David Weisburd discusses AI adoption challenges, Anthropic valuation, and early-stage investing strategies
Summary
David Weisburd argues that enterprise AI is in its 'first inning,' citing a Deloitte statistic that 95% of AI pilots fail due to poor data infrastructure and cultural resistance. He highlights that 80% of valuable data remains behind enterprise firewalls, creating a moat for companies that can securely access it. Weisburd shares case studies of portfolio companies like Inspire and Bark, demonstrating the value of physical AI and specialized use cases. He also critiques high pre-seed valuations, advocating for price discipline and rational underwriting of high-probability outcomes. Finally, he notes that agentic coding is increasing founder efficiency, creating alpha opportunities in early-stage rounds.
Episode sections
- 0:00 – 1:49
Anthropic Valuation and AI Market Size
David Weisburd discusses his investment in Anthropic at a $4 billion valuation and references Lucas Twischer's three S-curves of AI adoption, identifying developers as current early adopters.
- 1:49 – 3:51
Enterprise AI Adoption Challenges and Failure Rates
Weisburd cites a Deloitte statistic that 95% of enterprise AI pilots fail, attributing this to poor data infrastructure, lack of expertise, and cultural resistance in legacy industries.
- 3:51 – 6:47
Data Silos and Cultural Barriers
The discussion highlights that large enterprises suffer from complex data silos and inconsistent tagging, requiring top-down cultural leadership for successful AI adoption.
- 6:47 – 8:53
The 80% Behind-the-Firewall Data Moat
Weisburd states that 80% of the world's data is behind enterprise firewalls, arguing that competitive advantage lies in unlocking this proprietary data through secure workflows.
- 8:53 – 16:07
Case Study: Inspire (Physical AI in Healthcare)
Inspire uses computer vision and sensors in senior living facilities to extract real-time data, preventing falls and predicting health issues, demonstrating the value of physical AI.
- 16:07 – 17:52
Case Study: Bark (Child Safety AI)
Bark uses natural language processing to monitor children's online activity for signs of depression or predators, cited as a positive AI use case.
- 17:52 – 22:35
Case Study: Evercoast (4D Human Capture)
Evercoast captures 4D in-motion data of humans for media clients, creating datasets useful for training robotics and self-driving cars.
- 22:35 – 27:29
AI Business Models and the 'Do the Unscalable' Principle
Weisburd discusses the shift from SaaS to usage-based models and advocates for solving problems manually first to ensure value creation before automation.
- 27:29 – 30:09
AI Efficiency and Seed Stage Alpha
Speakers discuss how agentic coding allows technical founders to achieve product maturity faster, creating alpha opportunities in early-stage investing.
- 30:09 – 31:51
Valuation Discipline and Risk Allocation
The speaker argues against high pre-seed valuations, preferring to underwrite rational, high-probability scenarios rather than betting on rare, generational winners.
- 31:51 – 36:52
Anthropic Valuation and Competitive Landscape
A heuristic is proposed to value Anthropic by comparing it to a $100 million run rate, while warning of risks from open-weight models and low switching costs.
- 36:52 – 39:12
Adoption S-Curves and Market Timing
The conversation outlines three S-curves of AI adoption, noting that AI is still in the 'first inning' with significant room for growth in vertical solutions.
- 39:12 – 43:44
Vertical vs. Horizontal AI and Sales Integration
Vertical AI wins by leveraging domain expertise to build trust, while salespeople remain crucial for gathering customer insights and driving upsells.
- 43:44 – 49:18
Post-Sales Data and Product Roadmap
Known Well uses AI to analyze post-sales service interactions, providing insights into customer needs that feed back into the product roadmap.
Assets discussed
| Asset | Kind | Stance | Context |
|---|---|---|---|
| Anthropic | private company | mixed | So I did invest in the anthropic at a $4 billion valuation. |
| SpaceX | private company | neutral | So if I'm the CEO of Delta, let's say I just put my foot in my mouth with a Starlink and saying that I'm don't want Starlink and don't like SpaceX, all these things. |
| TeslaTSLA | equity | neutral | Just think about Tesla, for example, with their self-driving car and their self-driving technology. |
“Stance” is how the speakers talked about the asset, not a recommendation.
Claims worth verifying
David Weisburd invested in Anthropic at a $4 billion valuation.
Unidentified speaker ·0:00
How to verify: Check Weisburd's public investment disclosures or portfolio announcements from Vein Capital Ventures or his personal investment records.
There are 50 million developers in the world out of 8 billion people.
Unidentified speaker ·0:55
How to verify: Cross-reference with global workforce statistics from the International Labour Organization (ILO) or World Bank data on software developers.
Pre-seed rounds today average a $100 million valuation.
Unidentified speaker ·1:26
How to verify: Review recent venture capital market reports from sources like PitchBook, CB Insights, or Preqin for 2025-2026 pre-seed deal data.
95% of enterprise AI deployments/pilots are failing.
Unidentified speaker ·2:45
How to verify: Locate the specific Deloitte report cited by the speaker containing this statistic and verify the methodology and sample size.
Approximately 80% or more of the world's data is behind firewalls.
Unidentified speaker ·6:50
How to verify: Search for enterprise data management studies or reports from firms like IDC, Gartner, or the specific enterprise partner mentioned by Weisburd that generated this stat.
Inspire closed a $170 million funding round from Insight and New View Ventures.
Unidentified speaker ·8:53
How to verify: Check Crunchbase, PitchBook, or official press releases from Insight Ventures and New View Ventures for Inspire's funding history.
Inspire was able to eliminate 60 to 70 percent of falls in senior living facilities.
Unidentified speaker ·10:48
How to verify: Review case studies, white papers, or press releases from Inspire detailing their clinical or operational outcomes.
Jensen Huang predicts physical AI will be a $50 trillion market.
Unidentified speaker ·13:22
How to verify: Search for Jensen Huang's speeches, interviews, or NVIDIA earnings call transcripts where he discusses the physical AI market size.
Bark ingests about 7 million kids' data real-time across multiple apps.
Unidentified speaker ·16:25
How to verify: Check Bark's technical documentation, investor presentations, or press releases for user base and data processing metrics.
Vista Equity has deployed over $100 billion in software.
Unidentified speaker ·
How to verify: Review Vista Equity Partners' official website, annual reports, or press releases regarding their total deployed capital in software.
Anthropic is reported to have a $2 trillion valuation.
Unidentified speaker ·32:01
How to verify: Review financial news reports and funding announcements regarding Anthropic's valuation in 2026.
Anthropic had a reported $100 million run rate in 2026.
Unidentified speaker ·32:11
How to verify: Review financial disclosures or credible news reports on Anthropic's revenue for 2026.
95% of AI projects in enterprises fail.
Unidentified speaker ·37:24
How to verify: Search for industry studies or reports from 2025-2026 citing the failure rate of enterprise AI implementations.
There are 50 million developers in the world.
Unidentified speaker ·37:07
How to verify: Check global workforce statistics from organizations like the World Economic Forum or developer surveys (e.g., Stack Overflow) for 2025-2026.
Consumer willingness to pay for AI is between zero and ten dollars per month.
Unidentified speaker ·33:40
How to verify: Review consumer surveys or market research reports on AI subscription pricing willingness in 2026.
The speaker invested in Anthropic at a $4 billion valuation.
Unidentified speaker ·34:52
How to verify: Check the speaker's public investment disclosures or fund announcements regarding Anthropic investments.
Average engineers are getting 4-5x efficiency gains from AI, while 10x engineers get 100x gains.
Unidentified speaker ·45:21
How to verify: Look for internal studies or public statements from Anthropic or OpenAI regarding AI tool efficiency metrics.
Anthropic tends to be more enterprise-focused than OpenAI, which is more consumer-focused.
Unidentified speaker ·36:34
How to verify: Analyze customer case studies, product offerings, and market positioning of Anthropic (Claude) vs. OpenAI (ChatGPT) in 2026.
GoDaddy increased revenue by putting salespeople into customer service calls.
Unidentified speaker ·43:13
How to verify: Review business case studies or articles detailing GoDaddy's sales and customer service strategies.
Do your own research
- Verify the Deloitte statistic regarding the 95% failure rate of enterprise AI pilots by accessing the original Deloitte report.
- Research Anthropic's latest funding rounds and valuation metrics from primary sources like Crunchbase or PitchBook.
- Investigate the data infrastructure challenges in legacy industries by reviewing case studies on data silos and governance.
- Analyze the competitive landscape of open-weight AI models and their impact on incumbents like Anthropic.
- Review the financials and business models of portfolio companies mentioned, such as Inspire, Bark, and Evercoast.
- Assess the current market rates for pre-seed valuations to contextualize the critique of $100 million entry points.
- Examine the adoption rates of agentic coding tools among technical founders to validate claims about efficiency gains.
Full transcriptexpand
- 0:00
- Unidentified speaker: So I did invest in the anthropic at a $4 billion valuation.
- 0:03
- Unidentified speaker: How would you underwrite it today?
- 0:09
- Unidentified speaker: Right.
- 0:09
- Unidentified speaker: So how big is the market likely to be?
- 0:11
- Unidentified speaker: And then what percentage are you expecting to capture as anthropic vis-a-vis or relative to open AI and some unknown entrance?
- 0:19
- Unidentified speaker: So I think that's hard to handicap on the who's coming next, right?
