Ep19. State of Venture, AI Scaling, Elections | BG2 w/ Bill Gurley, Brad Gerstner, & Jamin Ball
What this covers
Open Source bi-weekly convo w/ Bill Gurley and Brad Gerstner on all things tech, markets, investing & capitalism. This week they are joined by Jamin Ball (Partner at Altimeter) to discuss the state of venture capital and incentives, implications of overcapitalization, the future of power law returns, AI’s impact on investment strategies, the future of AI and CapEx, the market’s reactions to the 2024 election, the evolution of search and AI integration, Nasa, SpaceX, & more. Enjoy another episode of BG2.
Timestamps:
(00:00) Intro (00:52) Bill’s Dad | from Nasa to SpaceX (06:20) State of VC and Incentives w/ Jamin Ball (12:31) Implications of Overcapitalization (25:08) The Future of Power Law Returns (38:28) The Impact of AI on Investment Strategies (46:00) Future of AI and Capital Expenditure (52:21) Market Reactions to 2024 Election (01:00:45) Evolution of Search and AI Integration
Available on Apple, Spotify, www.bg2pod.com
Produced by Benny Beausoleil Music by Yung Spielberg
Follow: Brad Gerstner @altcap https://x.com/altcap Bill Gurley @bgurley https://x.com/bgurley BG2 Pod @bg2pod https://x.com/BG2Pod Jamin Ball @jaminball https://x.com/jaminball
Shownotes Clouded Judgement Substack: https://cloudedjudgement.substack.com/
#BillGurley #BradGerstner #Bg2Pod
Source description (no synthesized summary yet).
Venture capital has transitioned from a high-margin cottage industry to an institutionalized asset class with misaligned incentives between GPs and founders, where larger fund sizes enable GPs to get rich from management fees regardless of outcomes, leading to overcapitalization of companies and erosion of the medium-outcome return paths that historically created value for all stakeholders.
- A $4B fund generates ~$80M/year in management fees (2%), enabling GPs to become wealthy independent of fund performance, whereas historically smaller funds required carry to get rich
- Overcapitalized companies spread resources across too many initiatives, destroy focus, and eliminate the $100-200M acquisition outcome path that creates life-changing returns for founders and employees
- Large funds must spread bets across many companies to justify capital deployment, making it mathematically difficult to achieve historical venture-like returns while incentivizing faster deployment over better selection
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
Venture fund management fees on a $4 billion fund generate approximately $80 million per year in guaranteed compensation (2% management fee), enabling managing partners to earn $10-15 million annually independent of fund performance or outcomes, which fundamentally misaligns GP incentives from founder and company success compared to historical models where GPs made only a couple hundred thousand per year and had to rely on carry to get rich.
“that $400 million fund is now 4 billion that guaranteed portion of comp the 2% the management fee you can now get rich on and the reality is is a path exists today in that case roughly 80 million a year that's right for a $4 billion fund right that's right and so they now exist and you could imagine like the managing partner taking quarter of that third of that yes you know so 10 15 million bucks a year yeah right”
Services markets are typically 10x larger than software markets; if AI replaces services (call center agents, insurance form fillers, back-office workers), the capex payoff is enormous because it targets the much larger services market, not just software spend.
“Services markets usually 10x bigger than software markets if we look at how can spend attack it budgets as a whole it's not just let's replace this software spend with this software spend it's let's replace these call center agents let's replace these insurance people filling out forms in the back office it's let's replace a lot of this me most of these firms replace you know employ hundreds and hundreds of thousands of people and we're seeing those numbers”
The timing and sequence of which AI-powered job categories fall to automation will be critical to understanding the actual economic impact of AI: high-error-cost jobs like programming and self-driving cars require very low failure rates (near-zero percent) before they can be fully automated, whereas customer service jobs with historically high error rates (5-10%) may tolerate similar error rates from AI and achieve rapid automation, meaning the economic value of AI will concentrate in categories with high baseline human error rather than precision-critical roles.
