls you for decisions.
41:13
Unidentified speaker: She must get a very incoherent answer.
41:15
Unidentified speaker: I know, I know.
41:16
Unidentified speaker: But is this, is this, is this I just wanted some curtains done.
41:19
Unidentified speaker: Yeah, you know, with the curtains, I just like you pick one of them.
41:21
Unidentified speaker: I've seen Simpsons and like Rainer Wolfcastle becomes the president, and they're like, which one do you want to pick?
41:26
Unidentified speaker: You know, one to seven.
41:28
Unidentified speaker: He's like, I was born to lead, not to read.
41:31
Unidentified speaker: Number four.
41:33
Unidentified speaker: Yeah.
41:33
Unidentified speaker: Can I ask?
41:34
Unidentified speaker: We've seen this real shift, I think, in in emphasis, but also in resource allocation from pre-training to post-training.
41:41
Unidentified speaker: How do you feel about that?
41:42
Unidentified speaker: And do you think that is the new normal?
41:45
Unidentified speaker: I don't know.
41:45
Unidentified speaker: I I think post-training works because there is a lot of tasks that people do in real life that have fairly finite sort of feedback mechanisms.
41:54
Unidentified speaker: So coding is like a great example, right?
41:56
Unidentified speaker: Like you can do coding with reinforcement learning pretty well because you know you can have it can you can run code in sandboxes.
42:01
Unidentified speaker: And a lot of labs are running modal sandboxes for that exact purpose.
42:05
Unidentified speaker: You start to look at cybersecurity as another area.
42:07
Unidentified speaker: Math, obviously, with lean.
42:08
Unidentified speaker: I mean, we've seen some fantastic project.
42:10
Unidentified speaker: I think it's kind of incredible, all these like open AI proofs the other day.
42:13
Unidentified speaker: So law, maybe to some extent, you can you know do like eval.
42:17
Unidentified speaker: So I I think I think post-training right now, like there's like such an incredible focus on it because you have all these tasks that traditionally only humans can do, but are actually well within reach for agents and for LLMs.
42:28
Unidentified speaker: But like I don't know if it extends beyond that.
42:30
Unidentified speaker: Like I don't know.
42:30
Unidentified speaker: Like there's a lot of tasks out there that are like have way longer feedback loops that I don't know if like post-training quite yet are is within reach.
42:38
Unidentified speaker: A lot of people ask me like what's gonna happen with software engineers in the long term.
42:41
Unidentified speaker: And I think you know, we're probably gonna automate everything that an agent can do within 24 hours or a week or something like that.
42:47
Unidentified speaker: But I kind of feel like you know, engineers like they they work on a longer timescale and they have feedback loops that are longer.
42:52
Unidentified speaker: So I have a feeling like reinforcement learning, we're gonna start to run into some limitations around that.
42:56
Unidentified speaker: But I've been wrong for so long, so maybe I'll be wrong and just keeps going.
43:00
Unidentified speaker: And eventually we can re-do reinforcement learning on things that take six months for a human.
43:04
Unidentified speaker: I don't know.
43:04
Unidentified speaker: Aaron Powell What does the day look like in three years for an engineer?
43:07
Unidentified speaker: I mean, I'm extremely bullish on software engineering.
43:09
Unidentified speaker: I've been coding for 30 years and I've seen so many different changes with like programming languages and the cloud and you know, any so many different frameworks.
43:17
Unidentified speaker: And every single time we make programmers more efficient, it just increases demand for software.
43:22
Unidentified speaker: There's so many parts of the world that desperately need software.
43:24
Unidentified speaker: And so I'm very, very bullish on software and the need for software engineering.
43:28
Unidentified speaker: And I think one of the big biggest misconceptions from the outside is that developer is about like translating from English to code.
43:34
Unidentified speaker: Like that has never been in the job of software engineers.
43:37
Unidentified speaker: It's like to understand the requirements in the first place and be kind of an interface between the humans you're building for humans at the end of the day, and like trying to figure out like what does that even mean in terms of processes and workflows.
43:47
Unidentified speaker: And that continuously moves up in the stack.
43:49
Unidentified speaker: You know, uh back in the 60s, we used to write assembly and punch cards, and then it moved to COBOL and then to C and then to Python.
43:56
Unidentified speaker: And now we're using agents.
43:57
Unidentified speaker: And that's just like one like it's definitely like the biggest thing that I've seen in the last 30 years.
44:02
Unidentified speaker: But I do think we have many, many analogs and many parallels throughout the the history of software and computers.
