Limitless

Hosts analyze Anthropic's IPO risks, citing governance, regulatory headwinds, and funding sustainability concerns.

1h 1m Listen to the original
AI InfrastructureIPO AnalysisCorporate GovernanceRegulatory RiskCloud Computing

Summary

Dr. Thaon Simms and Dr. Matthew Preston examine Anthropic's proposed IPO, highlighting its reliance on third-party cloud distribution and high capital expenditure requirements. They discuss the political risks associated with potential Democratic control of Congress and its impact on data center regulations. The hosts express skepticism about the sustainability of current AI funding models and the lack of a durable competitive moat. Dr. Simms declines to invest due to valuation uncertainty and governance concerns, while the co-host prefers to wait for a market correction.

Episode sections

  1. 0:00 – 1:12

    Intro and Anthropic's Business Model

    Dr. Thaon Simms and Dr. Matthew Preston introduce Anthropic, the company behind Claude AI, and discuss its proposed IPO. They note that developing advanced AI models requires billions in funding and that profitability is uncertain. The hosts outline their intent to analyze how Anthropic makes money, its dependence on tech partners like Google, and the sustainability of current AI investment levels.

  2. 1:12 – 8:10

    Claude Use Cases and Revenue Structure

    The hosts discuss practical uses for Claude, including coding, stress-testing investment theses, and generating presentations. They explain that Anthropic's revenue comes primarily from usage-based consumption (approximately 80%) rather than subscriptions (approximately 20%), similar to how OpenAI operates.

  3. 8:10 – 14:47

    Large Language Models and User Behavior

    Dr. Simms explains that large language models are predictive tools trained on vast amounts of text to generate appropriate outputs. Dr. Preston cites a September 2025 study by OpenAI and Harvard, which found that the majority of ChatGPT usage involves learning, tutoring, and practical advice, while coding and search represent smaller portions of usage.

  4. 14:47 – 19:50

    Product Features and Security Concerns

    The hosts review Anthropic's product suite, including Claude Code, Claude Assistant, and Claude Security. Dr. Simms expresses significant concern regarding AI security risks, particularly the potential for AI agents to escape their containers and execute unauthorized transactions or attacks, citing a lack of trust in current AI safety mechanisms.

  5. 19:50 – 20:47

    Political Risk and Data Center Regulation

    Dr. Simms discusses the political landscape following recent US midterm elections, noting Democratic control of the Senate and House. He highlights the potential for stricter scrutiny on AI development and infrastructure.

  6. 20:47 – 23:41

    AI Data Center Moratorium and Geopolitical Shifts

    The hosts discuss the potential for data center moratoriums and how geopolitical shifts might impact AI infrastructure development.

  7. 23:41 – 25:32

    Data Center Economics and Resource Constraints

    Dr. Preston suggests that if the US restricts data centers, companies might relocate to countries with excess energy, such as solar power. Dr. Simms counters that data centers also require significant water for cooling, making relocation difficult and subjecting companies to new regulatory and resource constraints in different jurisdictions.

  8. 25:32 – 29:20

    Regulatory Hurdles and Data Center Costs

    Dr. Simms discusses the regulatory and community pushback against data centers, citing issues like water access, power consumption, and lack of local job creation. He notes that these delays and compliance costs could cost Anthropic billions and that the company should have listed earlier to avoid such timing risks.

  9. 29:20 – 33:43

    Competitive Landscape and Model Distillation

    Speaker 1 compares Anthropic to competitors like DeepSeek and OpenAI, highlighting how DeepSeek's constraints led to efficiency innovations and open-source models. They discuss 'distillation,' where competitors use Anthropic's models to improve their own, potentially siphoning knowledge and revenue.

  10. 33:43 – 37:59

    Customer Lock-in and Trust Dynamics

    The speakers analyze Anthropic's strategy to lock in enterprise customers through consultants and integration, contrasting it with the ease of switching between models like ChatGPT and Claude. Dr. Simms highlights how trust and ethical stances, such as Anthropic refusing autonomous weapons contracts, drive user preference.

  11. 37:59 – 44:42

    Distribution Channels and Google's Dual Role

    Dr. Simms explains the difference between direct sales and cloud marketplace distribution, noting that 47% of 2025 revenue came from cloud providers. He details Google's complex relationship with Anthropic, acting as a competitor (Gemini), supplier (compute/TPUs), and distributor (Google Cloud), while also being an investor.

  12. 44:42 – 52:24

    AI Training Costs and the 'Bubble' Debate

    Speaker 1 describes the high, recurring costs of training and serving AI models, comparing the industry to smartphones where constant innovation is required to survive. Dr. Simms expresses concern about the sustainability of current funding models and the potential for an AI valuation bubble if returns are unrealistic.

  13. 52:24 – 54:02

    AI Funding Sustainability and Profitability Pressures

    Dr. Thaon Simms argues that the current funding model for AI, involving billions in capital expenditure, is unsustainable without significant increases in subscription revenue or a pivot to advertising. He notes that lenders may demand higher interest rates (e.g., 12%) if companies cannot generate sufficient cash flow, which would negatively impact stock prices.

  14. 54:02 – 1:00:59

    Anthropic's Share Class Structure and Founder Control

    The speakers discuss Anthropic's complex share structure, including Class A, N, F, and E shares. They highlight that Class F shares are held by founders, granting them majority control over the company despite public offerings. This structure allows founders to make decisions without shareholder approval, raising questions about governance and alignment with public investors.

Assets discussed

AssetKindStanceContext
AlphabetGOOGLequityneutralYeah. And so the cloud company would help because they help the customer, they help anthropic, you know, find customers, connect them, pay for the service too. So you're paying. So if I'm buying, if I'm accessing cloud through Google Cloud, I'm paying Google Cloud, and then Google Cloud has to pay Cloud. Right? So it can combine the bill or so on. But uh I think of it like uh like a bakery, right? Direct is you buy directly from the bakery because you like those products, and then marketplaces like you're buying from a grocery store that sells those baked goods, the same bakery's bread, and so on. So the bakery baker still makes the bread, the grocery store helps um shoppers to find and pay for it. And they keep a fee as well. So that's how I see it. And when you look at the numbers based on the prospectus, it said that customers purchased through cloud provider marketplaces accounted for 47% of um 2025 full-year's revenues. And then 44% of the 2026 20s 26's is revenue. So it's around the same in mid-40s. in terms of their revenues. So that's still a big part of it. Now, when we were talking about competitors and so on, I started to think even about Google because I think Claude's relationship with Google is quite interesting still. Because uh Google matters so much to Anthropic in many different roles, many different ways.

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

Claims worth verifying

Anthropic's revenue is approximately 80% from consumption and 20% from subscription.

Dr. Thaon Simms ·8:02

How to verify: Anthropic S-1 Filing (SEC EDGAR) or IPO Prospectus; verify revenue breakdown by usage vs. subscription. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Primary source

A study by OpenAI and Harvard in September 2025 titled 'How People Use Chat GPT' found that just under 30% of usage was for learning, tutoring, brainstorming, and personal advice.

Unidentified speaker (speaker_1) ·11:29

How to verify: OpenAI Research Blog or Harvard University publication; search for the specific study title and date. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Primary source

The same OpenAI/Harvard study found 24% of usage was for research and information, 24% for writing and editing, 7% for images and multimedia, 5% for conversation/entertainment, and 5% for coding.

Unidentified speaker (UNKNOWN) ·11:48

How to verify: OpenAI Research Blog or Harvard University publication; verify the specific percentage breakdowns in the study. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Dr. Simms has been paying for ChatGPT since 2022 or 2023.

Unidentified speaker (UNKNOWN) ·12:47

How to verify: OpenAI Account Profile or subscription history; verify start date of ChatGPT Plus subscription. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Senator Bernie Sanders and AOC announced an 'AI data center moratorium act' to delay or halt the construction of certain data centers.

Dr. Thaon Simms ·21:06

How to verify: US Senate or House of Representatives official press releases; search for 'AI data center moratorium act' and sponsors. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Democratic control of the Senate and House occurred following recent midterm elections, with a new Congress beginning in January 2027.

Unidentified speaker (UNKNOWN) ·20:12

How to verify: US Election Results (FEC or state election boards); verify party control of Congress post-midterms and start date of 120th Congress. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Anthropic has products including Claude Code, Claude Assistant, and Claude Security.

Dr. Thaon Simms ·14:47

How to verify: Anthropic official website or product documentation; verify product names and features. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Data centers require significant amounts of water for cooling systems.

Dr. Thaon Simms ·25:12

How to verify: Technical reports from data center operators (e.g., Google, Microsoft) or industry studies on water usage in data centers. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Anthropic's upgrades are done by code, with AI altering its own code to become smarter.

Dr. Thaon Simms ·18:51

How to verify: Anthropic technical blog posts or research papers on 'constitutional AI' or automated model improvement; verify claims of self-modifying code. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Anthropic's prospectus states that 5-7% of revenue came from usage that violated usage policies or involved distillation.

Unidentified speaker (speaker_1) ·32:49

How to verify: Review Anthropic's IPO prospectus filed with the SEC EDGAR, specifically the 'Risk Factors' or 'Management’s Discussion and Analysis' sections regarding revenue sources and policy violations. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Primary source

Customers purchased through cloud provider marketplaces accounted for 47% of Anthropic's 2025 full-year revenues.

Dr. Thaon Simms ·40:40

How to verify: Review Anthropic's IPO prospectus filed with the SEC EDGAR, specifically the 'Business' or 'Financial Statements' sections detailing revenue breakdown by distribution channel. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Primary source

Google supplies computing capacity, including TPUs, to Anthropic for training and running models.

Dr. Thaon Simms ·43:06

How to verify: Review Anthropic's IPO prospectus filed with the SEC EDGAR for disclosures on 'Concentration of Business' or 'Related Party Transactions' with Google, and Google's investor relations page for partnership announcements. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Primary source

Google invested in Anthropic.

Dr. Thaon Simms ·43:46

How to verify: Review SEC filings (e.g., Form D or 13F) or press releases from Google/Alphabet and Anthropic confirming the investment. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Primary source

DeepSeek is owned majority by a Chinese investor named Liang Weng Wen Feng.

Unidentified speaker (speaker_1) ·30:40

How to verify: Search for corporate registry filings or news reports regarding DeepSeek's ownership structure and key shareholders. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Primary source

Anthropic refused to use its model for autonomous weapons, while OpenAI agreed.

Dr. Thaon Simms ·36:45

How to verify: Search for news reports and official statements from Anthropic and OpenAI regarding contracts with government or defense entities for autonomous weapons systems. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

It costs approximately $1 billion to train a model like Opus 6.

Unidentified speaker (speaker_1) ·46:15

How to verify: This is a speaker estimate; verify by comparing against industry reports on AI training costs or Anthropic's disclosed R&D expenses in their prospectus. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Anthropic's current funding model may not be sustainable due to high costs and regulatory risks.

Unidentified speaker (UNKNOWN) ·49:52

How to verify: Analyze Anthropic's cash flow statements and debt obligations in their IPO prospectus filed with the SEC EDGAR. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Primary source

SpaceX should have listed a long time ago, not after the midterms.