- 0:22
- Unidentified speaker: Conversation.
- 0:23
- Unidentified speaker: It's the unknown unknown.
- 0:25
- Unidentified speaker: Who's going to have the idea or the resources?
- 0:27
- Unidentified speaker: That's the unknown unknown.
- 0:29
- Unidentified speaker: Um, but you have other folks that are devoting a lot of resources to the space that have billions of dollars in reasons to be highly defensive.
- 0:36
- Unidentified speaker: The Google, the Metas, the Amazons, uh, et cetera.
- 0:39
- Unidentified speaker: I'll tell you the full case, virtue of Lucas Twischer, he runs growth for Code 2 company.
- 0:44
- Unidentified speaker: He talked about the three S curves of adoption.
- 0:48
- Unidentified speaker: So the first S curve, the early adopters in terms of LLMs from a revenue standpoint have been developers.
- 0:55
- Unidentified speaker: What's crazy about that is that there's only 50 million developers in the world out of 8 billion people.
- 1:00
- Unidentified speaker: And on top of that, only a small percentage of developers have actually implemented.
- 1:04
- Unidentified speaker: If you look at worldwide kind of engineers, then you have all of enterprises, then you have consumers, and then you could argue a fourth S curve, which is agents.
- 1:13
- Unidentified speaker: So some smart people would argue that your camera's not actually 8 billion, it's 8 billion people and they're agents.
- 1:20
- Unidentified speaker: So in terms of the adoption curve, we started this conversation talking about we're in the first inning of AI.
- 1:26
- Unidentified speaker: It's reported that pre-seed round today on average $100 million valuation.
- 1:30
- Unidentified speaker: How do you understand?
- 1:33
- Unidentified speaker: I think you've been in around venture for nearly two decades with firms like Vein Capital Ventures, Samsung.
- 1:45
- Unidentified speaker: And you believe that AI is in its first inning.
- 1:49
- Unidentified speaker: Why?
- 1:49
- Unidentified speaker: Well, I think specifically enterprise AI is really just starting to scratch the surface of what it can do within enterprise.
- 1:56
- Unidentified speaker: Um, what we've seen at large enterprises, i.e., the customers of our portfolio companies, is that they're either scared to adopt or slow to adopt or have high failure rates when they adopt AI.
- 2:06
- Unidentified speaker: So I think because of all we're seeing from our customer base through the portfolio lens, um, I think there's so much more to come.
- 2:12
- Unidentified speaker: And I'm excited to see what it can do.
- 2:14
- Unidentified speaker: Which one of those is the real issue?
- 2:17
- Unidentified speaker: Being scared to adopt or being slow to adopt or just being poor poor at adopting today?
- 2:21
- Unidentified speaker: Aaron Powell It's a new muscle for a lot of enterprises.
- 2:23
- Unidentified speaker: I mean, they've gone through data transformation over the many years, they've uh built sort of database architecture, and now as they start to use that foundational data to adopt AI best practices, it's all new for them.
- 2:35
- Unidentified speaker: And so they either need external help.
- 2:37
- Unidentified speaker: Uh, could be a startup, could be a growth company, could be a consultant.
- 2:41
- Unidentified speaker: But internally, their resources are generally pretty poor right now, from what we've seen.
- 2:45
- Unidentified speaker: 95% of enterprise AI deployments are failing these days.
- 2:50
- Unidentified speaker: And that's I think that's per Deloitte.
- 2:51
- Unidentified speaker: So um that only has a lot of room to scale, a lot of room to grow.
- 2:54
- Unidentified speaker: And obviously, if you've done something a few times, as in deployed AI across different verticals or different data sets inside of an enterprise, you only get better over time.
- 3:02
- Unidentified speaker: Aaron Powell That number you said from Deloitte, 95% of pilots fail.
- 3:06
- Unidentified speaker: I haven't heard that number, but I experienced it firsthand.
- 3:10
- Unidentified speaker: And you're seeing you're seeing these hundreds of enterprises implementing AI over and over every day.
- 3:16
- Unidentified speaker: What are the best practices in terms of the 5% that do succeed?
- 3:19
- Unidentified speaker: If they come from a deeply technical background that's already been doing deep data analysis and has great data infrastructure plus security, then they have the foundation of success, right?
- 3:29
- Unidentified speaker: So that's like already tipping the scale in the favor of tech companies that are pretty well poised to do that.
- 3:35
- Unidentified speaker: Some of the legacy companies, legacy industries that are lower tech, um, we've seen implementations where once we get under the covers, the data is pretty unruly.
- 3:44
- Unidentified speaker: And so getting the data to a point where, from an infrastructure standpoint, it's ready to adopt AI is half the battle.
- 3:51
- Unidentified speaker: Now, I think once we cross that chasm, a lot of these AI best practices can start to proliferate and accelerate.
- 3:57
- Unidentified speaker: What does it mean to have good data versus bad data when it comes to implementing AI on an enterprise level?
- 4:02
- Unidentified speaker: You see in large enterprises, right, they're the larger the enterprise, the more complex, generally speaking, and they have lots of data silos.
- 4:08
- Unidentified speaker: That means data tagging, fidelity across different uh fields and tags can be different.
- 4:14
- Unidentified speaker: And so when you try to run enterprise grade AI on top of it, you have all sorts of sometimes conflicting um tags and data structures, which have to be resolved before you can really do it in a meaningful way.
- 4:25
- Unidentified speaker: And so those are some of the data transformation projects that are foundational to then running great AI and analytics on top of enterprise.
- 4:32
- Unidentified speaker: Aaron Powell Something that I hear over and over about AI implementation is that it's mostly a cultural issue.
- 4:39
- Unidentified speaker: Yeah, sure, there's data, there's processes, but starts at the culture level.
- 4:43
- Unidentified speaker: How much does culture play a role in terms of enterprises being able to implement AI within the organization?
- 4:49
- Unidentified speaker: That's right.
- 4:49
- Unidentified speaker: And it starts from the top uh within organizations, right?
- 4:52
- Unidentified speaker: If if senior leadership is embracing it, it has to be in a lot of cases top down to reach sort of the most mission critical uh parts of the enterprise.
- 5:00
- Unidentified speaker: And so, you know, some leaders, especially within tech companies, are thinking about this on a daily basis.
- 5:05
- Unidentified speaker: And I can tell you for a fact that some companies leading huge banks, huge consumer companies, huge airlines are not native to this.
- 5:11
- Unidentified speaker: And so there's a lot of education and foundational work that needs to happen for them to really reap the benefits of AI's power and uh and possibility.
- 5:21
- Unidentified speaker: This episode is presented by Juniper Square, the operations partner for private markets.
- 5:27
- Unidentified speaker: So if I'm the CEO of Delta, let's say I just put my foot in my mouth with a Starlink and saying that I'm don't want Starlink and don't like SpaceX, all these things.
- 5:37
- Unidentified speaker: But let's say I want to rehash my image and I want to show that I'm leaning into technology and leaning into AI.
- 5:43
- Unidentified speaker: Is that the problem set that you would go into, which is elastic pricing for seats?
- 5:49
- Unidentified speaker: What would you advise the Delta CO to do?
- 5:51
- Unidentified speaker: What's the next steps?
- 5:53
- Unidentified speaker: It's looking at a lot of historical data to understand how they've approached it in the past, but then also having a horizontal view of pricing within the industry.
- 6:01
- Unidentified speaker: One of the ways you can get a data moat, which is very important these days, is to have horizontal uh insights about an industry, meaning you see hundreds of airlines.
- 6:11
- Unidentified speaker: The airline you're talking to probably sees mostly its own data most of the time.
- 6:16
- Unidentified speaker: And so when you bring horizontal insights around best practices and price elasticity and uh price per mile, what type of seat uh across different uh organizations, you can then bring that to bear within large uh high willingness to pay customers.
- 6:29
- Unidentified speaker: And so I think it's A, looking at their own historical data, but B gleaning a lot of model insights across thousands of other uh potential opportunities.
- 6:38
- Unidentified speaker: This connects to what we were talking to a couple of weeks ago, which is you said that the vast majority of data is not accessible to LLMs.
- 6:47
- Unidentified speaker: Why is that?
- 6:47
- Unidentified speaker: One of our favorite stats, and it actually came through an enterprise that we work very closely with, which is that about 80% or more of the world's data is behind firewalls.
- 6:56
- Unidentified speaker: Um LLMs have done an absolutely heroic job, right?
- 6:59
- Unidentified speaker: Going from uh infantile phase a few years ago to in the pockets of almost every American and around the world.
- 7:06
- Unidentified speaker: Uh so would never speak ill of what they've accomplished, but they've accomplished it with the publicly available data, similar to how Google crawls the public web.
- 7:14
- Unidentified speaker: Um to get a lens into private data, you have to earn customer trust, you have to bring unique insights, and oftentimes you need a subject matter expert to be able to have the trust and expertise to enter a market like airlines or banking or insurance.
- 7:28
- Unidentified speaker: When we see companies that are starting to unlock those types of value, they get access in a highly secure way to proprietary behind-the-firewall data, the 80%.
- 7:38
- Unidentified speaker: When they do that, they can start to produce new insights for the company and for the industry they operate in.
- 7:43
- Unidentified speaker: And that's where really high value opportunities exist.