44:08
Unidentified speaker: To me, that makes me feel like you know, software engineers are gonna be fine for the next 10, 20 years.
44:12
Unidentified speaker: How much money do you spend on coding agents a month at Modal?
44:15
Unidentified speaker: I think a half a million or something like that.
44:17
Unidentified speaker: How many engineers do you have?
44:18
Unidentified speaker: 150?
44:19
Unidentified speaker: Dude, that's so that's so worryingly though.
44:21
Unidentified speaker: You want to should be token maxing more?
44:25
Unidentified speaker: Yeah, maybe.
44:25
Unidentified speaker: I don't know.
44:26
Unidentified speaker: Brandon uh McCall um that they spend three X their human salaries on inference.
44:32
Unidentified speaker: Yeah, we're not quite there yet.
44:33
Unidentified speaker: But I will say like our software engineers are extremely agent pills, including myself.
44:36
Unidentified speaker: I mean, I I write code with agents like all the time.
44:38
Unidentified speaker: But I think like the big question if you're looking at like an anthropic IPO, and don't worry, I'm not doing a Bloomberg and like pulling you into be like, what do you think of the anthropic IPO?
44:45
Unidentified speaker: I hate creating it.
44:48
Unidentified speaker: Yeah, it's great.
44:48
Unidentified speaker: It's uh no one in AI is like, uh, not a fan.
44:53
Unidentified speaker: Everyone is against it.
44:54
Unidentified speaker: Go.
44:55
Unidentified speaker: But I do think a big the mother load, so to speak, is claw code.
44:59
Unidentified speaker: It's $60 billion of revenue.
45:01
Unidentified speaker: It largely kind of dictates the $2 trillion price.
45:05
Unidentified speaker: And you look at Mark Benioff, who says that they spend 5% of engineers' salaries, 300 million, on claw code today.
45:12
Unidentified speaker: Now, if that comes way down as efficiencies improve, you name it, that's a challenge.
45:18
Unidentified speaker: Or will that go up as it just eats more and more of engineers' workflows?
45:22
Unidentified speaker: Yeah, it's just gonna eat more.
45:23
Unidentified speaker: I mean, like again, like we've seen this so many times.
45:25
Unidentified speaker: Like you make something more efficient, it makes something more powerful, it just unlocks like five new use cases, and then demand just goes up in the end.
45:31
Unidentified speaker: Like that's that's just the history of like the last 50 years of computers and software.
45:35
Unidentified speaker: But like I wouldn't be here if I wasn't such an optimist.
45:38
Unidentified speaker: That's why I started a company.
45:39
Unidentified speaker: Like, I'm just like the biggest like optimist that there is in the world about software and computers, and like I just love this shit.
45:45
Unidentified speaker: And I think it's just gonna keep growing.
45:46
Unidentified speaker: The data market is an interesting one.
45:49
Unidentified speaker: We have so many players that have scaled so fast.
45:53
Unidentified speaker: Yeah, it's incredible.
45:54
Unidentified speaker: If I was 20 and I just wanted to make a lot of money, I would go out and start an environment company and just sell data to Frontier Labs.
46:00
Unidentified speaker: There is so much money to be made in that space.
46:02
Unidentified speaker: It's kind of incredible.
46:03
Unidentified speaker: Let's go layer deeper.
46:04
Unidentified speaker: What would you do next?
46:05
Unidentified speaker: You specialize on a specific market?
46:07
Unidentified speaker: I mean, the coding space is where you see the most demand right now.
46:10
Unidentified speaker: Like if you can sell like code environments to the labs, like you can make a lot of money.
46:13
Unidentified speaker: It's a big space in that.
46:15
Unidentified speaker: Still.
46:15
Unidentified speaker: Still, yeah.
46:16
Unidentified speaker: Yeah.
46:16
Unidentified speaker: I meet startups all the time.
46:18
Unidentified speaker: I I angel invested in a bunch of that.
46:20
Unidentified speaker: I mean, they're incredible businesses.
46:21
Unidentified speaker: What I will say, like my one piece of advice is like you make a lot of money today, but like, is this a venture fund?
46:26
Unidentified speaker: Like, do you want to get on the venture treadmill?
46:28
Unidentified speaker: I'm not sure.
46:28
Unidentified speaker: Like, I think these are companies that are gonna make a lot of money today.
46:31
Unidentified speaker: I don't know if they're gonna make a lot of money in the very long run, but like who cares?
46:34
Unidentified speaker: Like, you know, take the money.
46:35
Unidentified speaker: It's like great money.