Unidentified speaker (UNKNOWN) ·27:33

How to verify: This is a subjective opinion; verify by checking SpaceX's public statements on IPO plans and market conditions during the relevant election period. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Google's Gemini is better than Claude at multimodal tasks like video and image analysis.

Unidentified speaker (UNKNOWN) ·41:37

How to verify: Compare independent benchmarks (e.g., from AI research labs or tech publications) evaluating Gemini and Claude on multimodal tasks. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Lenders may increase interest rates to 12% if AI companies cannot generate sufficient cash flow to service debt at lower rates (e.g., 8%).

Dr. Thaon Simms ·52:24

How to verify: Review loan covenants and interest rate trends for high-growth tech companies in SEC filings (e.g., Form 10-K) or bond prospectuses during periods of high capital expenditure. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Primary source

Anthropic's CEO stated he would not pursue advertising revenue as a monetization strategy.

Dr. Thaon Simms ·53:28

How to verify: Check Anthropic's IPO prospectus (S-1) filed with the SEC EDGAR database for 'Risk Factors' or 'Business' sections discussing revenue models and competitive landscape. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Primary source

Anthropic has multiple share classes including Class A, N, F, and E, with Class F held by founders.

Unidentified speaker (speaker_1) ·54:55

How to verify: Review the 'Capitalization' section of Anthropic's IPO prospectus (S-1) on SEC EDGAR to confirm share class designations and voting rights. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Primary source

Class N shares are allocated to strategic partners like Amazon or Google and lack voting rights.

Unidentified speaker (speaker_1) ·54:55

How to verify: Review the 'Capitalization' section of Anthropic's IPO prospectus (S-1) on SEC EDGAR to verify share class rights and any agreements with strategic partners. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Primary source

Anthropic may raise $100 billion at a $1.5 trillion valuation.

Unidentified speaker (UNKNOWN) ·57:15

How to verify: Monitor financial news outlets (e.g., Bloomberg, Reuters) and SEC filings for the final IPO pricing and size announcement. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Primary source