“I think it's going to be super interesting to watch which categories and which verticals fall into place first and why there was an article out this week that was looking at I think open AI as a scribe and they highlighted that it failed like a lot and so if you know and and this is also an issue with like self-driving like like if the cost of a 5% error or 10% error is super high it's going to take ai ai a long time to get there I would argue programming is that way you can't have five or 10% selfing self driving car is death it it may turn out that customer service that's just fine because the best in class today already has that error rate just because of human you know failure and so it it will take AI a while to get to where it has the kind of nines you need to solve certain problems and that's why I'm I'm so intrigued by which verticals are going to line up first and get knocked over”
There is nonzero probability that AI capex produces poor returns for some investors or creates supply-demand imbalances leading to diluted returns, or that technology costs decline over time such that faster spending is suboptimal to waiting.
“for sure exact formula for I mean listen there is some nonzero probability just as Elon says there's a nonzero probability that AI is going to end up being bad for Humanity there is a nonzero probability here that at least for some of the players they don't see a return on that investment or or or that you get a supply demand imbalance in the middle I or you get a diluted return right that that you all overspend and had you just waited and spent slower the cost of the technology would come down it would cost you less over time”
Companies that raise excessive capital relative to their ability to deploy it productively spread their focus across too many initiatives rather than concentrating on the three critical priorities needed for success, and this dilution of focus reduces competitive advantage and sustainability despite enabling faster short-term revenue growth, which gets destroyed when revenue slows and the company realizes it lacks competitive moat in the markets where it lacks advantages.
“there are companies that could be sold for 100 $200 million that results in life-changing outcomes for Founders and early employees if you raised $100 million too early that exit path is no longer on the table you can't sell the business for $100 million and make any money because that just Returns the preference stack as the call option investor your call option didn't hit but you get your money back well let's let's drill down let let's split this into two sections because I think that the this thing that's happening um and the reason I retweeted it has implications for both Founders and companies but the asset class also right so we we started down this path let's do it first you know what does it mean for a company to have maybe too much money crammed into it at too high a price yeah there's there's there's many reason there's there's many deep implications of this one thing that I've learned that I hold you know that I will argue with anyone on is every founder and every company and every board will tell you if we raise money we're not going to spend it we're going to stay Frugal that never happens y and instead of focusing on you might ask a Founder what are the three things that matter most this year that are critical for your success if you only focused on those three things you might maximize your chance of success when you have lots of money you're going to do six things at once which means you were diluting the three things that matter”
Over the last decade, data center engineering and construction breakthroughs at Meta and Google were largely taken for granted in the background, but data center engineering is now becoming a primary source of competitive advantage in the AI era.
“for the last decade data center engineering and construction has been in the background right like you know meta had breakthroughs Google had breakthroughs like we' talk about it but it wasn't something that people were posting videos uh on I think there's some Equinox shareholders that might disagree with you but but I I would say we generally took it for granted but now it's clearly becoming a much more important source of competitive advantage”
Only mega-cap companies at the scale of the Magnificent Seven can sustain $50 billion+ annual AI compute spend; for smaller companies, even at a $250B spend scale, the opportunity is inaccessible.
“if you look at Google in the quarter gcp accelerated its revenues by up to plus 35% year on-ear okay so just to put that in perspective they added about two and a half billion of new ARR in the quarter that's like adding a data bricks in the quarter right so it's only companies of that scale that can afford to spend $50 billion a year”
When founders raise too much money at too high a valuation, they eliminate the acquisition outcome path (e.g., selling for $100-200M) because the preference stack makes those exit prices non-viable for founders and early employees, even though such exits would have been life-changing, forcing all stakeholders into a binary outcome (10x or zero) rather than allowing the medium-outcome paths that historically created broad value distribution.