46:36
Unidentified speaker: Like it it's a fantastic business today, and that's all that matters.
46:38
Unidentified speaker: I think there's data marketplaces and then there's data and services.
46:42
Unidentified speaker: And when you look at the the leading providers, they started as the data requires and then just sell you the data.
46:48
Unidentified speaker: And that's kind of it.
46:49
Unidentified speaker: And I think they've really gone full lifecycle now to deployment, implementation, evaluation, monitoring with a real enterprise handholding ability, similar to the FDEs, where they will really be that partner to ensure full implementation.
47:03
Unidentified speaker: That's how you expand margin and have some form of business in the long term.
47:07
Unidentified speaker: For sure.
47:08
Unidentified speaker: Yeah, and I think it's gonna be a tough business in the very long run.
47:10
Unidentified speaker: But like today, you can have extremely good cash flow selling these data sets.
47:13
Unidentified speaker: Can I say when we think about that kind of implementation working with partners so closely?
47:18
Unidentified speaker: You said, and Lynn has said too from Fireworks about the importance of owning your own intelligence.
47:23
Unidentified speaker: Can you unpack how this looks?
47:26
Unidentified speaker: Because specialized models for companies in this way, the kind of jury is still out.
47:31
Unidentified speaker: You know, we have Harvey who says that it's fantastic and it's been amazing.
47:36
Unidentified speaker: And then you have Lagora who say ridiculous, the underlying models move so fast, it's the wrong strategy.
47:41
Unidentified speaker: Yeah, for some stuff it's good to own your own intelligence.
47:43
Unidentified speaker: For other stuff, it's not good.
47:44
Unidentified speaker: Like I don't think it matters.
47:45
Unidentified speaker: Like we have a lot of customers who are on proprietary models on model too.
47:48
Unidentified speaker: Like, and their customers, you know, they use a proprietary model.
47:51
Unidentified speaker: Like, I don't I don't think it makes sense for everyone to own their own intelligence.
47:54
Unidentified speaker: Why does it make sense for you two versus not?
47:56
Unidentified speaker: I I think for certain use cases, you can take an open source model and you can fine-tune it and you can get really good performance.
48:01
Unidentified speaker: And the nice thing is you know what the fine-tuned model is based on.
48:04
Unidentified speaker: It's not going to change, like they're not going to take away the model from you.
48:07
Unidentified speaker: It's nice to have that consistency and the ability to like continuously tweak it without without the underlying model changing, which has happened in some cases with the Frontier Labs.
48:16
Unidentified speaker: That's like, you know, that's nice.
48:18
Unidentified speaker: Like, you know, I support that, but like I, you know, I think there's many other things that matter to people.
48:22
Unidentified speaker: You know, owning your intelligence is like just one thing out of many.
48:25
Unidentified speaker: Like I think latency matters, I think cost matters, performance, I think, you know, hackability.
48:29
Unidentified speaker: I think you know, being, you know, having a platform of a lot of adjacent services.
48:33
Unidentified speaker: Like, these things matter too.
48:34
Unidentified speaker: So owning your own intelligence, like, yeah, I'm supportive, but you know, there's many other things.
48:38
Unidentified speaker: Aaron Powell, isn't that kind of going against your business?
48:40
Unidentified speaker: No, I don't think so.
48:41
Unidentified speaker: I I we the I love the honesty, but like, dude, it would be in your favor to say, yes, every company's gonna need it, and we're the partner to help you.
48:50
Unidentified speaker: I kind of fucking teed you up there and you Well, I mean, not not really.
48:54
Unidentified speaker: I mean, if you think about TypeSafe with Jav, right?
48:56
Unidentified speaker: Like, you know, it's a highly proprietary model.
48:58
Unidentified speaker: Like, I want people to use Java.
49:00
Unidentified speaker: I mean, it runs on modal.
49:01
Unidentified speaker: Like, like so we have many customers like that themselves are proprietary model, like that are hosting on modal that are serving their customers.
49:08
Unidentified speaker: And I'm very, very in favor of that too.
49:10
Unidentified speaker: I I think the beauty of modal is always that there's like a general purpose platform, and we support so many different things.
49:15
Unidentified speaker: That that to me is like you know the beauty.
49:16
Unidentified speaker: It's like, you know, we can if you want to use open source models, like that's great.
49:20
Unidentified speaker: Like we'll help you.
49:20
Unidentified speaker: And there's many customers that build incredible AI applications on top of open source models, fine-tuning them, reinforcement learning, have a bunch of you know adjacent uh sandbox use cases or training, whatever.