- 7:46
- Unidentified speaker: Years ago, we were just pure play enterprise software investors.
- 7:50
- Unidentified speaker: Now we always see those companies come with an AI strategy.
- 7:54
- Unidentified speaker: The best of those companies come with an enterprise software workflow that systematically helps companies to unlock their data and then run AI best practices on it in a highly secure way.
- 8:05
- Unidentified speaker: You made a very important but subtle point, which is the reason 80% of the data is behind the enterprise firewalls is because it comes from the customer.
- 8:14
- Unidentified speaker: And in order to get that customer data, you must provide a good product.
- 8:17
- Unidentified speaker: So everything is still downstream of having great product.
- 8:20
- Unidentified speaker: The difference in AI world today is that your data starts to be one of the most sustainable forms of competitive advantage.
- 8:27
- Unidentified speaker: So again, 95% of enterprises, according to Deloitte, fail in terms of their pilots.
- 8:33
- Unidentified speaker: The ones that are succeeding, what are some early wins that you're seeing in enterprise that are maybe not as obvious to the outside?
- 8:40
- Unidentified speaker: Enterprise software that is creating workflows that unlock new data feels like the oil rig company pulling new oil out of the ground for enterprise.
- 8:50
- Unidentified speaker: And what we're seeing is from inside the enterprise.
- 8:53
- Unidentified speaker: So one of my favorite ones to talk about is Inspirin, which is a portfolio company, just closed $170 million from Insight and New View Ventures.
- 9:02
- Unidentified speaker: And they are in the healthcare space, which is a space that generally speaking, if you asked me 10 years ago, Joe, are you going to invest in the healthcare space?
- 9:08
- Unidentified speaker: I probably would have said no.
- 9:09
- Unidentified speaker: But their business model, their workflow, the data that they extract is perfectly aligned with our thesis.
- 9:16
- Unidentified speaker: And the reason for that is that they have a very low cost device that they place across senior living facilities.
- 9:22
- Unidentified speaker: And it uses computer vision and other sensors to extract real-time data 24-7 about residents.
- 9:29
- Unidentified speaker: That data set prior to this device was primarily anecdotal.
- 9:33
- Unidentified speaker: So it's what is the nurse saying, what is the doctor saying, what have we logged, right?
- 9:37
- Unidentified speaker: This is a human problem.
- 9:38
- Unidentified speaker: Humans are not great at this organizational level of detail.
- 9:41
- Unidentified speaker: Um, nor can they go as granular as 24-7.
- 9:44
- Unidentified speaker: So what is 24-7 monitoring unlock?
- 9:47
- Unidentified speaker: Well, first of all, you can do it in a very anonymized, safe way.
- 9:50
- Unidentified speaker: So that's that has to be table stakes, right?
- 9:52
- Unidentified speaker: Um and then the second thing is you can unlock insights about patients, doctors, nurses, staff members over time.
- 10:01
- Unidentified speaker: And so that helps you do things like uh prevent falls.
- 10:04
- Unidentified speaker: If you're monitoring patients 24-7 and you see somebody that you have a known fall risk, right, that can be built into the um into the AI.
- 10:12
- Unidentified speaker: You can monitor them in the middle of the night.
- 10:14
- Unidentified speaker: If they stand up or they look like they're about to stand up, you can send a staff member minutes before they otherwise would have arrived to help that person with whatever they need in the night.
- 10:24
- Unidentified speaker: Could be a question, could be a run to the restroom.
- 10:26
- Unidentified speaker: But either way, if that person is standing up under medication in the middle of the night and they're 85 years old, the uh negative outcome is a fall, a broken hip, a lawsuit.
- 10:35
- Unidentified speaker: That's a tens of millions of dollars problem eliminated.
- 10:38
- Unidentified speaker: So inspiring, one of the first use cases, was both in hospital and senior living, was they were able to show they were could eliminate 60 to 70 percent of falls that would happen oftentimes in the night.
- 10:48
- Unidentified speaker: The second thing is you have longitudinal data now.
- 10:51
- Unidentified speaker: So you're watching uh seniors or occupants over time and how their activity changes.
- 10:58
- Unidentified speaker: What that can do is actually do predictive analytics and start to predict things like dementia and Alzheimer's based on activity levels, behavior levels, uh, and things like that.
- 11:08
- Unidentified speaker: We can also predict if somebody starts um exhibiting certain behaviors, things like infection, which is incredible.
- 11:14
- Unidentified speaker: And um, lastly, we're now creating a sort of time series global view of an of an entire operation.
- 11:21
- Unidentified speaker: So, you know, senior living facility may have hundreds of rooms, you know, dozens of staff members, dozens of doctors and nurses.
- 11:28
- Unidentified speaker: Now they get to see in real time who's doing what and where.
- 11:32
- Unidentified speaker: So where are the resources being deployed?
- 11:34
- Unidentified speaker: Logistics, essentially.
- 11:35
- Unidentified speaker: Right.
- 11:35
- Unidentified speaker: And so that was anecdotal before.
- 11:37
- Unidentified speaker: What and so you know, you would ask a staff member um how their day was, and they'd say, Oh, it was just, I was, I was so busy, way understaffed, right?
- 11:45
- Unidentified speaker: Classic response.
- 11:46
- Unidentified speaker: Maybe they noted it in their logbook, maybe they put it in EHR, maybe not.
- 11:50
- Unidentified speaker: Now we can actually say, wow, Joe, the staff member, was supposed to visit 10 rooms today, but he got pulled into 25 rooms.
- 11:57
- Unidentified speaker: That's a major problem.
- 11:59
- Unidentified speaker: By the way, Joe as a staff member is likely to leave if we're working two and a half times hard.
- 12:03
- Unidentified speaker: Yeah.
- 12:04
- Unidentified speaker: And so we're also eliminating burnout among staff because we can load balance according to the needs of the patients.
- 12:10
- Unidentified speaker: And lastly, you know, over time patients and and occupants start to exhibit signs of additional needs, right?
- 12:16
- Unidentified speaker: Like they tend to degrade, especially in senior living.
- 12:18
- Unidentified speaker: And so uh you want to appropriately staff up based on the patient's needs.
- 12:23
- Unidentified speaker: So if my father went into senior living tomorrow, he might need one visit a day.
- 12:27
- Unidentified speaker: But in two years, he might need 10 visits a day.
- 12:29
- Unidentified speaker: That is a different, different value prop.
- 12:31
- Unidentified speaker: That's a different staffing mechanism.
- 12:33
- Unidentified speaker: And my father might need to be moved to a different floor to achieve that, where he can get the level of care that he actually needs.
- 12:38
- Unidentified speaker: So it results in better staff satisfaction, better patient outcomes, better profitability at the end of the day.
- 12:45
- Unidentified speaker: And an all new data set that was previously anecdotal at best is now digitized moment by moment across the entire facility.
- 12:53
- Unidentified speaker: So that's that's oil, as I said, being pulled out of the ground.
- 12:56
- Unidentified speaker: It was sitting there, it's always been there.
- 12:58
- Unidentified speaker: But we've in SpyRon has pulled that data out of the ground and created a very sophisticated model to help multiple decisions, whether it's tactical decisions like preventing falls, or operational decisions like staffing decisions and care level decisions.
- 13:13
- Unidentified speaker: SpyRIN is a fascinating case study.
- 13:15
- Unidentified speaker: It's kind of like a minority report, but it's actually helping the patients.
- 13:18
- Unidentified speaker: And also in many ways, it's this gateway drug to physical AI.
- 13:22
- Unidentified speaker: Jensen Wang predicts that physical AI would be a $50 trillion market.
- 13:26
- Unidentified speaker: A lot of the top VCs that I know right now are trying to invest into physical AI because they see it as a new frontier once kind of all the software and all the uh databased AI gets integrated.
- 13:38
- Unidentified speaker: And most of that application is actually building from scratch.
- 13:42
- Unidentified speaker: So humanoids or something, sensors and factories, things like that.
- 13:47
- Unidentified speaker: But here it's actually going into existing systems and getting existing insights from systems versus kind of building these brand new technologies from scratch.
- 13:58
- Unidentified speaker: Yeah.
- 13:58
- Unidentified speaker: And all the new technologies you just mentioned live and exist in the future based on ingesting the data we're capturing now.
- 14:07
- Unidentified speaker: So imagine that instead of having a staff member, you had a robotic assistant.
- 14:11
- Unidentified speaker: Well, how is that robotic assistant going to learn over time?
- 14:14
- Unidentified speaker: Well, wouldn't it be great if you had a 24-7 data set of video capture for thousands of patients over thousands of days?
- 14:22
- Unidentified speaker: That sounds like the perfect training desk data set for that type of use case.
- 14:26
- Unidentified speaker: And so obviously the impact right now at Inspire and is tangible, but the data set they're creating is hasn't even been monetized.
- 14:34
- Unidentified speaker: And that's one of the things that's exciting for us.
- 14:36
- Unidentified speaker: I imagine they're going to be monetizing not per seat or not per device, but if you are, I think you mentioned 60 to 75% of falls are prevented with the device.
- 14:45
- Unidentified speaker: If you're able to keep these large, you know, multi-hundred thousand dollar lawsuits from happening over and over.
- 14:51
- Unidentified speaker: That's extremely profitable to the healthcare centers.