“there are companies that could be sold for 100 $200 million that results in life-changing outcomes for Founders and early employees if you raised $100 million too early that exit path is no longer on the table you can't sell the business for $100 million and make any money because that just Returns the preference stack as the call option investor your call option didn't hit but you get your money back”
The burden of proof is on mega-funds to demonstrate they can pick power law outcomes at a higher hit rate than traditional venture, or they risk merely expanding the 'long tail' of investments, which will dilute overall returns.
“I think that the burden of proof is on the picking of the large funds to say we will pick and hit those power L outcomes but then have a much higher hit rate on the rest and so it won't be this big Power LW outcome it will have the power LW and then be more evenly distributed in the long tail which it can happen it's just really hard especially when your incentives are just to deploy to raise the next fund”
The venture capital industry has transitioned from a 'high margin cottage industry' to an 'institutionalized lower margin industry' over the past 10-15 years, with significant implications for GP compensation models, LP relationships, and founder incentives, representing a fundamental structural shift in how venture capital operates.
“The Venture markets have transitioned from a high margin cottage industry to a institutional ized lower margin industry and that has a lot of implications implications for LPS implications for GPS and implications for Founders”
Brett Taylor's company (Sherpalo Ventures or similar) is showing that customer service can be 100% replaced with AI while improving customer satisfaction, which would validate the massive capex spending thesis if that pattern holds across multiple verticals.
“but if if Sierra Brett Taylor's company is right and you can replace 100 % of your customer service agents with this technology and get a more satisfied customer and if there are multiple verticals like that then you're right this could this could go on for a long long time”
A private equity firm made $1-1.2 billion on a successful exit (Double Click acquisition by Google for $3.1 billion) within a $4 billion fund, which represented a power-law outcome but only returned approximately 25-30% of the fund size, illustrating why even single-hundred-billion-dollar wins cannot get mega-funds to 4-5x returns because the denominator is too large.
“I was at one point in time invited into the office office of this PE firm I didn't know PE very well but they had had back double click and and Google bought it for 3.1 billion and this particular Firm made I forget they made they they had a lot of it so they made like a billion or a billion two and I said to them I we say congratulations man that's incredible he goes well it's in A4 billion doll fund wow you know so you have an incredible outcome right but the denominator is so big right that that you it's just hard to claw it back it's hard to get back to that number which is why you can't can't get to the 45x because you have a you you have a power law outcome and it returns 25% or a third of the fund correct right”
Constraints drive creativity and focus, as illustrated by Steve Jobs' design philosophy of extreme specificity ('I want it this thin') and the principle that companies forced to operate in only 2 markets develop stronger competitive advantages than companies spread across 5 markets because constraints force trade-off decisions that clarify strategy and concentrate effort.
“constraints Drive creativity I think you know I've heard famous stories about Steve Jobs where he felt his job was just to say I want it this thin go figure it out and by saying that he limited the space right we're not going to build a phone and then and then people can focus and be creative um you you've constantly heard you know we're going to make a better decision because we're going to have constraints and we're only going to be in two markets not five”
Early-stage companies showing traction get flooded with investment offers and capital, which can be risky because founders may take excessive personal payouts before product-market fit, destroying the founder's incentive to execute the business plan.
“It leads to companies early companies that show early traction get flooded with offers and investment dollars and we can talk about why over capitalizing and overvaluing your business actually can pose a risk I think you know one of the reasons I wanted to write this post was I thought we would have learned a lot of lessons coming out of 2021 that it's it's very clear we haven't or we just choose to ignore because the incentives are no longer there but for Founders I think understanding and appreciating are you partnering are you kind of marrying a firm that is more focused and caring about maximizing the value of your company together or are you partnering with a firm that just wants to deploy dollars as quickly as possible”
OpenAI put its Foundry (chip manufacturing) ambitions on hold and is instead talking to TSMC about building a chip, suggesting recognition that internal foundry operations would require excessive capital that conflicts with other priorities like model development and training, and indicating some restraint or reevaluation of spending despite reported high burn rates.