49:31
Unidentified speaker: But there's a lot of companies that have built highly successful businesses, you know, Suno, Runway, TypeSafe, and many others like running their own proprietary model on Modal, like and their customers don't care.
49:41
Unidentified speaker: And that that I think is like you know, puts us in a really good spot.
49:44
Unidentified speaker: I want to ask two more questions before I do a quick fire.
49:46
Unidentified speaker: Okay.
49:47
Unidentified speaker: Model routing, as the dear Americans say.
49:49
Unidentified speaker: Is that a defensible business?
49:51
Unidentified speaker: What do you say in the UK rooting?
49:52
Unidentified speaker: Rooting.
49:53
Unidentified speaker: I love that.
49:53
Unidentified speaker: Well, I mean, we are actually the original language America just took ours and balsadized it.
49:58
Unidentified speaker: But yes.
49:59
Unidentified speaker: Uh my question to you, is that a defensible segment of this value chain, or is that a commoditized product that everyone just builds themselves?
50:07
Unidentified speaker: Aaron Powell I think it's possible to build a successful business in model routing.
50:12
Unidentified speaker: I think you can do that today as a standalone product.
50:16
Unidentified speaker: I wonder if long-term it has to be part of a more comprehensive platform.
50:20
Unidentified speaker: At the end of the day, I think what what these companies want, they want evals, they want you know ways to fine-tune, they want you know more control over performance, like latency, throughput, like all these things.
50:30
Unidentified speaker: And I think it's like one of those things that kind of makes sense today.
50:33
Unidentified speaker: Like these are very good businesses.
50:34
Unidentified speaker: I invested in open router personally, and they had a great outcome.
50:38
Unidentified speaker: I actually think they could have done even better on a standalone basis.
50:40
Unidentified speaker: But I think in the very long term, it feels like one of those categories, like, you know, and but that's my view on everything in AI.
50:46
Unidentified speaker: It should really be a part of you know a larger, more comprehensive set of tools.
50:49
Unidentified speaker: Aaron Powell Congratulations, you're on the open router board and the acquisition offer comes in.
50:54
Unidentified speaker: Please make the case for why they would have read.
50:56
Unidentified speaker: Project.
50:57
Unidentified speaker: Yeah.
50:57
Unidentified speaker: Yeah, please make the case for why.
50:58
Unidentified speaker: I think open router, like the growth rate and the revenue, like this is and and the brand, frankly, like it they have a really strong brand.
51:04
Unidentified speaker: I mean, there's a lot of like there's 500 other competitors, but but the the space is growing so fast and there's so much revenue I would have told them to keep going.
51:12
Unidentified speaker: Final one.
51:12
Unidentified speaker: I think a really telling question is what do you believe that does not go in the favor of your business?
51:19
Unidentified speaker: And so for example, I think that seed investing today is kind of dead and bullshit.
51:24
Unidentified speaker: This idea of like two on ten or slowly build for 18 months.
51:29
Unidentified speaker: Well, it's kind of bad.
51:29
Unidentified speaker: Like it's kind of bad if you want to make money, right?
51:31
Unidentified speaker: But it's great if you want to develop a relationship with engineers.
51:33
Unidentified speaker: Or me, I mean with founders.
51:35
Unidentified speaker: Sure.
51:35
Unidentified speaker: But that goes against what I do.
51:39
Unidentified speaker: You're you're against the relationship with founders, or no, I'm saying like I think the death of seed investing, the two on ten and build for eighteen months, is dead.
51:46
Unidentified speaker: Yeah.
51:46
Unidentified speaker: No, I understand what you're saying.
51:48
Unidentified speaker: So what would yours be?
51:49
Unidentified speaker: I I don't know.
51:49
Unidentified speaker: I I mean I think I think one one thought I have is like obviously at some point this AI stuff is going so fast and we're here and we're gonna go here, and then there's gonna be pullback, you know, 20%, 30% in the stock market, everyone's gonna freak out, and like people are gonna say it's the end of the world.
52:04
Unidentified speaker: I'm almost certain it's gonna happen, but like I'm so bullish because then we're gonna keep going.
52:08
Unidentified speaker: And like, you know, like so so that that would be like one thing that I'm like, you know.
52:12
Unidentified speaker: There's gonna be a pullback.
52:13
Unidentified speaker: I'm okay with that.
52:14
Unidentified speaker: Because I think the long-term opportunity is so fucking big.
52:17
Unidentified speaker: It's like so large.