- 14:54
- Unidentified speaker: And if you monetize off of the losses that you're preventing, that that could be quite a sizable business.
- 14:59
- Unidentified speaker: Yeah, today we actually do have kind of a SaaS pricing model.
- 15:02
- Unidentified speaker: So like a per-room uh deployment pricing model.
- 15:05
- Unidentified speaker: Um, but you're right.
- 15:07
- Unidentified speaker: I think in the selling process, that is one of the most tangible aspects uh to get in the door in the first place and then to win a pilot.
- 15:15
- Unidentified speaker: But what keeps us there over the long run is that the staff, the administrators, the owners all become addicted to this new data.
- 15:24
- Unidentified speaker: Once you've pulled that new oil out of the ground, uh putting it back would be pretty difficult, right?
- 15:28
- Unidentified speaker: You've already realized these staff retention benefits, these profitability benefits, and these patient improvement benefits in addition to those first use cases, which is hey, if we could prevent falls, would that be helpful?
- 15:39
- Unidentified speaker: Right?
- 15:39
- Unidentified speaker: That gets us in the door.
- 15:40
- Unidentified speaker: What keeps people is that um that long-term reliance on this new data set.
- 15:45
- Unidentified speaker: It's a fascinating example of physical AI.
- 15:48
- Unidentified speaker: In terms of physical AI, maybe you could double-click on additional use cases you're seeing today in your portfolio.
- 15:53
- Unidentified speaker: There's certainly some things to be uh fearful of in AI, but when you see tangible impacts helping people's lives, um, it starts to it it's the other side of the coin, right?
- 16:05
- Unidentified speaker: It's like you can't get all the positives without the negatives.
- 16:07
- Unidentified speaker: The negatives can be mitigated with cybersecurity and hardening and data protection and things like that, obviously personal uh personal protections as well.
- 16:15
- Unidentified speaker: Um but when you see the positive impact in something like healthcare uh or a company in our first portfolio called Bark, which protects kids online using natural language processing.
- 16:25
- Unidentified speaker: So we're ingesting about 7 million kids' data real time across multiple apps and across the OSs to establish emotional baselining for those kids.
- 16:35
- Unidentified speaker: We can then predict with very high accuracy when somebody's experiencing uh depressive episodes, uh, or specifically has flagged, even if they're using tons of slang, uh, you know, self-harm, harm to others, or if a child predator has entered the chat.
- 16:49
- Unidentified speaker: And so the impact coming out of that company, uh, as somebody who has three kids especially gives me shivers every time I open it, right?
- 16:56
- Unidentified speaker: The uh the VC muscle memory tells you to open up a company update and say, what was the revenue?
- 17:03
- Unidentified speaker: What was the growth?
- 17:04
- Unidentified speaker: But in this case, I go straight to the impact.
- 17:06
- Unidentified speaker: Uh, and that's what gives me shivers.
- 17:08
- Unidentified speaker: It's dozens of school shootings, hundreds of suicides prevented, child predators escalated to the FBI dozens every year.
- 17:15
- Unidentified speaker: Um and so doing something like protecting kids with a complex natural language processing engine plus predictive analytics and AI is just another use case where um I think you can get behind AI if you see use cases like that.
- 17:29
- Unidentified speaker: But it again, there's plenty to be fearful of, and I think there's plenty of companies being built to protect against some of the dark side of AI.
- 17:37
- Unidentified speaker: Um, and that's a thesis that we're leaning into going forward.
- 17:40
- Unidentified speaker: Um and then on your physical AI question, um, I would cite a company called Evercoast, which does 4D meaning in-motion capture of primarily human beings.
- 17:52
- Unidentified speaker: And they use a camera array, which is um pretty much off-the-shelf cameras with very sophisticated processing capabilities in the background to reconstruct human beings or objects in motion.
- 18:04
- Unidentified speaker: So effectively video in 4D.
- 18:07
- Unidentified speaker: Um, and those the early use cases for that business were primarily in the media space.
- 18:11
- Unidentified speaker: So we worked with the Olympics and captured every Olympian in Paris, for example, doing their specialty with a 4D camera array around them.
- 18:19
- Unidentified speaker: That created digital humanoid reproducible assets from those gymnasts or from those uh from those gymnasts and and other athletes that NBC could use in other media um generatively.
- 18:31
- Unidentified speaker: And so that was the initial use case.
- 18:33
- Unidentified speaker: We've done that now um tens of thousands of times with hundreds of customers, uh including Apple, Netflix, and others, and have a really nice business doing that capture processing asset management and dovetailing it with the production workflows.
- 18:51
- Unidentified speaker: That business is interesting, right?
- 18:53
- Unidentified speaker: They're already making millions of dollars in revenue.
- 18:55
- Unidentified speaker: What's really interesting is they have the world's largest data set of human 4D captured uh assets.
- 19:01
- Unidentified speaker: And why does that matter?
- 19:03
- Unidentified speaker: Exactly what we said before.
- 19:04
- Unidentified speaker: Robotics training, right?
- 19:06
- Unidentified speaker: Generative humans for robotics training, real world applications.
- 19:10
- Unidentified speaker: Um and so to do that at scale, you can imagine a data set like Evercos becoming a pipeline uh to create generative human beings at scale to further train robots, self-driving cars, uh, you know, the digital nursing example we gave earlier, home care, et cetera.
- 19:29
- Unidentified speaker: And so that's sort of step two of unlocking the data set there.
- 19:32
- Unidentified speaker: And, you know, notably, that data set is very valuable to a lot of folks.
- 19:36
- Unidentified speaker: And so we have to be very careful about like who do we partner with?
- 19:39
- Unidentified speaker: Do we sell the data?
- 19:40
- Unidentified speaker: Do we build our own internal models?
- 19:42
- Unidentified speaker: And so that's a conversation we have almost weekly with the founder there.
- 19:44
- Unidentified speaker: One thing I've learned from talking to hundreds of investors is that great investment firms aren't built on investment returns alone.
- 19:50
- Unidentified speaker: The firms that under are great at the things most people don't see their operations, their relationship with LPs, and the quality of information they use to make decisions.
- 19:58
- Unidentified speaker: And here's what AI has changed.
- 20:01
- Unidentified speaker: Every firm now has access to the very same models, so the intelligence isn't the edge anymore.
- 20:06
- Unidentified speaker: The edge is what you could feed it.
- 20:07
- Unidentified speaker: A firm with its fund operations and data in one connected record can actually put AI to work.
- 20:13
- Unidentified speaker: A firm running on disconnected systems simply can't.
- 20:17
- Unidentified speaker: That's why thousands of GPs run their funds on Juniper Square.
- 20:20
- Unidentified speaker: Juniper Square puts your fund operations, data, and administration together in one connected record.
- 20:25
- Unidentified speaker: That means less time managing disconnected systems and more time investing, working with LPs, and building your firm.
- 20:32
- Unidentified speaker: This episode is brought to you by JuniperSquare, the operations partner for private market GPs.
- 20:38
- Unidentified speaker: Learn more at junipersquare.com slash how I invest.
- 20:41
- Unidentified speaker: That's junipersquare.com slash how I invest.
- 20:44
- Unidentified speaker: A lot of people have trouble thinking second order effects or third order effects, but as the LLMs continue to get better and the inference continues to be more and more sophisticated, the moat will continue to be the data.
- 20:57
- Unidentified speaker: Right.
- 20:57
- Unidentified speaker: Just think about Tesla, for example, with their self-driving car and their self-driving technology.
- 21:02
- Unidentified speaker: I think it took a decade for them to gather the information, just driving on the streets of Austin, San Francisco.
- 21:08
- Unidentified speaker: And that was the bottleneck, it was the 10 tens of year, uh, 10 years.
- 21:11
- Unidentified speaker: And then obviously they had to build the models, which was non-trivial at the time.
- 21:14
- Unidentified speaker: This was pre-Chat GPT.
- 21:16
- Unidentified speaker: Now I'm sure they're using a lot more LLMs.
- 21:18
- Unidentified speaker: But in a post-LLM world where you have all the models where you have basically genius level physics, PhDs, you know, 10,000 of them that could work on your problem set, the data, the moat actually becomes the data.
- 21:33
- Unidentified speaker: We hear this word a lot in our kind of investment committee, which is you know, either flywheel, which is people use a lot, but we also talk about like pro-cyclical advantage, meaning once you're Tesla and you get a lot of data, then your model becomes better and you get more users, and then you get even more data.
- 21:51
- Unidentified speaker: That goes back to the customers being the source of the data.
- 21:53
- Unidentified speaker: Right.
- 21:54
- Unidentified speaker: You have to prime the pump.
- 21:55
- Unidentified speaker: Sometimes that's costly.
- 21:56
- Unidentified speaker: In Tesla's case, it was building a car that had autonomous driving features years ago.
- 22:00
- Unidentified speaker: Now it's a pro-cyclical event where every incremental user often chooses Tesla because of its self-driving, and so they further cement their data lead.
- 22:10
- Unidentified speaker: So that's a pro-cyclicality we talk about a lot, um, where that becomes part of your defensibility over time.
- 22:16
- Unidentified speaker: A new entrant, starting at zero or even even above zero, will still be so far behind a Tesla or an Evercoast in this example to catch up.
- 22:26
- Unidentified speaker: And so incremental users choose those companies again and again and further deepen their data mode.