Do your own research

  • Review Anthropic's S-1 filing for detailed revenue breakdowns and risk factors.
  • Verify the September 2025 OpenAI/Harvard study on LLM usage patterns.
  • Check recent US legislative proposals regarding data center energy and water usage.
  • Analyze Anthropic's share class structure (A, N, F, E) for voting rights and control.
  • Assess the financial health of cloud providers like Google and Amazon as key distribution partners.
  • Compare Anthropic's capital expenditure requirements against its cash flow generation.
  • Monitor political polling and election results for potential shifts in AI regulatory oversight.
Full transcriptexpand
0:00
Dr. Thaon Simms: Could the shares actually make sense at the price investors are being asked to pay? Google can help Anthropics sell and operate cloud while also competing with it. So
0:10
Unidentified speaker (speaker_1): you're then gonna have to then spend more money and then what do you have to spend two billion now? It really takes a lot of money to continuously like fund this.
0:18
Dr. Thaon Simms: So you have to constantly be getting better. It's either that or you die. Are these the catalysts that we're seeing now that will make this AI bubble pop? I just think that currently how it's being funded, like all these billions of dollars being poured into it. I just don't think it's that sustainable.
0:34
Unidentified speaker (speaker_1): I thought the F was almost like for F U
0:36
Unidentified speaker (UNKNOWN): because
0:36
Unidentified speaker (speaker_1): they're saying, hey, this is our ship. So you wouldn't buy it at all?
0:41
Dr. Thaon Simms: No.
0:53
Dr. Thaon Simms: Welcome back, Limit Breakers. I'm Dr. Theon Sims and this is the Limitless Podcast, where we make money, markets, and big ideas simple. As always, I'm joined by my co-host, Dr. Matthew Preston. Hail up the people there, Matthew. Hello, Hell up the Limit Breakers.
1:10
Unidentified speaker (UNKNOWN): Hey
1:10
Unidentified speaker (speaker_1): guys, what's
1:11
Unidentified speaker (UNKNOWN): good?
1:12
Dr. Thaon Simms: And today we're getting into
1:15
Dr. Thaon Simms: Anthropic, a company called Am Anthropic, their proposed IPO, right? Their proposed initial public offering. That's the company behind Cloud. Right. And Cloud can do several things. Um I don't know if you guys have heard about Cloud, but I'm just gonna ask Preston. Preston, do you use Cloud?
1:34
Unidentified speaker (UNKNOWN): Uh
1:34
Unidentified speaker (speaker_1): yeah, I do. Do you?
1:37
Dr. Thaon Simms: Yeah, yeah. What what do you what do you use cloud for?
1:41
Unidentified speaker (speaker_1): Um that's a good question. I use it uh it's pretty good at coding. And uh also just to like even make like also graphics or stuff like that.
1:55
Dr. Thaon Simms: Nice.
1:57
Unidentified speaker (UNKNOWN): Yeah,
1:57
Dr. Thaon Simms: so it's
1:57
Unidentified speaker (UNKNOWN): I can help
1:58
Unidentified speaker (speaker_1): stress test. it can help stress test. Like say I'm like, okay, I think
2:03
Unidentified speaker (speaker_1): um
2:04
Unidentified speaker (UNKNOWN): I
2:05
Unidentified speaker (speaker_1): build like evaluation I'm trying to decide if I'm gonna buy a company or something like that. I can
2:09
Unidentified speaker (UNKNOWN): like build
2:09
Unidentified speaker (speaker_1): up okay, this is my thesis and then make it like try to poke holes in
2:13
Unidentified speaker (UNKNOWN): it.
2:14
Dr. Thaon Simms: Ah wow, okay. All right.
2:16
Unidentified speaker (UNKNOWN): What
2:17
Dr. Thaon Simms: do you guys think?
2:17
Unidentified speaker (UNKNOWN): What do you advance? What do you use it for?
2:21
Dr. Thaon Simms: Um
2:22
Unidentified speaker (UNKNOWN): d
2:22
Dr. Thaon Simms: I mean, okay, so you know, you just said you used cloud to code at I also started using cloud a bit more to code. Um more than codecs in the past few days. Uh, you know, thing things I code are just some apps that help me um with just personal tasks, like some stuff that
2:42
Dr. Thaon Simms: you know, like my pay slip reader, my budget app, those kind of things, which
2:47
Unidentified speaker (speaker_1): maybe we'll launch one day, we'll see.
2:49
Unidentified speaker (UNKNOWN): Um
2:50
Dr. Thaon Simms: you know, but uh yeah cloud cloud itself though I was just going to jump into just general use cases of cloud. So cloud can write code. Um and it sounds more technical than it actually is. I find that coding nowadays um is now suited for non technical users with
3:10
Dr. Thaon Simms: AI, you know, uh or large language models uh like cloud like uh ChatGPT, and we'll go into that. But cloud can do things like write code, analyze documents, just help people get work done really and truly. Um but I feel like the question for us, you know, going through Anthropic's IPO is can Anthropic turn all
3:30
Dr. Thaon Simms: of that into
3:32
Dr. Thaon Simms: um something sustainable, something
3:35
Dr. Thaon Simms: yeah, profits, exactly, and lasting profits, right? You know, could
3:39
Unidentified speaker (UNKNOWN): it
3:39
Dr. Thaon Simms: could the shares actually make sense at the price investors are being asked to pay? So I think in this episode we're just gonna be following the dollar. Um,
3:49
Unidentified speaker (UNKNOWN): you
3:49
Dr. Thaon Simms: know, who pays for cloud, where the money goes, and what could be left for shareholders, you know. We probably can look at some of the competition and so on, and maybe the huge promises um in that whole AI space. So
4:03
Dr. Thaon Simms: yeah.
4:05
Dr. Thaon Simms: Funny. This this feels like it goes take on Artica's um
4:10
Dr. Thaon Simms: framework, but I don't I don't necessarily think so. Unless you
4:13
Unidentified speaker (UNKNOWN): want to
4:14
Dr. Thaon Simms: jump into it like that. Because that's how I analyze companies now. So
4:18
Unidentified speaker (UNKNOWN): yeah. But
4:19
Dr. Thaon Simms: we can start with how how cloud makes money. Press Preston. done. Talk to us. Um, Like if someone is paying for Claude, right? What are they actually buying? Like what what what what what do people pay for?
4:35
Unidentified speaker (speaker_1): Okay. Um, so from my understanding, not a software engineer or anything like that, but basically, Anthropic, the company that sells cloud, and then I think the good comparison, like a lot of people may not know a lot of it's probably like more like nerds or like more technical people here. about oh, anthropic or cloud. But um people have probably heard about
4:55
Unidentified speaker (speaker_1): like OpenAI. And then OpenAI is the company that makes products like ChatGPT. So OpenAI is so anthropic, as ChatGPT is the cloud. Um basically, what these companies do is that they um
5:13
Unidentified speaker (speaker_1): take a lot of information. Usually it's a lot of like text and stuff like that. So it's like documents, um, stuff like like people, your stuff online,
5:24
Unidentified speaker (speaker_1): papers, things like that. And they basically put it into these what you'd call
5:30
Unidentified speaker (UNKNOWN): I
5:30
Unidentified speaker (speaker_1): guess computers or like processors and stuff like that. And then they are able to make or p kind of predict the next um answer, the next what you'd call the next word or the next token. Um and that all of
5:45
Unidentified speaker (UNKNOWN): that really beautiful.
5:49
Unidentified speaker (speaker_1): Yeah. That's
5:50
Unidentified speaker (UNKNOWN): really what
5:50
Unidentified speaker (speaker_1): um I guess Claude is. Um but I guess if you at the end of the day, really, um most of these companies they either do how would you say I'd say they do kind of two things. They'll do they will oh I guess it's more than two, but they will sell you access to this model that can, you know, read documents, write
6:10
Unidentified speaker (speaker_1): text, write emails, those kind of stuff, uh as
6:13
Unidentified speaker (UNKNOWN): part of
6:14
Unidentified speaker (speaker_1): like a subscription. So I mean all of us are probably on subscriptions. People use like Netflix, so you can pay for a net I don't know how much how much does Netflix cost now?
6:22
Unidentified speaker (UNKNOWN): I
6:23
Dr. Thaon Simms: actually
6:23
Unidentified speaker (UNKNOWN): don't know Netflix
6:24
Dr. Thaon Simms: I I stopped that. Yeah, I I cancelled I my subscription. Something is telling me like thirteen dollars a month. That could be around.
6:31
Unidentified speaker (speaker_1): Thirteen, okay. But yeah, usually I think most subscriptions for like I guess retail or smaller people, it's like oh it starts at like ten, twenty dollars, and then sometimes you can get higher subscriptions. And then just like on Netflix, you could watch an unlimited amount of movies um for that whatever time period, whatever dollar value you're paying for. So if you're paying the ten $10, dollars fifteen.
6:50
Unidentified speaker (speaker_1): $15, $20, you can watch an unlimited amount of movies. With the AI stuff, it's not as oh unlimited, but you'll usually because as much as they'll say, Oh, you could you pay the $10 or $20. I think most of them it's around $20 per month, unless you pay like annually, then you can save a lot of money.
7:06
Unidentified speaker (UNKNOWN): But they
7:07
Unidentified speaker (speaker_1): will allow you to use their model so you can go on the website just like you can talk to they
7:11
Unidentified speaker (UNKNOWN): have
7:11
Unidentified speaker (speaker_1): like the chat bots where it's like, oh, you can go to the just like you'd go to chat GPT. you say oh which thing should I buy or um how do I fix this or um
7:19
Unidentified speaker (UNKNOWN): yeah how
7:20
Unidentified speaker (speaker_1): much calories I don't know what people typically ask it.
7:23
Unidentified speaker (UNKNOWN): Um
7:24
Unidentified speaker (speaker_1): but you can go to it for answers and stuff like that. You can say oh hey look at my um expenses tell me where I'm spending the most or
7:32
Unidentified speaker (UNKNOWN): um
7:33
Unidentified speaker (speaker_1): those kind of all those kind of different stuff. So they
7:35
Unidentified speaker (UNKNOWN): have what they call the
7:36
Unidentified speaker (speaker_1): subscriptions and then they
7:37
Unidentified speaker (UNKNOWN): have um
7:37
Unidentified speaker (speaker_1): people who pay for usage. So it's just based on so it's like, okay, based on if you think of it as it costs one dollar to read ten documents or it costs one dollar to write one email. I
7:48
Unidentified speaker (UNKNOWN): mean
7:48
Unidentified speaker (speaker_1): it's not that simple, but so they pay on based on your usage. So you could the more you use, the more you're gonna pay. And
7:54
Unidentified speaker (UNKNOWN): most
7:55
Unidentified speaker (speaker_1): of their um revenue or sales comes from people who pay from usage. I think it's about eighty 80% percent based on the pressed Prost access.
8:02
Dr. Thaon Simms: Yeah. it's about 80% is consumption subscription, I believe, is 20%.
8:09
Dr. Thaon Simms: Um
8:10
Unidentified speaker (UNKNOWN): just
8:10
Dr. Thaon Simms: something I wanted to touch on. So you were talking about large language models, because that that's a term that we hear a lot in this whole AI space. It's just a a model trained to So statistics, status it's just stats, right? It's trained
8:23
Unidentified speaker (UNKNOWN): to
8:23
Dr. Thaon Simms: process and generate language. Um so you were talking about tokens. So text can be represented as tokens and if you're using it, you'll see tokens a lot too. Um that that token can be words, so maybe like the
8:38
Unidentified speaker (UNKNOWN): word
8:38
Dr. Thaon Simms: and um or other units or like parts of words, like pieces of words. So
8:43
Unidentified speaker (UNKNOWN): if it's
8:43
Dr. Thaon Simms: like building, it could be build and then ing, I don't know. Um so that's that's each one
8:48
Unidentified speaker (UNKNOWN): of them. And it's almost like
8:49
Unidentified speaker (speaker_1): prefix, suffix or those kind of stuff.
8:51
Dr. Thaon Simms: Exactly,
8:52
Unidentified speaker (UNKNOWN): right?
8:53
Dr. Thaon Simms: So um basically, you know, and then they do certain things to get smarter, so we call training um to get better in terms of prediction. So all all all my understanding of large language model is is just a predictive thing. So the model gets better at predicting and it gets
9:13
Dr. Thaon Simms: better at producing like appropriate. outputs um from the context that you give it. So let's say you're asking it to um brainstorm date ideas to bring your girlfriend
9:29
Unidentified speaker (speaker_1): you use it for
9:31
Dr. Thaon Simms: no
9:31
Unidentified speaker (UNKNOWN): no
9:32
Dr. Thaon Simms: maybe I have I I feel like I've used it to no, I used it when I was traveling um. and I went to Miami.
9:40
Unidentified speaker (UNKNOWN): I was looking
9:41
Dr. Thaon Simms: at different restaurants in
9:43
Unidentified speaker (UNKNOWN): certain
9:43