“there was a there's something that just came out yesterday that open AI is talking to tsmc about building a chip and has put their Foundry ambitions on hold always thought The Foundry Ambitions were a little too ambitious but but in some ways that's an odd recognition that oh well that may have been too much money and and with the burn rate that's rumored that they have like oh okay maybe you know and so there is there is there is some wrestling”
When AI tools like Copilot are deployed in coding, they provide productivity lift in the 15-30% range according to CIOs, which falls short of the 100% (full replacement) productivity needed to justify single-person headcount reductions, suggesting that the actual measured productivity gains vary significantly by job category and are lower than conceptual maximums in knowledge work like programming.
“when we started you know every oh we're going to replace Engineers but but if you talk to someone that a CIO type you say what kind of lift do you get from a co-pilot product they're kind of generally in the 15 to 30% range which isn't you know if you're replacing somebody that's 100% or infinite percent and so um whether or not and maybe customer service is different than coding in this case where you can you know replace 100% or 90% of the people um but it'll be interesting to see as the analysis come out of these things what those different numbers look like for each of these different verticals”
There was speculation that Google would release 'Jarvis', their version of agentic AI, this week, but Google has already discounted this risk by trading at the lowest multiple in the Mag 7, suggesting markets are pricing in future search disruption.
“a lot of people thought that Google was going to release Jarvis this week which is their version of this someone just funded a new browser company Kleiner did I think open AI will certainly do this agentic you know the fact of the matter is Google is already discounting this it's trading at the lowest multiple uh you know in the mag 5”
AMD is trading down approximately 10% after hours despite 'blistering' revenue growth because the company's guidance suggests it will remain at ~3% of the GPU data center market while Nvidia maintains ~90%, indicating that AMD is not closing the competitive gap despite strong execution, and the opportunity to pierce larger share of the market remains elusive despite Lisa Su's leadership.
“amd's down about about 10% after hours despite you know continuing to have blistering growth and I think what's going on there is there was some talk at the beginning of the year the AMD could capture maybe 10% of the data center market and the fact of the matter is when you they gave some forward guidance tonight that suggest they're going to stay at about 3% of the GPU kind of data center Market relative to Nvidia at about 90% the company's done extraord well Lisa is an amazing you know CEO and I think there are a lot of good things in the works but you know they're not closing the Gap in terms of share of Market”
Never in his 30-40 year career has Bill seen companies take out 10,000+ people and simultaneously see customer satisfaction scores increase, yet under Zuckerberg's 'flatter is faster, leaner is better' approach, companies are reporting that they're achieving more with fewer people, suggesting that AI is enabling a fundamentally different productivity dynamic than historical labor force reductions.
“never have has he seen or have I seen the ability for companies to take out 10,000 people and actually see customer satisfaction scores increase right there's there's always a cost to letting people go but in this case you know you hear Zuckerberg talk about take out 20,000 people flatter is faster leaner is better we're getting more done with less right”
Looking at the 1,400 unicorns that remain private from the pre-mega-round period (roughly 2010-2020), fewer than 10% (possibly less than 5%) can currently raise an up round, meaning over 1,000 private companies are effectively stuck at current valuations and cannot raise capital at higher prices, demonstrating the massive downside risk from raising at too high a valuation in prior years.
“we're baring a number from the cot presentation but they said there were what 1,400 unicorns that are still private from what I like to call the pre-m period yes um what percentage of those could raise an up round right now yeah what is the number I think it's less than 10% it might be less than 5% today so so that means there's over a thousand private unicorns that that are somewhat stuck right now and could not raise an up round”
The Kaufman Foundation published a famous 2011 report finding that billion-dollar-plus venture funds had never achieved good returns, yet after that report, the industry responded by having everyone raise billion-dollar funds anyway, suggesting the lesson was ignored or disbelieved despite empirical evidence.
“in 2011 there was this famous report by The Kaufman Foundation that that that argued that that Billion Dollar Plus funds had never had good returns and of course after that everyone raised billion dollar funds”
Piff (Public Investment Fund) and other sovereign wealth funds are consolidating the number of GPS they work with and are increasingly prioritizing investments within their home countries over global diversification, which may reduce capital available for overseas venture funds.