- 22:32
- Unidentified speaker: I want to go back to this question of business models.
- 22:35
- Unidentified speaker: I don't think there's been a technological revolution that has not led to significantly new business models, but yet people are continuing to charge in this kind of SaaS SaaS model.
- 22:43
- Unidentified speaker: What do you see in terms of business models and how they evolve in a post-AI world?
- 22:49
- Unidentified speaker: Yeah, I think you have to meet your customer where they're comfortable, especially when you're a startup.
- 22:54
- Unidentified speaker: Um, so will SaaS modeling be the traditional or most frequent use case in the future?
- 23:00
- Unidentified speaker: Possibly, possibly not.
- 23:02
- Unidentified speaker: But right now, that's where large enterprises have been trained to purchase and been, you know, when they think about unit economics, going back to the inspiring example, you know, how much is our customer paying us per month?
- 23:13
- Unidentified speaker: How much does this cost us per month?
- 23:15
- Unidentified speaker: What's our cost savings?
- 23:16
- Unidentified speaker: What's our cost savings per user?
- 23:18
- Unidentified speaker: Unit economics leads you to think about it in terms of a cost per unit.
- 23:23
- Unidentified speaker: Oftentimes that rolls up to a SaaS pricing model.
- 23:26
- Unidentified speaker: Um now in more avant-garde tech, you might see usage-based models, which we're already seeing a lot of.
- 23:31
- Unidentified speaker: You may see even outcomes-based models, right?
- 23:33
- Unidentified speaker: Where if we can produce this for you, large enterprise, you'll give us this, right?
- 23:38
- Unidentified speaker: Um, those are a little bit harder sells.
- 23:40
- Unidentified speaker: I mean, I think that they're very aligned with outcomes-driven and ROI-driven implementations.
- 23:45
- Unidentified speaker: So don't get me wrong, I kind of love that.
- 23:47
- Unidentified speaker: But the buying behavior is new.
- 23:50
- Unidentified speaker: And so there's some education and some comfort level, particularly with large enterprises, some of whom have been around for a hundred years, right?
- 23:56
- Unidentified speaker: Um and certainly over the last 20 years, they've been habituated to price based on SaaS per seat or per month.
- 24:04
- Unidentified speaker: Um, so getting them out of that will take some time.
- 24:06
- Unidentified speaker: I don't think it's impossible, but I I think it'll we'll see a slow ramp away from it.
- 24:10
- Unidentified speaker: What comes next?
- 24:12
- Unidentified speaker: I think the usage-based, consumption-based, token-based models are all things that are in the conversation right now.
- 24:18
- Unidentified speaker: What we've seen is some hybrid approaches, so some SaaS plus utilization.
- 24:22
- Unidentified speaker: We've seen some pure play utilization or token-based models.
- 24:26
- Unidentified speaker: Um, those can be scary for the CEO, CFO of a large enterprise because you have uh some opacity around what am I going to pay?
- 24:34
- Unidentified speaker: You know, I know what my budget is.
- 24:36
- Unidentified speaker: Uh, we've all bought into the budget.
- 24:38
- Unidentified speaker: We told our board our budget's X.
- 24:40
- Unidentified speaker: What if we exceed that?
- 24:41
- Unidentified speaker: Right?
- 24:41
- Unidentified speaker: That can create problems.
- 24:43
- Unidentified speaker: Um, and so understanding that and creating visibility is important for enterprises to really get comfortable with that buying behavior.
- 24:50
- Unidentified speaker: I think you can cap it, you can tier it.
- 24:52
- Unidentified speaker: That creates more hybrid approaches that look sort of like SaaS plus utilization.
- 24:56
- Unidentified speaker: Um, that may be the on-ramp process that gets them comfortable over time to say, look, we've been doing this for two years now.
- 25:03
- Unidentified speaker: Our utilization is X.
- 25:04
- Unidentified speaker: We've paid a million dollars a year for the last three years.
- 25:07
- Unidentified speaker: This year we'll probably use it 20% more.
- 25:09
- Unidentified speaker: So we've budgeted $1.2 million.
- 25:11
- Unidentified speaker: That'll take some time, right?
- 25:12
- Unidentified speaker: They need to have the data behind that and the comfort level to make that type of decision.
- 25:17
- Unidentified speaker: But um, you know, it's how most markets develop over time, and I don't think it'll be different here.
- 25:21
- Unidentified speaker: Sitting where you're sitting, a couple of weeks ago, I had David Breach, the Vista Equity president.
- 25:26
- Unidentified speaker: They've deployed over $100 billion in software.
- 25:29
- Unidentified speaker: So they certainly thought long and hard on this issue.
- 25:32
- Unidentified speaker: And he believes that the future of software is not just selling software, but selling solutions.
- 25:36
- Unidentified speaker: It's kind of do-it-yourself versus done-for-you model.
- 25:40
- Unidentified speaker: Do you see that as the evolution in software itself?
- 25:42
- Unidentified speaker: In the early days, we're seeing a lot of that.
- 25:45
- Unidentified speaker: Um, consultative selling.
- 25:47
- Unidentified speaker: I think we've seen portfolio companies with very powerful technology uh enter Fortune 500 companies, take the covers back, or open the hood and be surprised by how disparate and unruly the data is.
- 26:02
- Unidentified speaker: And that very quickly turns into a data consulting project.
- 26:04
- Unidentified speaker: For a VC, that's not what I love to hear, right?
- 26:06
- Unidentified speaker: We want to hear this this will work out of the box, right?
- 26:10
- Unidentified speaker: Um, and so I think you have to have um your customers have to continue to deploy best practices in in data management, fidelity, uh, infrastructure in order to unlock the plug and play value that a lot of portfolio companies and software companies will will ultimately bring.
- 26:28
- Unidentified speaker: But the on-ramp in the early days can be consultative.
- 26:31
- Unidentified speaker: Um that that can be a positive to the startup in our world because they learn a tremendous amount about the problems that the enterprise is feeling.
- 26:39
- Unidentified speaker: Once they've felt that and solved that, they can then replicate it and even create agentic workflows to replicate it across other use cases.
- 26:45
- Unidentified speaker: So that the next Fortune 500 they go into takes 10 times is 10 times faster for the implementation.
- 26:51
- Unidentified speaker: But these are baby steps, I think, in the process of really unlocking the true enterprise value that will come in in years.
- 26:57
- Unidentified speaker: Reminds me of the polygram advice, which is do the unscalable.
- 27:00
- Unidentified speaker: If you want to build a build great business, do the unscalable, figure out what the problem is, and then automate.
- 27:06
- Unidentified speaker: Same with the Elon algorithm.
- 27:07
- Unidentified speaker: A lot of people think the Elon algorithm starts with automation, actually ends with automation.
- 27:12
- Unidentified speaker: Because the number one problem you could have in engineering is that you automate something that doesn't actually add value.
- 27:16
- Unidentified speaker: So you have to figure out, make sure that it adds value, run it, kind of make sure that that ROI is done.
- 27:21
- Unidentified speaker: And then and only after then, then you automate.
- 27:23
- Unidentified speaker: People have this, especially venture capitalists, have this bias, oh, I want something scalable, I want something automatable from day one.
- 27:29
- Unidentified speaker: But sometimes it can be counterproductive even to the design of the product itself.
- 27:33
- Unidentified speaker: Yeah, potentially.
- 27:34
- Unidentified speaker: And and I think like when you think about funding startups, you know, we always enter at seed.
- 27:38
- Unidentified speaker: I think you need to create some realistic benchmarking and understanding that that will happen.
- 27:45
- Unidentified speaker: Um and that leads to things like, you know, how much cash flow do you need to achieve a particular thing, uh, a particular outcome?
- 27:51
- Unidentified speaker: So what does it take to reach a series A, series B?
- 27:54
- Unidentified speaker: Um, I think on one hand, from a product standpoint, you can reach product maturity faster than ever.
- 28:01
- Unidentified speaker: So like the dollars that we're putting into companies are being used with an efficiency that has been never seen before.
- 28:06
- Unidentified speaker: Double-click on that.
- 28:07
- Unidentified speaker: Why why are companies able to scale so much quicker with so much less capital?
- 28:11
- Unidentified speaker: I mean, I think it starts when you're thinking about product and code.
- 28:14
- Unidentified speaker: I think agentic coding has been massively impactful, especially to people who know how to wield it.
- 28:19
- Unidentified speaker: So early stage founders that are technical, that have engineering backgrounds themselves, are the people that we've seen in sort of the data research we've done that are benefiting the most.
- 28:28
- Unidentified speaker: The most efficiency right now is being accrued to the engineers.
- 28:32
- Unidentified speaker: Um, you know, I think next on the list is something like customer service, which is just a great data set, great use case that you can augment quickly.
- 28:39
- Unidentified speaker: It's digestible.
- 28:40
- Unidentified speaker: But when you think about like the biggest delta in efficiency, it's at engineering.
- 28:45
- Unidentified speaker: And so talking to large enterprises, you know, we have a number of advisors, technical advisors on our team that work at places like Meta, Amazon, Oscar Health.
- 28:53
- Unidentified speaker: And when we talk to them, we say, what does your job look like now as an engineering or AI manager?
- 28:59
- Unidentified speaker: And they say, I can accomplish four times as much in half the time.
- 29:02
- Unidentified speaker: And so when we have some.