52:18
Unidentified speaker: Do you take secondaries?
52:20
Unidentified speaker: Very little.
52:20
Unidentified speaker: Why?
52:21
Unidentified speaker: I mean, I personally don't need a lot of money.
52:24
Unidentified speaker: I'm pretty happy with life and I have enough.
52:27
Unidentified speaker: So personally, I don't I I actually support it for my employees, though.
52:30
Unidentified speaker: It's it's quite important actually as a founder that you make the equity worth as much as possible.
52:35
Unidentified speaker: And one of the ways to do that and to tell future candidates that no, this isn't just pay paper money, this is real money, is to have a precedent of doing employee secondaries.
52:44
Unidentified speaker: And you can point to last round, we let employees sell this much or whatever, and you know, up to 10%.
52:50
Unidentified speaker: Like I think you have to have guardrails.
52:51
Unidentified speaker: Like you don't want employees to sell too much.
52:53
Unidentified speaker: You want them to still be very, you know, have a lot of skin in the game and and be very incentivized to make the company successful.
52:59
Unidentified speaker: But I think it's a it's a way to like you know, like create a very liquid hard currency out of the equity.
53:04
Unidentified speaker: Aaron Powell I totally agree.
53:05
Unidentified speaker: And as you think about a talent marketplace as more competitive, it's fair.
53:09
Unidentified speaker: It's extremely competitive out there.
53:11
Unidentified speaker: Is it nuts?
53:12
Unidentified speaker: It's nuts.
53:12
Unidentified speaker: It's nuts.
53:13
Unidentified speaker: I mean, you know, it like the the offers we see for we've lost new grads going to you know some of the labs for like insane, like talking, you know, 20 plus million.
53:22
Unidentified speaker: I mean, these are exceptional people, exceptional.
53:25
Unidentified speaker: What makes an exceptional, exceptional new grad?
53:28
Unidentified speaker: You know, like just fucking sharp people.
53:31
Unidentified speaker: You know, the these are like I I know it when I see it.
53:34
Unidentified speaker: Like they're just you know, they're just extremely hungry, hardworking.
53:38
Unidentified speaker: I mean everything about technology.
53:40
Unidentified speaker: I mean they go to some of the big labs.
53:41
Unidentified speaker: Is that it?
53:41
Unidentified speaker: They go to open air and anthropology.
53:43
Unidentified speaker: Yeah, the big labs, yeah.
53:43
Unidentified speaker: They pay a lot.
53:44
Unidentified speaker: But is that it?
53:45
Unidentified speaker: Is that where the talent comes from?
53:46
Unidentified speaker: There's like a few other ones, but but I mean, mainly those people are are known for paying very a lot of money.
53:51
Unidentified speaker: And you won't pay that much?
53:53
Unidentified speaker: We pay a lot.
53:54
Unidentified speaker: We can go very aggressive.
53:56
Unidentified speaker: We want to be out there paying 95 fifth percentile for in the market.
53:59
Unidentified speaker: And there's like a few companies that I know, like they're probably gonna pay a little bit more.
54:02
Unidentified speaker: I think we have way more upside in the equity.
54:04
Unidentified speaker: And going back to what I said about secondaries, I think it's important that for that reason, also like to create like a real tangible value and the currency in the equity that makes us, you know, our offers seem very, very strong.
54:14
Unidentified speaker: Plus the fact that we have a lot of upside.
54:16
Unidentified speaker: I mean, you you you said it, right?
54:17
Unidentified speaker: Like our stock seems cheap.
54:18
Unidentified speaker: Like, I kind of agree.
54:19
Unidentified speaker: There's a lot of upside in our stock.
54:21
Unidentified speaker: So that enables us to not necessarily have to match with the labs, but I think, you know, create a compelling sales pitch for why our comp package is like basically better than a labs.
54:31
Unidentified speaker: And we're willing to go very far to get the best talent.
54:33
Unidentified speaker: Okay, we're gonna do a quick fire round because otherwise I could talk to you all day.
54:36
Unidentified speaker: You have a great angel portfolio.
54:38
Unidentified speaker: Open router, you mentioned lovable before.
54:41
Unidentified speaker: You can put all of your money into only one.
54:44
Unidentified speaker: I love TurboPuffer.
54:45
Unidentified speaker: I know Simon really well, Danish guy.
54:47
Unidentified speaker: So you have the Scandinavian connection.
54:50
Unidentified speaker: Just a fantastic entrepreneur.
54:51
Unidentified speaker: I just love the business.
54:52
Unidentified speaker: How big can that business be?