“they said we're going to deploy more of piff's dollars on investments inside of Saudi Arabia than outside of Saudi Arabia right so I think there was like and I've heard this as well from you know the UAE they're going to consolidate the number of GPS and so I think there is increasing Focus coming out of sovereigns as well I hear the same I was just down in Texas talking to the Texas pensions the same thing happening there”
Venture capital funding deployed globally has declined from a peak of over $700 billion in 2021 to approximately $300 billion currently, representing a return to 2017-2018 levels, but this aggregate number masks more important consolidation trends: the number of first-time funds and second-time funds raising has plummeted, with capital concentrating among a few large platforms like General Catalyst and Lightspeed that are raising $13-14 billion in capital over 24-30 month periods.
“we peaked in key in in Zer in 2021 um at over 700 billion deployed right we've come back down to about 300 billion deployed which is still meaningfully above where we were in 2014 2015 but we're back to about 2017 2018 levels but it it kind of hides a couple things number one the number of firsttime funds raising a second time fund has plummeted the number of firsttime funds have plummeted and so really what's happening is you have the consolidation among a few big platforms um you know just this week we saw that General Catalyst has raised 8 billion in New Capital now of course you know I love my friends at General Catalyst but you know just just 24 months ago they raised 4 a half billion so that's 13 billion raised in that 24mon period similarly Right light speed recently annc that they rais 7 billion after raising about 65 billion so another 13 or 14 billion over kind of that 28 30-month period of time”
TSMC announced a broad 20% price increase across its foundry business, which Brad anticipated would occur and viewed as overdue evidence that semiconductor supply is tight and capacity constraints are real, indicating that mega-capex spending is running up against material supply constraints and cost escalation.
“tsmc said they're going to do across the board 20% price increase and that's I was like this should be happening and so seeing that also indicative of this right”
xAI announced expansion of the Colossus cluster in Memphis from 100,000 GPUs to 200,000 GPUs, and other mega-cap companies (Google, Meta, Microsoft, Amazon) are reporting capex at a ~$250 billion annual run rate, with all companies discussing how to secure facilities capable of powering 1-3 gigawatt clusters, indicating that the prior question of 'will we need 200,000 GPU clusters' is now settled and the focus is on facility capacity and power infrastructure.
“we saw out of xai the Colossus cluster in Memphis which I discussed at length with Jensen they announced this week that they're scaling El said we're now going to go from 100,000 gpus to 200,000 gpus in that uh facility it's expected that capex will continue to you know continue on at this $250 billion run rate Google came in tonight at $1 13 billion just above the 12.7 billion that was expected so they revised that high higher uh a bit Bill what do you just make out out of you know as we sit here toward the end of 2024 um do you expect that this is going to you know uh continue at this pace”
The searches most at-risk of disruption from generative AI are 'encyclopedic' queries (factual knowledge lookups) rather than high-value transactional searches like travel, real estate, and automotive purchases, and Google's monetization is primarily derived from the high-value transactional searches, so the long-tail of low-monetization encyclopedic searches being disrupted may not materially impact Google's revenue.
“the searches that were I guess stolen most quickly by a chat engine or more of these kind of encyclopedic type searches where you're looking for information whereas the big money searches for Google or travel and new car and a home and those kind of things big purchases and so those may not in fact be the ones where they're inserting much of this and so so you may be you may be talking about the long tail of their searches anyway that aren't that relevant to the big dollars”
Companies are scaling faster and achieving larger outcomes (in terms of revenue and market cap) faster than in previous decades (e.g., OpenAI reached $4.5 billion in annual revenue much faster than Google and Meta did), which mathematically justifies why total VC deployment and fund sizes should be larger today even if the rate of increase seems excessive, but there is still a cap on the fund size that can deploy for traditional venture-like returns.