- 29:04
- Unidentified speaker: Yeah.
- 29:04
- Unidentified speaker: And so when you when you hear that, you start to think about, and we see this just in practice, right?
- 29:08
- Unidentified speaker: It's like the founders that we're backing are being way more cash efficient to get to product maturity in a shorter period of time.
- 29:15
- Unidentified speaker: And so that's part of the reason why we spend our time at seed.
- 29:17
- Unidentified speaker: I think the delta, the alpha creation opportunity from seed to A is that delta is greater than ever right now.
- 29:26
- Unidentified speaker: And so when we think about risk allocation and investment portfolio decisions, we think in sort of sharp ratios, right?
- 29:34
- Unidentified speaker: We're allocators and buyers of risk.
- 29:36
- Unidentified speaker: And for every unit of risk, we need to pay the appropriate price.
- 29:40
- Unidentified speaker: As things go from non-obvious pre-product to post-product and obvious, meaning some product market fit, customer references, revenue.
- 29:50
- Unidentified speaker: During that curve, that curve is happening shorter and faster than it's ever happened.
- 29:53
- Unidentified speaker: And as a result, you see series A's and series Bs being priced where series D's and E's were prior.
- 29:59
- Unidentified speaker: And so I wouldn't like to play that game as much as I'd like to capture the alpha going from non-obvious seed to relatively mature product in a matter of year, short, short number of years.
- 30:09
- Unidentified speaker: I don't think I've heard a single intelligent investor take the counter on AI disruptive, AI has intrinsic value.
- 30:17
- Unidentified speaker: I've I haven't found anyone to say, oh, it's all BS, it's all hype.
- 30:20
- Unidentified speaker: That being said, AI being valuable and all, it's reported that pre-seed rounds today on average $100 million valuation.
- 30:30
- Unidentified speaker: How do you underwrite that?
- 30:32
- Unidentified speaker: I think again, you're buying risk, you're allocating dollars behind some level of risk.
- 30:38
- Unidentified speaker: At early stage venture, these businesses are inherently risky.
- 30:42
- Unidentified speaker: So if that's true, it's very hard for my team to pay that kind of price.
- 30:46
- Unidentified speaker: Um because even if they're sophisticated, if they have the world's best team, you are still betting on an outcome that is extremely low probability.
- 30:54
- Unidentified speaker: And the numbers are sort of the numbers.
- 30:56
- Unidentified speaker: Now, those numbers have shifted, right?
- 30:58
- Unidentified speaker: Like the big outcomes are now tens of billions, not a billion.
- 31:02
- Unidentified speaker: Um so there is a world where that can happen, but it's still very low probability.
- 31:06
- Unidentified speaker: And so when I think about portfolio construction, those are the nice to have, not the need to have, in terms of how I underwrite.
- 31:12
- Unidentified speaker: Like if we're fortunate enough to get into a company that's a $20, $50 billion, $100 billion company, that's fantastic, obviously.
- 31:18
- Unidentified speaker: But we are underwriting a more rational, high probability outcome, and that gives us some price discipline at entry.
- 31:25
- Unidentified speaker: Um but I think that's why we also play in the intersection of what would be formerly enterprise software and AI core use cases, not in the sort of lab-based world that requires billions of investment into the top talent, which has just become more expensive over time.
- 31:42
- Unidentified speaker: Um we'd like to be the beneficiaries of that great work that's being done by those model companies without investing and owning that team, right?
- 31:51
- Unidentified speaker: This varies conversation with Matt Harris from Bain Capital Ventures, where I know you used to be.
- 31:55
- Unidentified speaker: And the heuristic you used is in evaluating, for example, anthropics valuation.
- 32:01
- Unidentified speaker: It's reported it's going to be $2 trillion valuation, is one way to think about it is just to take off three zeros.
- 32:07
- Unidentified speaker: So imagine taking off three zeros from two trillion, you get two billion.
- 32:11
- Unidentified speaker: And imagine taking three zeros from a hundred billion reported, kind of 2026 and end run rate.
- 32:18
- Unidentified speaker: So a hundred million.
- 32:19
- Unidentified speaker: And a 20x revenue is not that unreasonable.
- 32:22
- Unidentified speaker: It's just people get, and I think rationally so, or at least evolutionarily so, people get so hung up on the sticker shock that they can't imagine they're still being upside at 2 trillion.
- 32:33
- Unidentified speaker: But then when you look at the revenue, it starts to make more sense.
- 32:36
- Unidentified speaker: I think that's a fair argument.
- 32:37
- Unidentified speaker: Um, my counterpoint to that would be the probability of getting that type of outcome is just very, very long tail, right?
- 32:45
- Unidentified speaker: It's like a six sigma type outcome or a generational outcome.
- 32:48
- Unidentified speaker: Um, if you rewind 20 years, it would be like being the investor that found Google, right?
- 32:53
- Unidentified speaker: Those guys did okay.
- 32:54
- Unidentified speaker: Similar to that, the folks who found OpenAI, Anthropic, et cetera, will do very, very well.
- 33:00
- Unidentified speaker: Now, as I think with like my public market investor hat or my late stage investor hat, those companies are also facing new threats, both from each other, right?
- 33:09
- Unidentified speaker: Every model release uh is a step change in terms of functionality.
- 33:13
- Unidentified speaker: And they're fighting in sort of a duopolistic market right now, uh, but new entrants are not uh a foredrawn conclusion.
- 33:20
- Unidentified speaker: Like they may come, right?
- 33:21
- Unidentified speaker: Open weight models may come into uh higher use.
- 33:24
- Unidentified speaker: Foreign models that we that we aren't even in the conscious right now uh could come into the into the purview.
- 33:29
- Unidentified speaker: And so I think rising competition for a massive, right?
- 33:32
- Unidentified speaker: You're you're they've demonstrated the TAM and the growth of the TAM.
- 33:36
- Unidentified speaker: What you see in markets like that is that you have new entrants, lower cost entrants.
- 33:40
- Unidentified speaker: Um, the consumer willingness to pay or expectation to pay is between zero and ten dollars per month.
- 33:47
- Unidentified speaker: Like uh, and those are some of the early adopters.
- 33:50
- Unidentified speaker: Over time, that probably goes down while models and switching costs between models become uh switching costs become lower, right?
- 33:58
- Unidentified speaker: I've I switch between four to six models in a week and agents.
- 34:03
- Unidentified speaker: Um so switching costs are low.
- 34:05
- Unidentified speaker: I'm always chasing the best and fastest LLM for my use cases.
- 34:10
- Unidentified speaker: And so what that results in it potentially is that it's harder to monetize your huge user base because of competition.
- 34:16
- Unidentified speaker: And we're not even talking about some new entrant that could be the next anthropic.
- 34:20
- Unidentified speaker: Um if we rewind two years ago, we wouldn't have been talking about anthropic necessarily.
- 34:24
- Unidentified speaker: We would have been talking about open AI.
- 34:26
- Unidentified speaker: And looking at that graph can be a cautionary tale in the sense that open AI was inflecting at a rate that nobody'd ever seen, and then here comes anthropic and actually apexes it.
- 34:35
- Unidentified speaker: And so seeing that um is an interesting thing to watch, let's say.
- 34:40
- Unidentified speaker: Um I was not fortunate enough to invest in open AI or anthropic, but uh some of my LPs were, and they've asked me over the last several years like when to sell.
- 34:48
- Unidentified speaker: And I've given a similar argument for them, especially depending on where their basis is.
- 34:52
- Unidentified speaker: So I did invest in anthropic at the $4 billion valuation.
- 34:55
- Unidentified speaker: So maybe you could give me some advice because I'm trying to get a balanced view in terms of uh the upside and the downside moving forward.
- 35:02
- Unidentified speaker: How would you underwrite it today, October 2026?
- 35:06
- Unidentified speaker: Aaron Powell Well, I think I would look at the implied forward capture of the market, right?
- 35:10
- Unidentified speaker: So how big is the market likely to be, and then what percentage are you expecting to capture as anthropic vis-a-vis or relative to open AI and some unknown entrant or entrance?
- 35:21
- Unidentified speaker: So I think that's hard to handicap on the who's coming next, right, conversation.
- 35:25
- Unidentified speaker: Um It's unknown, unknown.
- 35:27
- Unidentified speaker: Who's gonna have the idea or the resource?
- 35:30
- Unidentified speaker: The unknown, unknown.
- 35:31
- Unidentified speaker: Um but you have other folks that are devoting a lot of resources to the space that have billions of dollars in reasons to be highly defensive, the Googles, the Metas, the Amazons, uh, et cetera, right?
- 35:43
- Unidentified speaker: And they they may take a different approach or a similar approach to building very um very robust models, but that will only up the ante in terms of competitiveness over time.
- 35:51
- Unidentified speaker: And then if you look at uh or make an assumption around whether there'll just be two players in the future, I think it's very unlikely that there will be two dominant players in the LLM world.
- 36:01
- Unidentified speaker: So if you start to model, and you can sort of stress test this, right, a few more entrants, and then you look at the market size, of course it's going to continue continue to grow.
- 36:10
- Unidentified speaker: What do you what's the implied capture rate of that future revenue in three, five years that Anthropic would have to have to justify their current valuation?