Dr. Thaon Simms: areas that we can like us as a group can go and eat um and stuff like that. So into like chilies. That was really good. Um but yeah, you can use it like that. So you give it the context. So I I I'll tell it, you know, I'm traveling to this particular area in Miami. Look for restaurants with, you know, this kind of meal or blah blah blah or this kind of
10:03
Dr. Thaon Simms: atmosphere to sell drinks, etcetera.
10:04
Unidentified speaker (speaker_1): Or you could ask
10:05
Unidentified speaker (UNKNOWN): it to look
10:05
Unidentified speaker (speaker_1): at just the reviews, like oh what did people say about it? And it
10:09
Unidentified speaker (UNKNOWN): can basic
10:10
Unidentified speaker (speaker_1): scrape the website. So it can look on
10:11
Unidentified speaker (UNKNOWN): the
10:11
Unidentified speaker (speaker_1): website just like you would and you'd go on website, read it and that stuff it can
10:15
Dr. Thaon Simms: it kinda
10:15
Unidentified speaker (UNKNOWN): can
10:15
Unidentified speaker (speaker_1): do that.
10:16
Dr. Thaon Simms: Read the menu, prices, all of that kind of stuff.
10:19
Unidentified speaker (speaker_1): Convert
10:19
Unidentified speaker (UNKNOWN): it,
10:20
Unidentified speaker (speaker_1): yeah.
10:20
Dr. Thaon Simms: Yeah. Yeah. So um they use um Transformers to do a lot of that. Like GPT is the T is transformer. That's the
10:29
Dr. Thaon Simms: the infrastructure, architecture. I don't know the word, but you know, that's how it kinda is is and then, you know, every ever so often they do these training cycles. So I think a lot of the the things is that's just to make them upgrade and get better. So before we're using um I guess Cloud, Claude,
10:50
Dr. Thaon Simms: so since we're s we're sticking with anthropic, they have different models. So they have like Opus, they have Sonnet and so on. And every time they go up on iteration, it's because they'd went through some training and it's gotten better. And so this is a a new version. Um so that
11:05
Unidentified speaker (UNKNOWN): was just
11:05
Dr. Thaon Simms: me touching on large language models just a little bit.
11:09
Unidentified speaker (UNKNOWN): Oh wait,
11:09
Unidentified speaker (speaker_1): I've I found a little tangent. So I actually found there was a study that apparently OpenAI did. with Harvard in September 2025, titled How People Use Chat GPT. And they actually brought down the usage. So they say like about just under 30% was learning, tutoring, brainstorm, personal advice.
11:29
Unidentified speaker (speaker_1): About twenty four percent was research and information, which included finding facts, answering questions, researching topics. About twenty 24% four percent was writing and editing, email. emails, rewriting, summarizing, translation. Um, 7% images and multimedia, 5% conversation, reflection, entertainment.
11:48
Unidentified speaker (UNKNOWN): And
11:48
Unidentified speaker (speaker_1): about 5%
11:50
Unidentified speaker (UNKNOWN): is
11:50
Unidentified speaker (speaker_1): coding, cal number one is practical advice.
11:53
Unidentified speaker (UNKNOWN): Well,
11:53
Unidentified speaker (speaker_1): this word that I include learning under practical advice. They also say So learning, tutoring, brainstorming, and then personal advice.
12:01
Dr. Thaon Simms: And
12:01
Unidentified speaker (UNKNOWN): I found thought it was interesting.
12:03
Unidentified speaker (speaker_1): 18% was looking up specific information. So it just kind of sounds like search.
12:07
Dr. Thaon Simms: Yeah. Oh, you
12:08
Unidentified speaker (UNKNOWN): were
12:08
Unidentified speaker (speaker_1): saying
12:08
Unidentified speaker (UNKNOWN): something, sorry?
12:09
Dr. Thaon Simms: No, no, no. That's how I started using um AI though. It was more brainstorming certain things. So,
12:15
Unidentified speaker (UNKNOWN): way
12:15
Dr. Thaon Simms: back, I believe, in my when I left clinical medicine and I was working for a a company. Um
12:23
Dr. Thaon Simms: I remember I had to be do some medical. social media campaigns. So I would brainstorm ideas first. And that's that was way back when GPT was three. GPT
12:36
Unidentified speaker (UNKNOWN): three.
12:36
Dr. Thaon Simms: Yeah. GP
12:37
Unidentified speaker (UNKNOWN): were you even paying
12:38
Unidentified speaker (speaker_1): for it then?
12:39
Dr. Thaon Simms: Yeah I was. I was because
12:40
Unidentified speaker (UNKNOWN): I've been paying
12:41
Dr. Thaon Simms: for it. You c it tells you how long you've been paying for. Um oh
12:44
Unidentified speaker (speaker_3): weird.
12:45
Unidentified speaker (UNKNOWN): But
12:46
Dr. Thaon Simms: uh
12:47
Unidentified speaker (UNKNOWN): I
12:47
Dr. Thaon Simms: don't remember
12:48
Unidentified speaker (speaker_1): I'll check
12:48
Dr. Thaon Simms: maybe profile or so on but I I'm pretty sure I've been paying for it since 2022 or could be 2023.
12:59
Dr. Thaon Simms: Well, somewhat. I'm still hesitant with AI, and we can talk about why I'm still a bit hesitant and like wary and so on a little bit later. But
13:09
Unidentified speaker (UNKNOWN): yeah.
13:10
Dr. Thaon Simms: Um so re I think with Claude, Cloud Claude is really good. at a lot of things. Um I find that I use it for this is just my personal use case before we jump back into the company. I use it a lot for presentations. I use it a lot for um
13:29
Dr. Thaon Simms: things that have to
13:31
Unidentified speaker (UNKNOWN): like
13:31
Dr. Thaon Simms: front
13:31
Unidentified speaker (UNKNOWN): end things.
13:33
Unidentified speaker (speaker_1): Already.
13:33
Dr. Thaon Simms: No, No,
13:35
Dr. Thaon Simms: no, this is you know, the limit breakers need for no. Um and if they if they don't care
13:40
Unidentified speaker (UNKNOWN): that
13:41
Unidentified speaker (speaker_1): less exclusive,
13:42
Dr. Thaon Simms: yeah, yeah, yeah. So I I use it for presentations. I no longer use PowerPoint actually. Everything is HTML files, everything is coded. It's
13:50
Unidentified speaker (UNKNOWN): a break Microsoft
13:51
Unidentified speaker (speaker_1): foot.
13:52
Dr. Thaon Simms: Yeah. Yo, do you realize that? Like these companies are just breaking each other's foot. Um, we soon go into
13:59
Unidentified speaker (UNKNOWN): all of
13:59
Dr. Thaon Simms: that too. But
14:00
Unidentified speaker (UNKNOWN): Microsoft
14:01
Unidentified speaker (speaker_1): owns a stake in them, so.
14:04
Unidentified speaker (UNKNOWN): let's
14:05
Dr. Thaon Simms: let's we soon get into that. Um, in fact, write that down, write that down. Um
14:10
Unidentified speaker (UNKNOWN): what
14:10
Dr. Thaon Simms: else do I use? I really use it for front end stuff. So even the design of how I
14:14
Unidentified speaker (UNKNOWN): want my approach.
14:16
Dr. Thaon Simms: Yeah, yeah, yeah. Bro,
14:18
Dr. Thaon Simms: you
14:20
Dr. Thaon Simms: you use cloud way more than I use cloud. So here no man. But um, yeah, I find that so that's some of the stuff that they do uh in terms of anthropic. Um and that's some of the stuff I use it for. It's mainly just how things look and feel. So presentations, powerpoints, um, decks.
14:40
Dr. Thaon Simms: Uh maybe if I build a website, I'd use cloud too and so on. So yeah. Um
14:47
Dr. Thaon Simms: really useful tools in terms of what cloud can actually do. If you look through the prospectus, the file and describes different things. So you have like cloud code, um where you you know, you can plan, write, edit, test and debug code, um and work through certain development tasks. You have cloud
15:07
Dr. Thaon Simms: assistant, you have cowork, but I no longer see cowork in the app. Cowork
15:11
Unidentified speaker (UNKNOWN): was meant mechanical. to be.
15:21
Unidentified speaker (speaker_1): And I
15:21
Unidentified speaker (UNKNOWN): think
15:21
Unidentified speaker (speaker_1): chat as our OpenAI is sounds like they're gonna do the same thing to
15:25
Dr. Thaon Simms: They have done the same thing. There's no there's no chat GPT work anymore. It's just either chat or count.
15:33
Unidentified speaker (speaker_1): I don't think everyone has that yet though.
15:35
Dr. Thaon Simms: Oh, just in my app, I guess.
15:37
Unidentified speaker (UNKNOWN): Like
15:37
Unidentified speaker (speaker_1): when I'm on no, I say when I'm on the browser. I say chat and I
15:40
Unidentified speaker (UNKNOWN): say worker. So I
15:42
Unidentified speaker (speaker_1): think we're
15:42
Unidentified speaker (UNKNOWN): supposedly gonna
15:44
Unidentified speaker (speaker_1): kill it. Yeah.
15:45
Dr. Thaon Simms: Mm-hmm.
15:47
Dr. Thaon Simms: Yeah, they have other things like cloud security. Um
15:51
Unidentified speaker (UNKNOWN): which
15:51
Dr. Thaon Simms: is
15:51
Unidentified speaker (speaker_1): they're going
15:52
Unidentified speaker (UNKNOWN): into
15:52
Unidentified speaker (speaker_1): almost every industry. I say healthcare, life sciences.
15:56
Dr. Thaon Simms: Mm-hmm.
15:57
Dr. Thaon Simms: Design. Is that cloud club
15:59
Unidentified speaker (UNKNOWN): personal
16:00
Dr. Thaon Simms: finance?
16:01
Unidentified speaker (speaker_1): Yeah, well, I think I see saw the announcement. I don't think I don't know if it's
16:05
Unidentified speaker (UNKNOWN): rolled out yet.
16:06
Unidentified speaker (speaker_1): Well I think it's like they usually stage the rollout and stuff. And I think they're supposed to be in the US, I saw initially. So
16:13
Dr. Thaon Simms: I'm
16:13
Unidentified speaker (UNKNOWN): interested to
16:14
Unidentified speaker (speaker_1): see that because that's almost like them going after um
16:18
Unidentified speaker (UNKNOWN): like
16:19
Unidentified speaker (speaker_1): budgeting apps. That's what I
16:20
Unidentified speaker (UNKNOWN): was thinking.
16:21
Unidentified speaker (speaker_1): Have you watched like people have done reviews on it and they showcase
16:24
Unidentified speaker (UNKNOWN): it?
16:25
Dr. Thaon Simms: Uh I haven't watched reviews. on it. I watched Claude's
16:28
Unidentified speaker (UNKNOWN): video,
16:29
Dr. Thaon Simms: Anthropic's video when they released it, but I just knew the things that it could do would be I would
16:34
Unidentified speaker (UNKNOWN): be
16:34
Dr. Thaon Simms: big. You know? Uh I think of like Bank of America's app or all of these other um banking apps that can showcase your expenses and so on. And I don't know. I I just think they're coming for a lot of these apps food.
16:48
Unidentified speaker (UNKNOWN): Not
16:48
Dr. Thaon Simms: necessarily the actual banks per se, but bank adjacent apps. So yeah.
16:53
Unidentified speaker (speaker_1): Do you think that I think an interesting thing is that like as much as like say
16:59
Unidentified speaker (speaker_1): um the one of the banks says, Okay, we're gonna allow for you know it to connect to your app.
17:07
Dr. Thaon Simms: Do
17:07
Unidentified speaker (speaker_1): you think that like I know that some people are gonna say, Hey, you know, what about the security issues? What about if it um
17:13
Unidentified speaker (UNKNOWN): you
17:13
Unidentified speaker (speaker_1): give it access and then it's like it just deletes or sends a transaction to a guy, uh a Nigerian prince?
17:20
Dr. Thaon Simms: Mm-hmm.
17:20
Unidentified speaker (speaker_1): Like, um,
17:23
Unidentified speaker (UNKNOWN): you
17:23
Unidentified speaker (speaker_1): know, I think that that is a risk. Um, and then people say, Oh, like you can supposedly set it up and then it should be able to not do that. But um, I think the interesting part is if one bank does it and then people really like it, don't you think that then puts pressure on the other banks to say, Hey, your your customer or your competitor offers it? And then people are really liking
17:43
Unidentified speaker (speaker_1): it. So it's almost like if the first one that does it may then put pressure on the others. I don't
17:47
Unidentified speaker (UNKNOWN): know.
17:52
Dr. Thaon Simms: I agree. Um I think there are a lot of security risks with AI right now, just on a whole. Like I I I would not give my banking app access and like Cloud or Chat GPT. I wouldn't give it access to my banking app because I just don't trust it, you know, uh to that level. Because imagine being able to
18:12
Dr. Thaon Simms: send all my money just like that to someone else and how do you reverse that? You can't. Um I think also if you look in the news too, the thing that concerns me the most is the amount of attacks or so on where these agents are leaving their containers and attacking certain
18:31
Dr. Thaon Simms: websites or organizations or companies and
18:34
Unidentified speaker (UNKNOWN): hacking
18:34
Dr. Thaon Simms: them and so on. I'm just like, yo, if if we have gotten to this point where artificial intelligence is this brilliant, um, you know, it's smarter than us as humans and then can because even now I think anthropic anthropic says it.
18:51
Dr. Thaon Simms: the upgrades are done by code you know it's it's AI altering its own code to make itself more like smarter and so on. Humans aren't really humans may stare it in the right direction but how much can it even stare it? You know there there's so many
19:11
Dr. Thaon Simms: I'm just worried about how quickly