54:54
Unidentified speaker: That's a good question.
54:54
Unidentified speaker: I mean, it is like somewhat specialized.
54:56
Unidentified speaker: They're doing vector databases.
54:57
Unidentified speaker: I think it could be a very large category.
54:59
Unidentified speaker: If I ran that company, I would have 500 different ideas of how to expand into many adjacent things around databases, data warehouses, other types of query, querying system.
55:07
Unidentified speaker: Like there's it could be a large category.
55:10
Unidentified speaker: What have you changed your mind on most in the last 12 months?
55:13
Unidentified speaker: Probably the magnitude of the sandbox space.
55:15
Unidentified speaker: I think it's kind of insane.
55:16
Unidentified speaker: Like I look at the revenue and it just keeps going up and I'm like, well, clearly it has to stop right now.
55:20
Unidentified speaker: And I think I'm like doing some sort of reverse psychology on the revenue because it just keeps going up.
55:24
Unidentified speaker: I'm like, what is going on?
55:25
Unidentified speaker: It just keeps growing and growing.
55:27
Unidentified speaker: And like I've been wrong so many times.
55:28
Unidentified speaker: I was like, you know, in a good way, like, you know, and and and so I'm uh it's incredible how large that space has gotten what do you think is the most overheated category within AI today?
55:38
Unidentified speaker: Oh boy, I I don't know.
55:39
Unidentified speaker: I I don't think anything is overheated.
55:41
Unidentified speaker: I I will say, like, one of the things that I'm a little nervous about is companies that are dependent on continual infusion of equity capital.
55:50
Unidentified speaker: I would not invest in them.
55:52
Unidentified speaker: And there's a lot of them that I I would be nervous about anything where like if there's a venture capital pullback, they will struggle to raise more money.
56:00
Unidentified speaker: I think it's important you raise the the every round as if it's your last round.
56:04
Unidentified speaker: And there's like a real path towards commercialization and revenue.
56:07
Unidentified speaker: You got two kids.
56:08
Unidentified speaker: Two kids.
56:09
Unidentified speaker: Biggest lesson on being a good dad and being a good CEO.
56:12
Unidentified speaker: I mean, sometimes I feel like it's kind of similar.
56:14
Unidentified speaker: Like you mean like in terms of like how to balance it, or do you mean in terms of just like being a good dad?
56:18
Unidentified speaker: What does it take to be a good dad?
56:20
Unidentified speaker: Setting them up for success in life.
56:22
Unidentified speaker: And I always like to think about like the right amount to push the kids.
56:25
Unidentified speaker: I'm not a tiger dad, but like I think a little bit you need to push the kids a little bit to like get them exposure.
56:30
Unidentified speaker: I don't want to create, you know, in the future a situation where my kids come back to me and say, Hey dad, I wish you'd made me do X, or I wish you'd made me do Y.
56:38
Unidentified speaker: But I also don't want to create a situation where my my kids come back to me when they're 20 and they say, Hey dad, why did you make us do X or why did you make us do Y?
56:46
Unidentified speaker: So a little bit of pushing.
56:47
Unidentified speaker: I make them play piano.
56:48
Unidentified speaker: I don't know why.
56:49
Unidentified speaker: But like I think it's like a good way to like learn something that I hope, you know, when they're 20, 30, they will actually appreciate.
56:55
Unidentified speaker: They don't like it that much.
56:56
Unidentified speaker: But I think stuff like that, you gotta push them a little bit, but not too much.
56:59
Unidentified speaker: My my parents are very lax and you know, I discovered math and and programming and they were like, Eric, why are you doing so much math?
57:06
Unidentified speaker: Like you should go out and play like the other kids.
57:08
Unidentified speaker: So I in a way I actually wish they would have pushed me even harder, but I pushed myself so hard anyway, so it doesn't matter.
57:13
Unidentified speaker: So you let that like right amount of pushes.
57:15
Unidentified speaker: Why did you push yourself so hard?
57:16
Unidentified speaker: You're sorry, you're actually from a nation where you are told that it's like bad to stand out.
57:22
Unidentified speaker: I'm forgetting that word, but I had a Swedish girlfriend.
57:28
Unidentified speaker: Yeah, where it's like it's bad to stand out.
57:30
Unidentified speaker: Yeah, don't think you're special.
57:31
Unidentified speaker: Yeah, why but I I don't know if that's true.
57:33
Unidentified speaker: Like I think there's a lot of misconceptions of Scandinavian culture.
57:36
Unidentified speaker: I do think there's a lot of like respect for people who follow their dream and like do things.