- 36:19
- Unidentified speaker: I think right now it is converging to a rational sort of pre-IPO price, so you're probably in a in a good spot, but that's absent my assumption around increased competition and lower cost players.
- 36:29
- Unidentified speaker: Now, I think the other angle or lens is what is there, who are the customers of all these LLMs?
- 36:34
- Unidentified speaker: OpenAI tends to be slightly more consumer.
- 36:36
- Unidentified speaker: Anthropic tends, especially with Claude Code, tends to be slightly more enterprise.
- 36:40
- Unidentified speaker: So for that reason, I would value every dollar of revenue differently across those two.
- 36:43
- Unidentified speaker: And so I give a little boost to Anthropic in that regard, and you can see that's probably why they've eclipsed OpenAI in terms of their growth rate.
- 36:49
- Unidentified speaker: I'll tell you the both case.
- 36:52
- Unidentified speaker: Virtue of Lucas Swisher, he runs Growth for Co, uh previous podcast guest, and he talked about the three S-curves of adoption.
- 37:00
- Unidentified speaker: So the first S-curve, the early adopters in terms of LLMs from a revenue standpoint have been developers.
- 37:07
- Unidentified speaker: What's crazy about that is that there's only only 50 million developers in the world out of 8 billion people.
- 37:14
- Unidentified speaker: And on top of that, only a small percentage of developers have actually implemented if you look at worldwide kind of engineers.
- 37:20
- Unidentified speaker: So we're still at really the small end of a small market.
- 37:24
- Unidentified speaker: Now, sure, enterprises have adopted it, but to your point, 95% of them have failed.
- 37:29
- Unidentified speaker: There's not that mass adoption.
- 37:32
- Unidentified speaker: There's not many enterprises that are super users today.
- 37:34
- Unidentified speaker: So we're still seem to be, and then obviously there's olive enterprises.
- 37:38
- Unidentified speaker: First, you have engineers, then you have olive enterprises, then you have consumers.
- 37:41
- Unidentified speaker: And then you could argue a fourth S-curve, which is agents.
- 37:46
- Unidentified speaker: So some smart people would argue that your TAM is not actually 8 billion, it's 8 billion people and their agents talk 57.
- 37:53
- Unidentified speaker: So in terms of the adoption curve, you we started this conversation talking about we're in the first inning of AI.
- 38:00
- Unidentified speaker: Well, I think you just made my point a little bit there, too.
- 38:02
- Unidentified speaker: I think there's an enterprise side, there's room to run, right?
- 38:05
- Unidentified speaker: There's definitely room to run.
- 38:06
- Unidentified speaker: So because you established that you know this market is growing and the TAM is expanding even from a very large base, it's extremely attractive.
- 38:13
- Unidentified speaker: And I think that will also attract new entrants, new players, whether it's you know, the Fang players or some player we haven't even started talking about.
- 38:21
- Unidentified speaker: Um but that could erode, you know, pricing power, right?
- 38:24
- Unidentified speaker: So it could erode in two ways, which is um over time you just see switching costs and you see the markets divided among more players.
- 38:30
- Unidentified speaker: Um you can see the entrance of sort of where we invest, which is bespoke enterprise software meets AI for specific industries that has foundational verticalization.
- 38:40
- Unidentified speaker: Yeah, that has more foundational um, you know, subject matter expertise baked into it, which then leads to trust within specific verticals like airlines, banking, insurance, et cetera.
- 38:50
- Unidentified speaker: Um and then over time, I think you also have a price and compression, which is like enterprises are going to be willing to pay more, but how much more will they be willing to pay?
- 38:59
- Unidentified speaker: In order for to really capture the line share of the wallet, you need to have much higher higher ROI than we're seeing right now in the in the first inning.
- 39:06
- Unidentified speaker: Um so it's sort of multi-factor.
- 39:08
- Unidentified speaker: It'll be fascinating to watch, but um, I don't think you can lose coming in where you can.
- 39:12
- Unidentified speaker: Also, investor in vertical AI solutions like Lagora.
- 39:16
- Unidentified speaker: How do you think about horizontal versus vertical?
- 39:20
- Unidentified speaker: Are there going to be different areas where vertical wins versus horizontal wins?
- 39:24
- Unidentified speaker: And what's your framework for that?
- 39:26
- Unidentified speaker: When we think about horizontal, we think about more utilities that are foundational to AI.
- 39:32
- Unidentified speaker: So it could be a company like Monitor, where it's helping any vertical or any one enterprise manage their own models.
- 39:39
- Unidentified speaker: It could be the data loss prevention uh example I gave with with sort of like where is my data flowing in and outside my enterprise.
- 39:46
- Unidentified speaker: That again would be a horizontal solution to avert potentially a vertical problem.
- 39:51
- Unidentified speaker: Um and then when I think about verticals that will win, um, you know, I think selling into any one of these verticals requires trust.
- 39:58
- Unidentified speaker: And especially in the early days, building that trust within an air within an industry as esoteric as airlines or insurance underwriting or banking underwriting, lending, um, that requires trust, understanding the customer process, the legacy process, understanding the data, the KPIs, et cetera, can give you an early lead to start building that data flywheel within legacy industries.
- 40:20
- Unidentified speaker: And so when we think about vertical uh AI, I think it's you know, companies that have domain expertise plus uh the ability to build product very, very fast.
- 40:29
- Unidentified speaker: So engineering-led teams that have AI-native workflows that also have deep understanding of what we call subject matter experts on the team that can earn the trust because they have some sort of uh asymmetric advantage to penetrate a market.
- 40:44
- Unidentified speaker: Um and so we're oftentimes like underwriting teams uh in that way, which is technical prowess, what have they built today?
- 40:51
- Unidentified speaker: How fast can they build their future roadmap?
- 40:54
- Unidentified speaker: But then also, what does the go-to-market motion look like?
- 40:57
- Unidentified speaker: Is it appropriate for their ACV, for example?
- 41:00
- Unidentified speaker: Um, so are they being rational about how they scale?
- 41:03
- Unidentified speaker: Um, but then the next thing is you know, we don't want to fight in a fair fight, right?
- 41:06
- Unidentified speaker: There's lots of players.
- 41:07
- Unidentified speaker: People can build companies faster than ever.
- 41:09
- Unidentified speaker: And so when we're looking at a company, we're looking for an unfair fight, bringing a gun to a knife fight, right?
- 41:15
- Unidentified speaker: It's who knows this industry better?
- 41:18
- Unidentified speaker: Who's better connected to this industry?
- 41:20
- Unidentified speaker: Who has the license to win in this industry?
- 41:23
- Unidentified speaker: That can often, in addition to having the best technology, you have to have base case, very strong, if not the best technology.
- 41:28
- Unidentified speaker: But having that other tipping point can really tip the scales in your favor in terms of how fast you can penetrate uh an esoteric market like airlines.
- 41:36
- Unidentified speaker: I think your answer will be very disappointing to engineers.
- 41:39
- Unidentified speaker: I think since 2008, when I started in the startup world, engineers have been trying to replace salespeople.
- 41:45
- Unidentified speaker: And of like this high school jocks, uh jocks versus nerds thing where uh for now almost 20 years, and engineers have been trying to replace salespeople.
- 41:54
- Unidentified speaker: The problem with that is they may be ultimately successful.
- 41:58
- Unidentified speaker: It might take 10, 20 years.
- 41:59
- Unidentified speaker: Why?
- 42:00
- Unidentified speaker: Because there's still human beings running the organizations and human beings need solution, they need trust, they need all these things.
- 42:05
- Unidentified speaker: So this idea that you could come to the market with the best AI solution still needs those salespeople that help with implementation, still needs those people that help actually on the operations level and all these things that still require real live human beings, at least for now.
- 42:20
- Unidentified speaker: Yeah.
- 42:20
- Unidentified speaker: And I would argue that's all true.
- 42:23
- Unidentified speaker: That may go away a little bit as you automate sales process, and that's fine.
- 42:27
- Unidentified speaker: But what still remains is understanding customer needs.
- 42:30
- Unidentified speaker: People who deeply understand their customer, some of them were formerly the customer they're targeting, right?
- 42:36
- Unidentified speaker: They bring back unique insights and problem sets that the best engineers thrive with.
- 42:43
- Unidentified speaker: Okay.
- 42:44
- Unidentified speaker: Generally, the best engineers, unless they're the founder of a company, are not going out and talking to the customer or have lived inside the customer for 20 years.
- 42:53
- Unidentified speaker: And so the salesperson can also be sort of a customer solutioning specialist that brings back unique insight or understanding to the engineers that informs product roadmap, solutioning, and making sure that whatever we're building within a startup has extremely high value day one to the customers because it came from them.
- 43:10
- Unidentified speaker: Do you know the story of how GoDaddy massively increased the revenue?
- 43:13
- Unidentified speaker: I don't.
- 43:14
- Unidentified speaker: They put salespeople into their customer service calls.
- 43:18
- Unidentified speaker: So they realize that the best time to actually upsell somebody and figure out what people need is at the time that they have issues.
- 43:25
- Unidentified speaker: It goes back to this kind of paradox in customer services.
- 43:28
- Unidentified speaker: When you have a problem and a company solves it for you, you may end up actually building affinity with the company more so than if you didn't have that problem.
- 43:35
- Unidentified speaker: So they actually come sort of de facto design partners with you.