19:14
Unidentified speaker (UNKNOWN): this
19:15
Dr. Thaon Simms: technology is developing and how limited I believe there is government regulation with this technology. Because I think there's there's nothing else that can stop um the growth of these companies outside of government intervention, in my opinion. Like rules, regulations, limitations, etc. Those are the
19:35
Dr. Thaon Simms: things that can provide the guardrails because leave it. up to capitalism ain't nobody stopping developing this. We're going to super intelligence, super AI, all of that very quickly. So um
19:50
Dr. Thaon Simms: which makes me want to touch on one other point. I don't know if you've been seeing the recent midterm elections and the results in the US. No? Um
20:01
Dr. Thaon Simms: well, all right, so I guess high level uh view of all of that. Uh basically what happened was I think a couple days, I don't remember which day it was. November
20:12
Unidentified speaker (UNKNOWN): a
20:12
Dr. Thaon Simms: few days ago though. I think the third. Or what happened was democratic control happened of the Senate and of the House. Uh that was the next day, so it it'll would be the fourth. Uh new Congress beginning. begins January 2027. So um I think
20:33
Dr. Thaon Simms: now with democratic control, and I don't know, you should there's a there's a video by Bernie Sanders that's on YouTube. You should you should watch that. Really, really good video. Um, but you know, we've had certain
20:46
Dr. Thaon Simms: um
20:47
Dr. Thaon Simms: members of the Democratic Party come out and speak. about things. Many things, many, many things. But one of the things that they touched on was artificial intelligence, especially data centers. So Senator Bernie Sanders and um AOC, they announced that
21:06
Dr. Thaon Simms: AI, um Alexandria Cortez, I don't remember how you pronounce her middle name or the the
21:13
Unidentified speaker (speaker_2): that
21:13
Dr. Thaon Simms: part of her last name. but AOC. Um she announced an AI data center moratorium act. Basically, all that is is they may delay or halt um the construction of certain data centers. Now,
21:33
Dr. Thaon Simms: data centers are crucial for, you know, cloud and chat GPT, but anthropic really. um as a company because that's where a lot of things are, that's where the compute happens, that's
21:44
Unidentified speaker (UNKNOWN): where a
21:45
Dr. Thaon Simms: lot of the the processing, the memory, all of those things, that's where a lot of the magic happens. All right. That's so halting the development of that would hurt um
21:57
Dr. Thaon Simms: would hurt those kind of companies, especially if let's say I'm anthropic. right? And I get investment. You spoke about Microsoft. Microf Microsoft is investing them. So Microsoft gives anthropic some, you know, some cash for maybe shares and so on. And it's based on maybe maybe it's predicated on the
22:17
Dr. Thaon Simms: construction of 10 data centers, right? Because now you'll have way more capacity because the demand is there, so you'll have way more capacity and so on. So and theoretically they're doing that because they're, you know, they're trying to anticipate that future demand and hopefully that turns into profits.
22:37
Dr. Thaon Simms: But now data centers, the construction may be halted. So your investment
22:44
Dr. Thaon Simms: was on the thesis that these data centers would be built and they would be operational by a certain time.
22:52
Dr. Thaon Simms: The time lost in construction of those things, uh, of those data centers, et cetera, also has a dollar dollar value, right? So they would be losing some some funds at some point based on their calculations as to how much money they'd be making, how many more subscriptions, etc. So,
23:11
Dr. Thaon Simms: I mean,
23:13
Unidentified speaker (UNKNOWN): and
23:13
Dr. Thaon Simms: that's billions. That's billions of dollars, I believe. So
23:17
Unidentified speaker (UNKNOWN): those
23:17
Dr. Thaon Simms: are so many steps, you know. Democrats winning, data centers can stop, billions can be lost. So I think that's a lot of people are skeptical about that. Um, and their position, the Democrats' position, like Bernie Sanders' position, is very clear, right? Um,
23:34
Dr. Thaon Simms: so I think that's something
23:37
Unidentified speaker (UNKNOWN): that
23:37
Dr. Thaon Simms: people are wary weary about. Like a lot of investors are weary about.
23:41
Unidentified speaker (UNKNOWN): So do you think that they
23:42
Unidentified speaker (speaker_1): would just go to another country then? If it's like the US blocking it unless well, unless the other countries then I'll do that. Because it
23:49
Unidentified speaker (UNKNOWN): when you imagine
23:50
Unidentified speaker (speaker_1): like this level of scale, it's like, okay, say the US says, hey, we don't want
23:54
Unidentified speaker (speaker_1): um
23:56
Unidentified speaker (speaker_1): we don't want data centers here, then
23:58
Unidentified speaker (UNKNOWN): what
23:58
Unidentified speaker (speaker_1): about someone say, okay, well, I'll do this. I want my
24:03
Unidentified speaker (speaker_1): um
24:05
Unidentified speaker (speaker_1): Yeah, like
24:06
Unidentified speaker (UNKNOWN): think
24:06
Unidentified speaker (speaker_1): about it. There are probably areas in the world where
24:08
Unidentified speaker (UNKNOWN): they
24:08
Unidentified speaker (speaker_1): have a lot of energy, like excess energy.
24:12
Unidentified speaker (UNKNOWN): Um,
24:12
Unidentified speaker (speaker_1): like they have a lot of solar power or whatever, then they might say, Hey, we laun we want to use this energy, or we have
24:19
Unidentified speaker (UNKNOWN): our energy
24:20
Unidentified speaker (speaker_1): compute. Yeah, and we want to use this, so we're looking for people to use the energy and looking for companies. to build it. Like it does
24:28
Unidentified speaker (UNKNOWN): I don't
24:28
Unidentified speaker (speaker_1): think they necessarily say, oh, it has to be a com um
24:32
Unidentified speaker (UNKNOWN): a
24:32
Unidentified speaker (speaker_1): company like Anthropic.
24:34
Unidentified speaker (UNKNOWN): It
24:34
Unidentified speaker (speaker_1): might be like, oh, we'd want it could have been like a manufacturer or something, but who's willing to pay more? So I it's not that I don't think it will it will hinder them, but it's gonna be hard it's honestly this is hard to stop. This is bigger than
24:50
Unidentified speaker (UNKNOWN): a lot of this can give us. Well,
24:52
Dr. Thaon Simms: I hear you. It is hard to stop. But even just let's say now you have to pivot because you you've identified a space. 'Cause remember, the data centers not just require energy, they also require some other key resources like water, right? Because and that's a big thing that people um
25:12
Dr. Thaon Simms: are looking at in terms of like use usage of um public water or
25:18
Unidentified speaker (speaker_2): you
25:18
Dr. Thaon Simms: know, just to that's like a lot of energy because water is needed to cool the system. Um so you do they have good access in these new areas. You now have to find the new areas in a different jurisdiction. So you have to go
25:31
Unidentified speaker (UNKNOWN): through
25:32
Dr. Thaon Simms: the legal parameters, centers, get um approvals, etc. Um, find areas with you know that water access and how is it gonna where's that runoff gonna go and all that kind of stuff and then you know are communities even um
25:47
Dr. Thaon Simms: in agreement with this kind of stuff because a lot of communities now have said they don't want data centers built. 'Cause if you're meds, data centers they because they use so much power that drives up the electricity cost of the entire community or communities around them. That's one. Two, resource intensive and not the right kind of
26:07
Dr. Thaon Simms: resources because they're not hiring a lot of community members to work. It's not like there's a factory, a new factory in town, and they're looking for community members to work certain hours or certain shifts doing this or that. Data centers barely, once they're fully built in, they barely require maintenance. And maintenance. and
26:27
Dr. Thaon Simms: in the terms of human resources. They don't need a lot of humans to be monitoring and so on. And the humans that they do need have need to have certain technical expertise. So it
26:38
Unidentified speaker (UNKNOWN): may
26:38
Dr. Thaon Simms: not even be people within that community that are even benefiting from the jobs and so on. So I
26:44
Unidentified speaker (UNKNOWN): think that's
26:45
Dr. Thaon Simms: there's there's so many different reasons why the uh them the Democrats are pushing back and have cr uh created this moratorium act or so on.
26:55
Dr. Thaon Simms: And so I can see it. But even if they do find solutions to all of those things, that delay costs billions. And so I feel like it will cost anthropic billions too. Um and then
27:09
Unidentified speaker (UNKNOWN): you
27:09
Dr. Thaon Simms: think about you know, they have now d let's say they borrowed money to find finance a lot of these data centers to be
27:17
Unidentified speaker (speaker_2): built.
27:18
Dr. Thaon Simms: They have to be paying interest on that. Whether it's built or not, they have to be paying back for those things. So, you know, at there's so much pressure now. In fact, I feel like Anthropic, I mean, if they were supposed to do this, they should have listed a long time. You see when SpaceX listed, they should have listed
27:33
Unidentified speaker (UNKNOWN): that
27:33
Dr. Thaon Simms: uh not now not after the midterms, you know. So
27:38
Unidentified speaker (speaker_2): yeah.
27:38
Dr. Thaon Simms: it's just it's poor timing in my opinion from them but it is what it is
27:45
Unidentified speaker (UNKNOWN): uh
27:46
Dr. Thaon Simms: you know but yeah there are there are
27:50
Unidentified speaker (UNKNOWN): uh
27:50
Dr. Thaon Simms: there's a lot to think about uh in in all of this because as as someone it I I find it so interesting because as someone that uses ai so much in his workflows and some so on, there's like a level of cognitive dissonance that I have where I'm just like, there are aspects of artificial intelligence that I definitely
28:10
Dr. Thaon Simms: do not agree with. But then again, it makes my life this much easier and it saves me time that I can spend, you know, uh
28:21
Dr. Thaon Simms: meds in my patients or meds in my family or doing all of
28:24
Unidentified speaker (UNKNOWN): these different
28:25
Dr. Thaon Simms: things or sleeping. which is I really have not been doing much of and medicine
28:30
Unidentified speaker (UNKNOWN): my health and
28:30
Dr. Thaon Simms: those kind of stuff. So, you know, um
28:34
Dr. Thaon Simms: yeah, cognitive dissonance, that's the phrase for sure.
28:39
Dr. Thaon Simms: Um
28:43
Dr. Thaon Simms: but uh I want to also talk about you spoke about other companies like um I guess Open OpenAI, AI.
28:51
Dr. Thaon Simms: um maybe Google,
28:53
Unidentified speaker (UNKNOWN): all these other
28:54
Dr. Thaon Simms: SpaceX, Metacross,
28:56
Unidentified speaker (UNKNOWN): DeepSeek, Meta,
28:57
Dr. Thaon Simms: Microsoft, oh
28:59
Unidentified speaker (UNKNOWN): Moonshot.
28:59
Dr. Thaon Simms: Moonshot, I think that's Kimmy. Like
29:01
Unidentified speaker (UNKNOWN): all these
29:02
Dr. Thaon Simms: different
29:02
Unidentified speaker (speaker_1): oh, okay.
29:03
Dr. Thaon Simms: Yeah.
29:04
Unidentified speaker (UNKNOWN): Um
29:06
Dr. Thaon Simms: how do you find that all of these things would affect anthropic? Like you know, it is it competition, is it synergistic?
29:15
Dr. Thaon Simms: like what what
29:16
Unidentified speaker (UNKNOWN): what's your
29:17
Dr. Thaon Simms: what's your take on a lot of these other companies and so on
29:20
Unidentified speaker (speaker_1): um that's a good question I mean I I think
29:25
Unidentified speaker (speaker_1): it's there both I guess pros and cons to it.
29:28
Unidentified speaker (UNKNOWN): Like
29:29
Unidentified speaker (speaker_1): on the pros side do you think that anthropic like my understanding is that for example the same people that founded Anthropic Antopic, like the what's his name Dario Ambassador Amade and his sister, I think they used to work at OpenAI. So they like gained experience. And there was a I think there are other people too that work at um
29:48
Unidentified speaker (speaker_1): Antopic now that you know came from OpenAI. What I see sometimes is with the labs, it's like old. I remember like But I think DeepSeek was the one, which I guess if people don't know, my understanding is that DeepSeek was basically this.
30:03
Unidentified speaker (speaker_1): Um
30:05
Unidentified speaker (speaker_1): they are
30:05
Unidentified speaker (UNKNOWN): a
30:06
Unidentified speaker (speaker_1): model. Yeah, it's like a Chinese. It's actually I think it was an interesting story. Like when I was looking it up, do you know that comparing DeepSeek to um
30:15
Unidentified speaker (speaker_1): some of the other
30:17