57:40
Unidentified speaker: And like whether, like I said, you know, it's music or writing books or you know, making furniture or art, like or being an entrepreneur.
57:47
Unidentified speaker: Like I actually think there's a lot of respect for I think there's like you know, self-actualization and following your dreams is like kind of highly regarded.
57:53
Unidentified speaker: And I think you know, like that's the flip side of having a very lax culture.
57:56
Unidentified speaker: It's like you kind of pursue people to follow their dreams.
57:59
Unidentified speaker: And I don't know.
57:59
Unidentified speaker: I my dream was like doing math and programming.
58:01
Unidentified speaker: I love that shit.
58:02
Unidentified speaker: So I did that.
58:02
Unidentified speaker: Aaron Powell Unless you're from Norway where a tax structure exists that basically makes it impossible to start a company.
58:07
Unidentified speaker: Aaron Powell Yeah.
58:08
Unidentified speaker: I mean Norway has some issues.
58:10
Unidentified speaker: Um they also barely work.
58:12
Unidentified speaker: Yeah, this is very true.
58:13
Unidentified speaker: But we do have a sovereign wealth fund.
58:14
Unidentified speaker: Yeah.
58:15
Unidentified speaker: So they are rich.
58:16
Unidentified speaker: Thank God for that.
58:17
Unidentified speaker: Maybe it's like the first post-economic society.
58:18
Unidentified speaker: I don't know.
58:19
Unidentified speaker: Literally all.
58:20
Unidentified speaker: In certain in in multiple meanings of the word.
58:22
Unidentified speaker: Aaron Powell Final one for you.
58:23
Unidentified speaker: What are you most excited for?
58:25
Unidentified speaker: It can be anything.
58:25
Unidentified speaker: Like I I look at drug discovery for kind of chronic conditions.
58:28
Unidentified speaker: My mom's actually drug discovery is like one of the applications I think is so incredibly cool.
58:32
Unidentified speaker: Like I can't wait to take AI generated drugs.
58:35
Unidentified speaker: You know, I'm like old and my skin is getting bad, and like I don't know, like you know, I every time I drink a beer, I feel like horrible the day after.
58:42
Unidentified speaker: Like, I don't know.
58:42
Unidentified speaker: Give me all the AI generated drugs I can take.
58:44
Unidentified speaker: Like I I want to live long, I want to live a healthy life.
58:47
Unidentified speaker: I uh you know, and I think it's Do you think that is actually a realistic future?
58:50
Unidentified speaker: Aaron Powell I think so.
58:51
Unidentified speaker: But but I think with the FDA and other things, it's gonna take a while.
58:54
Unidentified speaker: But I think there's just so much incredible progress today in like in AI-aided drug design.
58:58
Unidentified speaker: I think there's so many cool stuff, so many things.
59:01
Unidentified speaker: Dude, I've so enjoyed this show in a way that I I honestly wasn't expecting you to be this fun, open, charismatic.
59:08
Unidentified speaker: You've been amazing.
59:09
Unidentified speaker: So thank you so much.
59:10
Unidentified speaker: Thank you so much for having me.
59:11
Unidentified speaker: This is a lot of fun.
59:14
Unidentified speaker: But before we leave you today, Blitzy, the autonomous software development platform that's built for enterprise scale, is proud to sponsor today's show.
59:22
Unidentified speaker: Enterprise code bases have grown beyond human comprehension.
59:26
Unidentified speaker: Tens of millions of lines, decades of tech debt, complexity existing tools cannot fathom.
59:32
Unidentified speaker: Blitzy starts by understanding.
59:34
Unidentified speaker: Thousands of agents reverse engineer your code base into a proprietary knowledge graph that maps architecture dependencies and business logic.
59:41
Unidentified speaker: That graph surfaces every CVE and zero-day hiding in your code before anyone hands you a ticket.
59:48
Unidentified speaker: Grounded in that context, Blitzy then executes entire epics autonomously, modernize off legacy systems, ship new features 5x faster, or batch remediate CVEs at scale surfaced by the platform.
1:00:01
Unidentified speaker: Blitzy not only generates code, it validates and end-to-end tests every single line.
1:00:06
Unidentified speaker: And that's why Fortune 500 enterprises are shipping enterprise grade code five times faster.
1:00:13
Unidentified speaker: Go to blitzy.com forward slash 20vC to learn more.
1:00:17
Unidentified speaker: While Blitzy helps you build faster, Flex helps you run your finances smarter.