- 43:38
- Unidentified speaker: And uh they they figured this out a decade ago, and now we're kind of seeing this three emerge again in AI.
- 43:43
- Unidentified speaker: It's fascinating.
- 43:44
- Unidentified speaker: Um we invested into another AI company that was doing post-sales data gathering for service delivery companies.
- 43:52
- Unidentified speaker: So sales makes a sale, right?
- 43:54
- Unidentified speaker: They capture some data in their CRM about how that process went, and everybody has a steak dinner in high fives and says, great job, everyone.
- 44:01
- Unidentified speaker: And that's usually the it's like the day of your wedding, right?
- 44:03
- Unidentified speaker: It's like the happiest day of everyone's life is the day that you consummate the deal.
- 44:07
- Unidentified speaker: Um, and then after that, you start actually delivering the service.
- 44:10
- Unidentified speaker: You could be an invested banker, you could be a consultant, you could be a lawyer.
- 44:14
- Unidentified speaker: You're delivering service over a long time series.
- 44:18
- Unidentified speaker: Monitoring those interactions is what our company, Known Well, does.
- 44:21
- Unidentified speaker: And what it's pulling back is unique insights for the management to say, it's great that you sold this customer.
- 44:27
- Unidentified speaker: We're going to monitor minute by minute, interaction by interaction, A, how happy they are, but we're also going to distill down everything the customer's telling you about what they need.
- 44:36
- Unidentified speaker: You can take those insights and immediately put them back into those enterprises to probabilistically and sort of stochastically lay out what should we build next.
- 44:45
- Unidentified speaker: Because you have a time series, multiple customers, and frequency graph of what your customer is telling you.
- 44:52
- Unidentified speaker: When you digitize that and digest that with AI, it creates unique insights for the service delivery partner, which says, you know, it could say, wow, we just really don't like this team that we're working with, or so and so was not really great to work with, or you weren't very timely.
- 45:04
- Unidentified speaker: But it could also say, hey, the the software you sold me does XYZ, but we were really looking for, you know, tau.
- 45:12
- Unidentified speaker: And that can be unique insight that can then recursively get pulled back into the product roadmap.
- 45:17
- Unidentified speaker: When I talk to engineers at Anthropic, I also have some friends at OpenAI.
- 45:21
- Unidentified speaker: They're telling me that average engineers are getting four or five X efficiencies and that the 10X engineers are getting actually 100x.
- 45:28
- Unidentified speaker: So the 10X engineers, which oftentimes are twice as expensive, they end up being even better returns on investment.
- 45:35
- Unidentified speaker: Are you seeing the same phenomenon on the sales side where the top salespeople are actually getting these major efficiencies from AI?
- 45:42
- Unidentified speaker: I think it's still pretty early innings on that.
- 45:44
- Unidentified speaker: Um the best salespeople generally are still the best salespeople.
- 45:47
- Unidentified speaker: I think sales is probably slightly slower to adopt, or at least, you know, just like a lot of enterprise that we discussed earlier, is only starting to scratch the surface of the true scalability.
- 45:58
- Unidentified speaker: I think if you're an AI native, if you're a founder and not just like an enterprise salesperson, a founder doing sales can do things that they never could have done at a scale they've never done.
- 46:07
- Unidentified speaker: Um, even in like my own side projects, my agents are running business development workflows and analysis, you know, while I'm sleeping, right?
- 46:15
- Unidentified speaker: And so that's me being sort of like a hacky founder mind that's just running those in the side.
- 46:20
- Unidentified speaker: Uh similarly for fundraising at Studio BC, you know, lots of agents involved in identifying LPs, you know, drafting initial emails, keeping us in the loop in terms of follow-up, all these things, right?
- 46:31
- Unidentified speaker: So it's almost like we have a chief of staff there.
- 46:33
- Unidentified speaker: Um, but that's us as AI native, AI, you know, early adopters.
- 46:38
- Unidentified speaker: When you think about traditional sales, there's some defensibility.
- 46:40
- Unidentified speaker: There's a, well, how do I matter in a world where AI is doing a part of my job?
- 46:44
- Unidentified speaker: And so the willingness to adopt some of those technologies is a little slower.
- 46:48
- Unidentified speaker: And again, that comes back to my point earlier where it's like if the CEO is native and understands the impact that they could have on sales, that will start to trickle down quicker.
- 46:56
- Unidentified speaker: Um but if you're asking that ahead of sales, right, they're going to try to protect their fiefdom potentially.
- 47:01
- Unidentified speaker: Yeah, it makes me think about this alignment issue where salespeople within startups where they have equity may be faster to adopt than ones in traditional kind of Fortune 500 companies.
- 47:10
- Unidentified speaker: Yeah.
- 47:10
- Unidentified speaker: It's a great alignment question.
- 47:11
- Unidentified speaker: Yeah.
- 47:12
- Unidentified speaker: If you could go back to 2009 when you first started uh working in venture and you could give yourself one timeless piece of advice, what would that be?
- 47:20
- Unidentified speaker: Wow.
- 47:21
- Unidentified speaker: That this one to people take a long time to think about.
- 47:23
- Unidentified speaker: Two things coming from the culture of Bane Capital.
- 47:26
- Unidentified speaker: One is um, you know, some of the smartest 10 or 12 partners I've ever been around.
- 47:31
- Unidentified speaker: So it was an amazing place to start, be a fly in the wall, learn from and listen to.
- 47:35
- Unidentified speaker: Um, but there was also a lot of humility among those partners uh and as they interacted with founders.
- 47:41
- Unidentified speaker: What I I think the cultural fit for me, and and just to go back before that, when I was leaving investment banking, I thought I was going to go into private equity, just like all my colleagues did.
- 47:49
- Unidentified speaker: Um, but I got a call from Bain Capital Ventures, took the meeting, and just fell in love immediately.
- 47:54
- Unidentified speaker: It was the best interview I've ever had.
- 47:55
- Unidentified speaker: I called my parents and said that was the best interview I've ever had.
- 47:57
- Unidentified speaker: And the culture was such that, you know, they were extremely smart but humble.
- 48:02
- Unidentified speaker: They wanted to drive value in the ecosystem and they understood that this was a long game.
- 48:07
- Unidentified speaker: And so um, that's how I've tried to operate in my career, which is like even if you're not investing in a company, drive some value, right?
- 48:13
- Unidentified speaker: Like what's the worst could that could happen?
- 48:15
- Unidentified speaker: Maybe they come back a few years later or they send a friend.
- 48:18
- Unidentified speaker: But um the second thing I learned from Bane Capital was I think that they did still have some consulting DNA in terms of how they looked at deals.
- 48:27
- Unidentified speaker: Very complete uh analysis and diligence process at Bane Capital Ventures, which I was which I was impressed with, and I still keep uh in my back pocket to this day, which is you know, it's a combination of technology, financing, market strategy, um, and and reasons to win over time in a competitive market.
- 48:44
- Unidentified speaker: So that's what fascinated me about working there.
- 48:47
- Unidentified speaker: And, you know, as I've kind of progressed in my career, I've gone earlier and earlier towards seed stage because I think that's where the problem is most acute with founders.
- 48:55
- Unidentified speaker: And I love getting in there and helping them uh or creating creating the conditions that help them to solve those problems.
- 49:01
- Unidentified speaker: And, you know, hopefully I'm not the smartest guy in the room, like hopefully they are, right?
- 49:05
- Unidentified speaker: Um, and I can help create conditions and foundations and connections for them that help to make that happen in an asymmetric way compared to uh, you know, the average seed investor.
- 49:14
- Unidentified speaker: Well, Joe, this has been absolute masterclass.
- 49:16
- Unidentified speaker: Thanks so much for jumping on.
- 49:17
- Unidentified speaker: My pleasure.
- 49:17
- Unidentified speaker: Thank you so much, David.
More from How I Invest with David Weisburd
E456: Baylor Endowment on Rethinking Venture Capital, Concentrated Portfolios, and the Power of Compounding
Episode discussion: Baylor Endowment on VC Skepticism, Concentration, and Compounding
Speakers’ claims: Renee Hanna and guest discuss Baylor University's endowment strategy, emphasizing a concentrated portfolio of approximately 20 GP relationships to manage operational load and deepen partnerships. They reject early-stage venture due to inflated seed valuations, preferring expansion capital and later-stage investments that offer clearer liquidity paths and 3x-5x returns over 3-5 years. The team enforces a 15% net IRR hurdle for private investments and actively manages liquidity by selling underperforming assets early. They also highlight the role of behavioral biases and geographic location in maintaining disciplined, contrarian investment decisions.
E455: Hunter Walk on Why Venture is Getting Harder, Picking Great Founders, and Investing Against Consensus
Episode discussion: Hunter Walk explains why Homebrew switched to self-capital to avoid fund-size incentives and how AI crowds the market.
Speakers’ claims: Hunter Walk describes Homebrew's shift from institutional LPs to self-capital in 2022 to maintain alignment and avoid the deployment pressures of large funds. He argues that as venture funds grow, GPs are incentivized to chase hot names and deploy capital volume over ownership quality, forcing founders to raise excessive capital. Walk contrasts 'pre-consensus' investing with consensus-driven punditry, noting that AI-driven crowding requires identifying opportunities before market maps form. He also discusses Screendoor, a vehicle backing emerging managers, and warns against 'dub tape' decay in multi-generational venture firms.