Unidentified speaker (speaker_1): companies, the there's like a main guy that owns most of the shares. There's this Chinese like investor guy, and it's like he, I think owns the majority of the company. It's almost like maybe 80% or something like that at that time. So he like has funded it a lot himself. And so he's not, I think yeah, he really has that control. I'm
30:37
Unidentified speaker (speaker_1): trying to remember his name. Liang
30:40
Unidentified speaker (speaker_1): Weng Wen Feng.
30:43
Unidentified speaker (speaker_1): I think that DeepSeek was the first one.
30:47
Unidentified speaker (speaker_1): Um, Because they my understanding is that they were they were basically constrained. right? So they were in China, and there was something at one point where like you couldn't get them on, like I think a lot of the graphics cards were made in the US, or and it's like there were restrictions on how many graphics cards you could get. So it's like they need the graphics cards, which is like the processors that process and make the
31:07
Unidentified speaker (speaker_1): models, and then also even allow you to run the models.
31:11
Unidentified speaker (speaker_1): And because they were so constrained, like they couldn't, it's almost like, oh, in the US, they could have a million GPUs or a thousand GPUs and solve a problem, but they it's almost like oh, DeepSeek only had a hundred. And then they had to kind of figure out a way to be more efficient. And then that way change the game in AI. It's almost like they I think they were the ones that kind of came up with the whole thinking thing
31:31
Unidentified speaker (speaker_1): and stuff like that. And then decide the other models, the other because DeepSeek is what they call open source, where it's like, oh, anybody can basically dial. download the model and run it.
31:40
Unidentified speaker (UNKNOWN): Um
31:41
Dr. Thaon Simms: and alter the code.
31:43
Unidentified speaker (UNKNOWN): Yeah,
31:43
Unidentified speaker (speaker_1): the other companies then looked at that and they're like, oh, whoa, we could do this too. And then they so there's pros where okay, if one person launches this, they could be like, hey, we should launch this. So it's like you'll
31:53
Unidentified speaker (UNKNOWN): see
31:53
Unidentified speaker (speaker_1): one model do this and you see one model do that. And then the other Although that's before you even get into the whole thing. Did you you ever heard of um distilling models?
32:03
Unidentified speaker (speaker_1): That was interesting when I heard about that, where it's almost like you can apparently there was a point. I'm trying to remember how much. I think anthropic
32:11
Unidentified speaker (UNKNOWN): actually You'd have
32:12
Dr. Thaon Simms: to explain what that means. Yeah,
32:14
Unidentified speaker (UNKNOWN): distillation.
32:16
Unidentified speaker (speaker_2): How much
32:18
Unidentified speaker (speaker_1): so distillation is basically well, I guess distillation itself, just like oh you have oil, you can separate
32:24
Unidentified speaker (UNKNOWN): you have
32:24
Unidentified speaker (speaker_1): petroleum that you mine out of the ground and then you can or drill out of the ground. You can separate into the different stuff, like you can get gasoline for
32:30
Unidentified speaker (UNKNOWN): the car, you can
32:31
Unidentified speaker (speaker_1): get diesel,
32:32
Unidentified speaker (UNKNOWN): you can
32:32
Unidentified speaker (speaker_1): get asphalt for the roads. So you
32:34
Unidentified speaker (UNKNOWN): separate
32:34
Unidentified speaker (speaker_1): it out basically. Um what they're doing with this is like say I
32:41
Unidentified speaker (speaker_1): um I'm anthropic and I make a really good model that everyone's like, wow, this is the best model. Um
32:49
Unidentified speaker (speaker_1): you, another company in China. could say, hey, we can make our model better by running it against Anthropic's model. So it's almost like you're stealing or siphoning um the knowledge from their model. I think yeah, Anthropic said it in their in their prospectus, they said five to seven percent of their revenue
33:09
Unidentified speaker (speaker_1): came from that.
33:11
Unidentified speaker (speaker_1): Usage that well, there's they say a usage that violated their usage policy policies and then also things like distillation. So think about that's and then to put that into perspective guy that that is
33:21
Unidentified speaker (speaker_1): one that's like one to two billion US revenue.
33:25
Dr. Thaon Simms: Wow.
33:26
Unidentified speaker (speaker_1): Yeah.
33:27
Unidentified speaker (speaker_1): That's how much is
33:28
Unidentified speaker (UNKNOWN): coming
33:28
Unidentified speaker (speaker_1): from the same distillation point. So there's I guess pros and cons. Um so okay if someone drops something you could kind of say oh we could do that. That's a good idea. Um, But then they could be stealing I ideas, guess, from you. Um and I think you even touched on another good point, the whole thing of
33:43
Unidentified speaker (UNKNOWN): I
33:43
Unidentified speaker (speaker_1): mean, usually when we're looking at
33:45
Unidentified speaker (UNKNOWN): I've
33:46
Unidentified speaker (speaker_1): learned is that when you look at, I guess Buffett talks about it, when you're looking at businesses, you want to
33:51
Unidentified speaker (speaker_1): a good business has like what you'd call like a good moat where it can um
33:57
Unidentified speaker (speaker_1): it's not easy to kind of switch between um their products. So it's almost like, oh. this business is the only one that can do this. Like you're the only one selling this product in the area, and people want that product. So you you are you don't really have to um like drop your prices and stuff like that. So if you have a high
34:17
Unidentified speaker (speaker_1): demand and you're the only supply, you can be you know keep selling the product, keep selling product and stuff like that, as long as the demand stays the same. Uh I've heard a lot of people discuss, oh, the whole concept of with these models. Like for example, when people are using it for certain stuff, they could just go to Chat GPT, they just could just go to Claude, Gemini, whatever, and just ask it, oh, what color
34:37
Unidentified speaker (speaker_1): should I dye my hair? What should I what should I make for breakfast? Or those kind of stuff. So there's there's you can just easily siphon between them and it, oh, I hit my limit here, let me just go over to this other one. Um,
34:47
Unidentified speaker (speaker_1): But what I was seeing in the prospectus with Anthropic is that what they're trying to do is
34:53
Unidentified speaker (speaker_1): it's it's almost like they're trying to integrate into the business or really form like really good relationships with some of these companies. And they're trying to say, okay, like they will sometimes apparently send a consultant who works at Anthropic and then specifically send them to that business and help that business to integrate it. And they're trying to, I guess, as much as possible kind of lock in that customer to say,
35:13
Unidentified speaker (speaker_1): Hey, um
35:15
Unidentified speaker (speaker_1): you should really be abused to orb products. Um, so
35:19
Unidentified speaker (UNKNOWN): I think
35:19
Unidentified speaker (speaker_1): that's kind of interesting. So it's almost like they're trying to build these, they're really trying to lock it in. I guess we'll see if they um if that end up actually happening, if that ends up actually happening, because um if it doesn't, then it could be kind of like a race to the bottom where oh um Chat GPT said they'll do it for 10 million, and anthropic Tappe
35:37
Unidentified speaker (UNKNOWN): said
35:37
Unidentified speaker (speaker_1): they'll do it for 8 eight million. Let me just go the cheaper one because but then if you can actually show that hey my model is better or my or system so it's not just the model it's like okay
35:47
Unidentified speaker (UNKNOWN): how do you
35:47
Unidentified speaker (speaker_1): integrate the model with everything else you
35:50
Unidentified speaker (UNKNOWN): actually
35:50
Unidentified speaker (speaker_1): get a better outcome. So oh I spend 10 million with anthropic but I'll get 50 million. I spend five million with OpenAI but I won't get the to fifty 50 million I'll only get maybe ten million.
36:01
Dr. Thaon Simms: So
36:01
Unidentified speaker (UNKNOWN): if
36:01
Unidentified speaker (speaker_1): it I guess if you're a person your percentage gain is better with that one. So I think that's the other thing when you're evaluating this stuff. It's not just look at the cost. You also want to look at what's the output and making sure that
36:11
Unidentified speaker (UNKNOWN): the
36:11
Unidentified speaker (speaker_1): output you get from it is you get a return on your investment.
36:15
Dr. Thaon Simms: Yeah. That's true. That's true. And there are other things that people also look at too. Uh not just that, but like even trust. Like for me, I find that trust is a big thing. Um even so you you know I think I used Cloud when it initially you know came out and so on. But I was really using Chat
36:35
Dr. Thaon Simms: GPT heavily and it's not until I don't know if you remember this incident that happened but I think the government wanted to use Anthropic for some autonomous
36:45
Dr. Thaon Simms: weapons. I don't know it had something to do with um
36:49
Dr. Thaon Simms: warfare and
36:52
Dr. Thaon Simms: Anthropic said no. And then they went to Chat GPT or OpenAI. OpenAI said yes to use their basically their model for I think it was autonomous drone strikes or something like that. I can't remember what it actually is, don't quote me. However, when that happened, there was this big outcry on like
37:12
Dr. Thaon Simms: Twitter and all of these places about, you know. OpenAI selling out. And so a lot of people switched to cloud during that time. I was one of those people that said, All right, let me start using cloud again and see what it's about. Just because everyone was doing it as just like just hop
37:27
Unidentified speaker (UNKNOWN): on
37:27
Dr. Thaon Simms: the wave. So I feel like those kind of things in terms of trust, administration, um, you know, there are other things that I think people use cloud for that they prefer in terms of even.
37:39
Dr. Thaon Simms: its output in terms of how it writes, um and so on. So my friend uses cloud because they they say it writes really well. Um so it it depends, depends, depends on the person. One thing I didn't touch on though, in terms of the I guess the business itself is um
37:59
Dr. Thaon Simms: direct versus cloud distribution. Um
38:05
Dr. Thaon Simms: direct meaning so direct meaning the customer actually buying from Anthropic, buying cloud, like access to cloud from Anthropic through its website. So if you're doing a a subscription, you pay you $20 a month or $200 a On month, Monday say if you're if you're brawling broad in uh for the different
38:24
Dr. Thaon Simms: plans and you buy directly from Anthropocon Anthropic. So Anthropic makes that sale connection, agrees on commercial terms, blah blah blah. You can also use the API. I think you can purchase like API credits via straight from Cloud or straight from Anthropic.
38:39
Unidentified speaker (UNKNOWN): What is
38:42
Unidentified speaker (UNKNOWN): it?
38:43
Dr. Thaon Simms: API is like how the model interacts with other software. I think that's my understanding of API. I don't even know what API actually means. That's so funny. Do you know what API actually means?
38:55
Unidentified speaker (speaker_1): Uh application programming interface.
38:59
Dr. Thaon Simms: Ah, there you go.
39:00
Unidentified speaker (UNKNOWN): So
39:00
Unidentified speaker (speaker_1): it's allowed the interface, a connection
39:04
Unidentified speaker (speaker_1): between applications.
39:07
Dr. Thaon Simms: Cool.
39:09
Dr. Thaon Simms: Yeah. So that's the direct, but then you have the cloud distribution now where the customer already buys services from a large cloud company, like let's say Google Google Cloud or so on. And then they can access cloud through that company's platform. So through Google's platform. So
39:28
Unidentified speaker (speaker_1): is that middleman?
39:30
Dr. Thaon Simms: Yeah, it's like a middleman.
39:31
Unidentified speaker (speaker_1): Yeah.
39:32
Unidentified speaker (UNKNOWN): Google is
39:33
Unidentified speaker (speaker_1): paying for the API and then they're selling that to you as a
39:36
Unidentified speaker (UNKNOWN): part of the
39:36
Unidentified speaker (speaker_1): service. So they're bundling it, I guess.
39:39
Dr. Thaon Simms: Yeah. And so the cloud company would help because they help the customer, they help anthropic, you know, find customers, connect them, pay for the service too. So you're paying. So if I'm buying, if I'm accessing cloud through Google Cloud, I'm paying Google Cloud, and then Google Cloud has to pay Cloud. Right? So it can
39:59
Dr. Thaon Simms: combine the bill or so on. But uh I think of it like