1:00:22
Unidentified speaker: Business owners run their whole financial life on Flex.
1:00:26
Unidentified speaker: One platform from business revenue to their personal spend.
1:00:30
Unidentified speaker: Float every purchase for 60 days, tap capital that grows with your revenue and pay vendors in 170 countries across 32 currencies.
1:00:40
Unidentified speaker: Plus the whole back office, bills, expenses, accounting, all in one place.
1:00:46
Unidentified speaker: So you spend less time reconciling and more time growing.
1:00:49
Unidentified speaker: That's why thousands of owners use Flex, named one of Fast Company's most innovative companies of 2026.
1:00:56
Unidentified speaker: Visit Flex.1, that's f L E X dot O-N-E, and use the code 20VC.
1:01:05
Unidentified speaker: While Flex runs your finances, Vanta keeps your compliance covered.
1:01:09
Unidentified speaker: When you're building a company, you learn that trust is what closes deals.
1:01:12
Unidentified speaker: You may have the best product, but no buyers will sign these days without proof of your security.
1:01:17
Unidentified speaker: Here's what happens if you're not prepared.
1:01:19
Unidentified speaker: A prospect asks for proof of compliance.
1:01:21
Unidentified speaker: The deal stalls when you scramble.
1:01:23
Unidentified speaker: Your engineer gets pulled off the roadmap to audit prep.
1:01:26
Unidentified speaker: Every enterprise conversation turns into this horrible fire drill.
1:01:30
Unidentified speaker: That's where Vanta comes in.
1:01:31
Unidentified speaker: Vanta is the leading agentic trust platform that not only gets you compliant fast with frameworks like SOC2, ISO 27001, HIPAA, and GDPR, but keeps you compliant by continuously monitoring your controls.
1:01:44
Unidentified speaker: So your deals keep moving and your engineers really keep building.
1:01:47
Unidentified speaker: Now access the Vanta agent everywhere you work, even if your team lives in Claude or Cursor.
1:01:53
Unidentified speaker: And that's why Vanta's trusted by more than 16,000 companies like Ramp, Harvey, and Writer.
1:01:59
Unidentified speaker: So prove you're ready for business and get $1,000 off Vanta.
1:02:03
Unidentified speaker: When you go to vanta.com forward slash 20 VC, that's vanta.com slash 20 VC.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3 and interpreted by Qwen3.6-35B-A3B-GGUF:UD-Q4_K_XL. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.

More from The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch

The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch

20VC: Jev Creator on Why Data Is More Important Than Compute | Will US Open-Source Models Catch Chinese | The Cost of Intelligence and Tokens: What No One Sees But Needs to Know with Diogo Almeida

Episode discussion: Jev Creator argues AI reliability and data quality trump compute, predicting a shift from front-end apps to background automation.

Speakers’ claims: Jev Creator (Diogo Almeida) argues that the primary bottleneck for AI automation is not compute or speed, but the reliability of intelligence, defined as the consistency of decision-making rather than system uptime. He posits a 'bitterer lesson' where task definition and data quality are more critical than algorithms or compute, which he claims have plateaued in effectiveness since GPT-4.5. Creator criticizes current market hype, benchmark-maxing, and the 'AI coworker' model, predicting a 'SaaS Palooza' where incumbents win by integrating AI into workflows rather than replacing them. He also discusses TypeSafe AI's fundraising strategy, hiring philosophy, and the geopolitical landscape of open-source models.

AI ReliabilityData QualityVenture Capital
1h 14m2 assets22 claims to verify
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch

20VC: Taking on NVIDIA: Fractile CEO on the Future of Chips | What Everyone Gets Wrong About Capex Spending | The Winners and Losers in Neo Clouds with Walter Goodwin

Episode discussion: Walter Goodwin on Fractile's 3D-stacked memory chips, AI infrastructure bottlenecks, and NVIDIA's future valuation

Speakers’ claims: Walter Goodwin, CEO of Fractile, argues that frontier AI intelligence is not yet commoditized, distinguishing it from general cloud compute. He positions Fractile's 3D-stacked memory architecture as a solution to the memory bandwidth bottleneck that limits LLM inference speeds. Goodwin contends that land and power, rather than silicon, are now the primary constraints for AI infrastructure, favoring energy-efficient designs. He predicts NVIDIA will reach a $10 trillion valuation within five years, driven by the finite nature of data center power and the need for chip replacement cycles.

SemiconductorsAI InfrastructureVenture Capital
1h 16m3 assets25 claims to verify
All insights