40:04
Dr. Thaon Simms: uh like a bakery, right? Direct is you buy directly from the bakery because you like those products, and then marketplaces like you're buying from a grocery store that sells those baked goods, the same bakery's bread, and so on. So the bakery baker still makes the bread, the grocery store
40:24
Dr. Thaon Simms: helps um shoppers to find and pay for it. And they keep a fee as well. So that's how I see it. And when you look at the numbers based on the prospectus, it said that customers purchased through cloud provider marketplaces accounted for 47%
40:40
Dr. Thaon Simms: of um 2025 full-year's revenues. And then 44% of the 2026 20s 26's is revenue. So it's around the same in mid-40s. in terms of their revenues. So that's still a big part of it. Now, when we were talking about competitors and so on, I started
41:00
Dr. Thaon Simms: to think even about Google because I think Claude's relationship with Google is quite interesting still. Because uh Google matters so much to Anthropic in many different roles, many different ways. And I kind of had to separate it out based on my understanding as one, a
41:20
Dr. Thaon Simms: competitor. So Google also sells
41:24
Dr. Thaon Simms: access to their AI, which is Gemini,
41:28
Dr. Thaon Simms: that customers can choose from instead of cloud. So you can literally choose Gemini if you want, Gemini models, AI features, etc. And Gemini has
41:37
Unidentified speaker (UNKNOWN): it's is
41:38
Dr. Thaon Simms: very good at. certain things too. Better than Cloud at like multimodal stuff and so on. So, you know, it's analyzing videos and images. I think it does a way better job. I don't even know if Cloud can analyze videos like that. Um,
41:52
Dr. Thaon Simms: you know, Gemini can do a lot of different things in terms of summarizing documents and so on, but Google also has workspace, and I think that's what makes it a bit different from cloud.
42:04
Unidentified speaker (speaker_2): Uh,
42:04
Dr. Thaon Simms: because workspace is access to Gmail, and everybody almost everybody has Gmail, you know, Google
42:11
Unidentified speaker (UNKNOWN): Docs,
42:12
Dr. Thaon Simms: Google Sheets, and so on. So they can literally you'll see Gemini across all of their workspace though. Um and then if customers can get AI inside of tools they already use, like me, I already use Google Docs a lot, then they may not need to necessarily buy a separate cloud subscription. But it doesn't mean
42:31
Unidentified speaker (UNKNOWN): that
42:31
Dr. Thaon Simms: Gemini and Cloud are interchangeable because as I said, they have their their pros and cons, both of them.
42:38
Dr. Thaon Simms: I would prefer Cloud for certain tasks and I would never use Gemini for certain tasks. But in the And though
42:46
Dr. Thaon Simms: um in another sense, between Google and Cloud, as a supplier, Google provides computing capacity that Anthropic needs. I don't know if you've seen that, but you know, a lot of the compute that Anthropic needs, it's Google that provides that. So whether it's training the
43:06
Dr. Thaon Simms: models or even running those large AI models, they need chips, right? They need servers, they need data center capacity. they need a lot of those things. And Google actually supplies some of this capacity to them. I don't know if you've seen Google's TPUs, TPU chips, there's a specific chips uh that Anthropic uses for
43:26
Dr. Thaon Simms: compute power. Uh so you know, I feel like Anthropic is pretty reliant on Google. And even as a distributor, the same cloud, Google Cloud and so on, you can offer cloud through that platform. and so on. So meets that Google will be handling the billing, the collection for the marketplace purchases, and
43:46
Dr. Thaon Simms: then remit that to Anthropic share after some fees. Um you can kind of see that in their numbers if you look through the numbers. And it's so funny because Google has invested in Anthropic. Like you know um they've also put money into Anthropic, so they would benefit if Anthropic grows. So
44:06
Dr. Thaon Simms: it's it's it's kinda interesting because the relationship is so mixed. Like Google can help Anthropic sell and operate Cloud while also competing with it. I don't know if you can see that. So it's just it's just a very interesting relationship that they have. So they just I guess uh at Cloud and Anthropic
44:26
Dr. Thaon Simms: anthropic is growing in a way where it can still be independent from Google but they still do rely on Google for a lot of different things. So I I don't know if you can say that mixed relationship.
44:40
Unidentified speaker (UNKNOWN): But
44:40
Dr. Thaon Simms: yeah
44:42
Unidentified speaker (UNKNOWN): it's
44:42
Dr. Thaon Simms: just something something I thought of when when reading
44:45
Unidentified speaker (speaker_1): there's a lot one of the things that um I guess make I I know it's something that people talk about is that hey, first of all, um first of all, uh OpenAI and topic, they're not um profitable. And then I guess a lot of people are saying, hey,
45:01
Unidentified speaker (UNKNOWN): I
45:01
Unidentified speaker (speaker_1): mean obviously you'd want a business, ideally, if a business could not make money or make profits, you'd probably want it to make profits. Um
45:09
Unidentified speaker (speaker_1): and I think
45:12
Unidentified speaker (speaker_1): one of the things that is
45:15
Unidentified speaker (speaker_1): is hard to or is interesting with the AI models, right? So if you say I'm Anthropic and I want to make a new model and I'm gonna call it this is
45:28
Unidentified speaker (UNKNOWN): um
45:28
Unidentified speaker (speaker_1): Opus. And Opus is basically one of Anthropic's best models that they'll serve to like regular people, I guess.
45:35
Dr. Thaon Simms: And I
45:36
Unidentified speaker (speaker_1): think right now they're at Opus five 5.5. point five. So say I'm like, okay,
45:40
Unidentified speaker (UNKNOWN): I'm
45:40
Unidentified speaker (speaker_1): gonna do Opus 6, right? So it's supposed
45:43
Unidentified speaker (UNKNOWN): to
45:43
Unidentified speaker (speaker_1): be better, it's supposed to be smarter, it's supposed to be more efficient, supposed to be whatever. I spend, so we as we said, you have to buy all of these processors, computers, and do all of this stuff. You have to get data, you have to sometimes buy the data, or you just scrape it online.
46:00
Unidentified speaker (speaker_1): Um, and then you train them on. a model. And say
46:05
Unidentified speaker (speaker_1): it costs a hundred um I don't even like this. It's why when you meant the size of this, say it costs a billion, one billion dollars to train
46:15
Unidentified speaker (speaker_1): Opera 6, right? So I spend that $1 billion, it's I've now made it, right? And that included the data, it included using like people I that had to pay, all of that stuff, right? So we have Opus Six
46:28
Unidentified speaker (UNKNOWN): now.
46:29
Unidentified speaker (speaker_1): You then have to use more computers to actually process it. So it's almost like you have to the same computers you use to make it, you then have to use those same computers to serve it or allow
46:40
Unidentified speaker (UNKNOWN): people
46:40
Unidentified speaker (speaker_1): to use it. Like when people go and type, Oh, um, should I buy this or should I not buy this, whatever?
46:47
Unidentified speaker (UNKNOWN): And
46:47
Unidentified speaker (speaker_1): then that could theoretically what if that costs, I don't know, a hundred million depending on the usage or whatever?
46:53
Unidentified speaker (UNKNOWN): I
46:53
Unidentified speaker (speaker_1): think one of the tough parts is that
46:55
Unidentified speaker (UNKNOWN): say
46:56
Unidentified speaker (speaker_1): Opus 6 is like the best model ever. Like it's out, everyone's saying it's better. It's better than OpenAI, it's better than Google, it's better than DeepSeek, it's better than everyone, right? It's like say it's the best for like three months or six months, and then OpenAI AI makes their model and they call it, oh, this is um sold. soul
47:15
Unidentified speaker (UNKNOWN): uh
47:15
Unidentified speaker (speaker_1): I don't know seven or something like
47:17
Unidentified speaker (UNKNOWN): that.
47:18
Unidentified speaker (speaker_1): Right? It's it's then think about it. What do you do as an topic now? If their model is better than yours and people are actually using it and they're saying it's better, it's better. What are you then gonna do?
47:30
Dr. Thaon Simms: I don't
47:30
Unidentified speaker (UNKNOWN): know what would you run up
47:31
Dr. Thaon Simms: them.
47:32
Unidentified speaker (UNKNOWN): So
47:32
Unidentified speaker (speaker_1): you're then gonna have to then spend more money and then what do you have to spend two billion now?
47:37
Unidentified speaker (UNKNOWN): And
47:37
Unidentified speaker (speaker_1): then
47:38
Unidentified speaker (UNKNOWN): it
47:38
Unidentified speaker (speaker_1): just it feels like it's is is always kind of gonna be
47:42
Unidentified speaker (speaker_1): um in competition, I guess. Like you're always like
47:45
Unidentified speaker (UNKNOWN): they're
47:45
Unidentified speaker (speaker_1): gonna be spending like your model will be good for a while and then their model will be good for a while. Because I mean look at it, OpenAIS had the best models at some points, and then some people say, Oh, and Apic has the best models and it's this and this and this. Um I think that's something that is
47:59
Unidentified speaker (UNKNOWN): I
47:59
Unidentified speaker (speaker_1): guess kind kinda of tough to it's it's like it really takes a lot of money to. continuously like fund this. It's like you spend the money, it's not like oh you just it's not like oh like a Trans Jamaica or a highway company, you just um you build a highway and yes you do a little maintenance here and there, but it's nowhere near the level of building the highway and you can
48:19
Unidentified speaker (speaker_1): just reap rewards from it for years. Um, unless theoretically they made a model and it's like the model just was the best and no one needed a better model, but then
48:30
Unidentified speaker (UNKNOWN): I
48:30
Unidentified speaker (speaker_1): don't know. What do you think? How do you, I guess, view that? Like, do you think that there will be a point that um
48:38
Unidentified speaker (speaker_1): that makes sense? Like, how do you view that in terms of like the actual business model of being an AI company?
48:46
Unidentified speaker (UNKNOWN): Uh
48:48
Dr. Thaon Simms: I think of it like
48:51
Dr. Thaon Simms: this has to continue. That's how I see it actually.
48:55
Unidentified speaker (UNKNOWN): Uh,
48:55
Dr. Thaon Simms: in
48:55
Unidentified speaker (UNKNOWN): terms
48:55
Dr. Thaon Simms: of one upping the next model.
48:58
Unidentified speaker (UNKNOWN): Just
48:58
Dr. Thaon Simms: like in the smartphone industry, where you
49:02
Unidentified speaker (UNKNOWN): make an
49:02
Dr. Thaon Simms: iPhone 17, but their other Samsung comes out with their Galaxy, whatever, and their camera system is so much better. In
49:10
Unidentified speaker (UNKNOWN): the
49:10
Dr. Thaon Simms: next iteration of iPhone 18, you know, there's certain things you may have to improve on, whether it's battery. life, whether it's uh camera quality and low light or so on. Now they have variable aperture, all of these different things, promotion, etc. So you know,
49:24
Unidentified speaker (UNKNOWN): there
49:24
Dr. Thaon Simms: I think in terms of just how you know the tech industry works, you have to constantly be getting better. It's either that
49:33
Unidentified speaker (UNKNOWN): or
49:33
Dr. Thaon Simms: you
49:33
Unidentified speaker (UNKNOWN): We
49:33
Dr. Thaon Simms: die, you know. Um
49:36
Dr. Thaon Simms: so
49:38
Dr. Thaon Simms: the only
49:40
Dr. Thaon Simms: the two issues I find are one funding. I don't necessarily think the funding, the current funding, um how they are funding these companies is sustainable. Especially with all of
49:52
Unidentified speaker (UNKNOWN): these
49:52
Dr. Thaon Simms: other things that may come into play, like regulatory stuff or even them just not turning Earning a profit, maybe them not being able to pay back investors. Uh if it does pose a cash burden, like if they're not actually generating enough cash or so on.
50:07
Unidentified speaker (speaker_2): Mm-hmm.
50:08
Unidentified speaker (UNKNOWN): And
50:08
Dr. Thaon Simms: that's why they talk about this whole AI bubble. Uh, I think that's that's actually a big reason and a for a lot of the apprehension around AI on a whole right now, the so called possible bubble that may happen. Um and it's possible, right? You know, because the value
50:28
Dr. Thaon Simms: that they are trading at now,
50:31
Dr. Thaon Simms: you know, I a lot of investors, like you know, technology.
50:36
Dr. Thaon Simms: Well, a lot of these AI companies, maybe SpaceX, etc. Yeah.
50:40
Unidentified speaker (UNKNOWN): Um
50:40
Dr. Thaon Simms: but you know technology, I think technology can create lasting value. You know
50:49
Unidentified speaker (UNKNOWN): I
50:49
Dr. Thaon Simms: think AI itself is a revolution revolutionary technology for sure. But then the security itself, the listed company, etc., or the infrastructure, those things can be priced for unrealistic returns. So I've been seeing like a lot of bubble claims about valuation.
51:09
Unidentified speaker (speaker_1): That's also hilarious. There was a video that said so they were talking, sat they were saying, Oh, there's the AI bubble. And then
51:15
Unidentified speaker (UNKNOWN): they said
51:16
Unidentified speaker (speaker_1): there's the bubble of videos on the AI bubble.
51:21
Unidentified speaker (speaker_1): There were that many videos where people are like, Oh, the AI bubble, the AI bubble's gonna pop it.
51:27
Dr. Thaon Simms: And we never know.
51:33
Dr. Thaon Simms: You know, So
51:34
Unidentified speaker (UNKNOWN): and
51:35
Dr. Thaon Simms: then what would that even look like? Are these the catalysts that we're seeing now that will make this AI bubble pop? You know, uh
51:42
Dr. Thaon Simms: can the models become it you know, interchangeable? Are is it infrastructure overbuilt? Like are we just, you know, massively um creating all these data centers, but the use case for AI just isn't necessarily there and it's more for training? Is it is capital expensive? Is capital way too expensive now?
52:03
Dr. Thaon Simms: Because because it's seen now as very risky based on everything that we're seeing, the interest rates, let's say you're borrowing as anthropic and you're borrowing some funds from some bank because you really need to you know you need this billion dollars. So you're borrowing funds from either Google or whatever, right? But you have to pay them back. You
52:23
Dr. Thaon Simms: can either
52:23
Unidentified speaker (UNKNOWN): pay
52:24
Dr. Thaon Simms: them back in shares or in cash, or you can figure out how you pay them back. But let's say it is cash. If you're not generating the cash, the returns are just not there. Um that's a big problem. So companies are going to be more hesitant to lend you money and then the terms are not gonna be great. Right? I if I used to lend you this at eight percent, kinda risky, you know.
52:44
Dr. Thaon Simms: So I may have to lend you this at twelve 12% percent for the next couple of years and let's see what that what happens there. And then when companies get squeezed like that, you know what does that do for the stock price? So I don't know. Um
53:00
Dr. Thaon Simms: it's a it's a it's a tricky one to see where AI will go from here, but I just think that currently how it's being funded, like all these billions of dollars being poured into it, I just
53:11
Unidentified speaker (UNKNOWN): don't think
53:11
Dr. Thaon Simms: it's that sustainable. Uh subscriptions have to increase. tremendously for that, or they'd probably have to just pivot and have additional revenue lines, like you know what a lot of these other companies are doing, which is advertising. I've I don't know if you've been seeing that,
53:27
Unidentified speaker (UNKNOWN): but
53:28
Dr. Thaon Simms: you
53:28
Unidentified speaker (UNKNOWN): know,
53:28
Dr. Thaon Simms: ad revenue pays. Anthropic, I think the CEO specifically said he's not going down that route, but I'm like,
53:36
Unidentified speaker (speaker_2): yeah. you
53:37
Unidentified speaker (UNKNOWN): know,
53:38
Dr. Thaon Simms: give it a year or two.
53:39
Unidentified speaker (speaker_1): Pressure bus pipe.
53:41
Dr. Thaon Simms: Exactly. Pressure bus pipe. So give it a year or two and make we see.
53:46
Unidentified speaker (UNKNOWN): Make
53:46
Dr. Thaon Simms: we
53:46
Unidentified speaker (UNKNOWN): see.
53:47
Unidentified speaker (speaker_1): Look at this.
53:50
Unidentified speaker (UNKNOWN): Yeah,
53:51
Unidentified speaker (speaker_1): this is how I was saying the AI bubble video bubble
53:53
Unidentified speaker (UNKNOWN): for those that not
53:54
Unidentified speaker (speaker_1): seeing the screen.
53:58
Unidentified speaker (speaker_1): Yeah.
53:59
Unidentified speaker (UNKNOWN): This is real.
54:01
Unidentified speaker (speaker_3): It's real.
54:02
Unidentified speaker (speaker_1): Yeah.
54:02
Unidentified speaker (UNKNOWN): I
54:03
Unidentified speaker (speaker_1): mean one of the other things that stood out to me from the prospectus, did you what did you think of the you saw the amount of like share classes that there are?
54:11
Dr. Thaon Simms: Yes.
54:12
Unidentified speaker (UNKNOWN): Like
54:12
Unidentified speaker (speaker_1): there's different and then each of them has different stuff. Like some of them you you meant that they the founders of the company, I don't know if you do you think this
54:19
Unidentified speaker (UNKNOWN): is
54:19
Unidentified speaker (speaker_1): like a good or a bad thing. They basically the founders control the company, you know. Um they
54:25
Unidentified speaker (UNKNOWN): own
54:26
Unidentified speaker (speaker_1): like I think over 50% of the shares. Well,
54:30
Unidentified speaker (UNKNOWN): because
54:30
Unidentified speaker (speaker_1): like the shares that you're buying into now, which is like the or possibly buying into the class A shares. It's like the cloud the founders own a separate class called Class. Well, I should probably maybe I can show it on
54:43
Unidentified speaker (speaker_1): uh screen. Like there was this class A, class N. Hold on.
54:49
Unidentified speaker (speaker_2): Yeah.
54:55
Unidentified speaker (speaker_1): Class
54:56
Unidentified speaker (speaker_1): A, N, F T, E, and then the preferred shares, which will then convert into um so the class A is that what they're selling in the IPO? Class N
55:07
Unidentified speaker (UNKNOWN): is
55:07
Unidentified speaker (speaker_1): that the ones that they gave to like Amazon, well, Amazon or Google bot, so they don't actually have um voting rights. So they can't say, oh, don't. do this or don't do this, pay a dividend, don't pay a dividend. Class F, which I found hilarious. That it's I think it's I think the F is supposed to stand for founders. So I thought the F was almost like for F you because they're saying, Hey, this is our ship. Like if
55:27
Unidentified speaker (UNKNOWN): we wanna do
55:28
Dr. Thaon Simms: something,
55:29
Unidentified speaker (speaker_1): we control it unless we have a sell like this is
55:32
Unidentified speaker (UNKNOWN): our ship
55:33
Unidentified speaker (speaker_1): we're starting at that point, which is actually kinda kind of big for like even if you meant for a company their size to be able to
55:40
Unidentified speaker (speaker_1): for them to still own so much and then go public.
55:50
Unidentified speaker (speaker_1): I don't know. What did you think of that? Like they do you think because the thing is like I mean, if they're
55:56
Unidentified speaker (speaker_1): theoretically, do they are they gonna be the ones making the best decisions for the company? Um or even
56:03
Unidentified speaker (UNKNOWN): for I guess the
56:03
Unidentified speaker (speaker_1): shareholders, like what do the shareholders want? Like are is it are they gonna like what if they're like, hey, um, we don't want to do this or we don't agree here? Um, we're gonna do this or do that. Like,
56:14
Unidentified speaker (speaker_2): can't really say anything, you
56:16
Unidentified speaker (UNKNOWN): know?
56:17
Dr. Thaon Simms: That's that's true. Um
56:20
Dr. Thaon Simms: well, just So as an investor, you just have to be comfortable with who you're buying these shares from and who is steering the ship. Uh
56:29
Dr. Thaon Simms: you know, you i if you believe in Dario's vision, yeah, you buy in and so on. I I think
56:39
Dr. Thaon Simms: I don't know, I
56:41
Dr. Thaon Simms: I I'm not really putting a lot of credence to the class F shares versus to the different classes. Um I'm you
56:49
Unidentified speaker (UNKNOWN): know,
56:49
Dr. Thaon Simms: because I'm not gonna participate. So maybe that's probably why. Right? Maybe I'm looking
56:52
Unidentified speaker (UNKNOWN): at it. So you wouldn't buy
56:53
Unidentified speaker (speaker_1): it at all?
56:54
Dr. Thaon Simms: No. Or
56:55
Unidentified speaker (speaker_1): it's just like at the price?
56:57
Unidentified speaker (UNKNOWN): I
56:57
Dr. Thaon Simms: just wouldn't buy it. I don't think maybe at the price too. It would have to calm down a bit. And hey, I don't I don't know. I don't know.
57:04
Unidentified speaker (UNKNOWN): Because we still don't really
57:05
Unidentified speaker (speaker_1): we still don't know um confirmation at the time of recording. We still don't know confirmation of price
57:11
Unidentified speaker (UNKNOWN): how
57:11
Unidentified speaker (speaker_1): much they're well, even exactly how much they're raising. I'm seeing stuff. are
57:15
Unidentified speaker (UNKNOWN): saying it
57:15
Unidentified speaker (speaker_1): might be like a hundred billion and they may raise at like one point five trillion valuation or something like that.
57:22
Unidentified speaker (UNKNOWN): Um
57:23
Unidentified speaker (speaker_1): but I guess we'll have to see the final
57:25
Unidentified speaker (speaker_1): um yeah the final information and stuff.
57:29
Dr. Thaon Simms: Yeah, there's
57:30
Unidentified speaker (UNKNOWN): there's too
57:31
Dr. Thaon Simms: much we don't really know yet. But um
57:34
Unidentified speaker (UNKNOWN): I
57:34
Dr. Thaon Simms: don't know man. Uh
57:37
Dr. Thaon Simms: let's see. Let's see. actually. If it's if it makes sense, then of course I'll buy, but I don't see how this is making sense for me personally. Like I can see other things making sense, but as especially with how the Democrats won. And don't get me wrong, I'm not into politics, right? I yeah,
57:57
Dr. Thaon Simms: I don't really med those things just only when it influences uh a financial decision will I be looking at this and commenting on it on this podcast. But, you know, I have no affiliation. I'm not a US citizen. I can't vote. I I didn't vote. So don't don't crucify me. But with the
58:17
Dr. Thaon Simms: Democrats winning um, you know, both the, you know, the Senate and the House, I think that it's going to be a bit more difficult for AI to make the significant leaps and bounds that it has been making for the past couple of years. I think there
58:36
Unidentified speaker (UNKNOWN): will
58:37
Dr. Thaon Simms: be a bit more um they'll scrutinize things a bit more. They'll they won't just push things through. I think there will be a lot of pushback, especially based on public outcry for you know, the the the construction of data centers, um, what it can do for even like climate change, etc. I
58:57
Dr. Thaon Simms: think that's the party that will take those things a bit more into account. So with that looming, I am hesitant. So I don't think I would be partaking. Partake in.
59:10
Unidentified speaker (speaker_2): But
59:10
Dr. Thaon Simms: hey, let you guys can let us know in the comments down below. W what is this something you would buy? They would take my advice for it, because I'm not a software engineer either. I barely know anything in this whole realm. This is just off of this read for this prospectus. But um there could be a lot of other factors that can influence this that I'm not even looking at. Uh
59:30
Dr. Thaon Simms: you know, but speak with your licensed financial advisor if you guys are going to invest in this. Preston, what do you think? Is this something you would buy in?
59:38
Unidentified speaker (UNKNOWN): I
59:38
Unidentified speaker (speaker_1): want to buy in at IPO. I'd probably need some. There'd probably have to be some sort of um
59:46
Unidentified speaker (speaker_1): fear reaction before I'd consider like something that would the market or people feel is really bad, or it actually could be really bad for them. And it would
59:56
Unidentified speaker (UNKNOWN): have to probably
59:57
Unidentified speaker (speaker_1): drop a lot. And then I'd I'll have to say, okay, yeah, but I think there's this. Um but yeah, I think it's something. similar to I guess I'll continue to look at but
1:00:06
Unidentified speaker (speaker_1): no I'll stick stick with things that I can more easily predict right now.
1:00:10
Dr. Thaon Simms: Mm-hmm. Okay, fair. Yeah. And who knows, maybe a data center will be built in Jamaica.
1:00:16
Unidentified speaker (speaker_2): Mm-hmm.
1:00:17
Dr. Thaon Simms: Yeah. Maybe Wigton will build out the solar before it's For it. who knows. We shall we
1:00:24
Unidentified speaker (speaker_1): shall see.
1:00:25
Dr. Thaon Simms: One
1:00:25
Unidentified speaker (UNKNOWN): to
1:00:26
Dr. Thaon Simms: watch. Anyways, guys, if you love this episode, let us know. Leave a like, hit that subscribe bell. Please ring that notification bell because we make episodes like this every single week and drop them every Friday. Uh, hope you guys love it. And if you did, please subscribe. Thank you so much. And like always, folks, stay
1:00:45
Dr. Thaon Simms: limitless.
1:00:47
Unidentified speaker (speaker_1): Peace.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.
All insights