“there is no doubt that the outcomes that we're now talking about are much bigger than the outcomes we were talking about 10 or 20 years ago so it would make sense to me if you looked at a trend line for the industry the industry and the number of dollars deployed and the size of funds should be much bigger today because companies are scaling much faster today and the outcomes are much faster open AI got to you know $4.5 billion dollars of Revenue in a fraction of the time it took Google and meta and you know and so like that to me supports more dollars larger funds but I do think the law of economic gravity prevails there's a certain fund size I think that can deploy if you're aiming for traditional Venture likee return then I don't think there are 100 growth companies right that you can go put in a fund of 5 or 10 billion equally weight them and get a four to 5x over any reasonable period of time”
Google's search revenue growth has accelerated from 2-5% (early 2023) to 12-14% currently, and they reported that AI-enhanced search results (SGE - Search Generative Experience) monetize at the same level as non-AI-enhanced results, suggesting that Google can embed AI answers at the top of search results while maintaining ad revenue per search, contrary to concerns that generative answers would cannibalize clicks and revenue.
“if you look at this chart that shows the search Revenue growth over the course of last eight quarters I mean they've accelerated search Revenue growth um from 20 you know end of 22 and and early 23 from like you know 2% 5% % all the way up to 14% still at 12% today so part of the question is how long can search Revenue growth stay at this level and one of the key questions there was whether or not the AI enhanced search results were going to monetize at the same level as non AI enhanced search results so they call these sge um and it turns out that uh they they they said on the call tonight that they're monetizing at the same level”
Anthropic released a 'computer use' capability allowing Claude to take control of a user's browser and complete tasks like booking hotels or airline tickets, and Google reportedly was expected to release a similar capability called 'Jarvis' this week, indicating that agentic AI (AI that can autonomously perform multi-step tasks) is the next frontier of AI capability deployment.
“we've also as you know we saw this week that anthropic released computer use right to take control of my browser to you know what we've been discussing book that hotel or book that airline ticket it's cloy uh you know a lot of people you know but it but it's a step in the direction a lot of people thought that Google was going to release Jarvis this week which is their version of this”
Google faces a revenue-per-visit decline if it completes transactions directly (through generative search answers to travel booking queries like 'book a hotel'), because hotels and merchants currently spend 50-100% of first-transaction value on marketing to Google, but if Google or ChatGPT completes the transaction directly, merchants would only be willing to pay ~5% of transaction value to Google, creating a fundamental economics problem that blocks Google from fully automating travel/commerce transactions despite having the technical capability.
“well they have this other challenge that I that I think is hard for people to understand but I've explained it a couple times which is their revenue per visit is going to fall if they complete the transaction because someone is spending marketing dollars to take that customer to their website and run the transaction with the hopes that they'll come directly back to that website so they'll spend 50 to 100% of first transaction there's no one that's going to if if if Google or chat gbt or whoever convinces kayak or whoever to be a white label rails in the background the amount of money they're going to give you for that going to be like five% of the transaction”
Fred Wilson's rule of thumb (attributed to venture returns) suggests approximately one-third of companies fail, one-third underperform, and one-third deliver the 5-10x returns that actually drive fund returns, while Dreon's analysis shows that even more granularly, approximately 10% of companies produce the returns that get the fund to its target multiples (e.g., 10x fund return).
“I love you know Fred Wilson's rule of thumb you know a third of companies fail a third basically underperform and then a third get you your 5 to 10x returns and dreon breaks it down even more which you know says all comes down to the 10% that produce you know kind of that that that 10x fund”
Humanoid robots like Tesla's Optimus are preferred over non-humanoid robotic forms because humanoid robots work in environments (the physical world) that were already designed for human bodies, requiring no infrastructure redesign, whereas agentic AI using 'computer use' capability succeeds because it operates on existing web infrastructure designed for human interaction, providing similar 'use-as-is' advantage without requiring system redesign.
“I think the way you go over the top um and why I'm so comp so excited about you know the equivalent of computer use across the board is you know in the same reason perhaps that Elon says you use a humanoid versus a non-humanoid robotic form because a humanoid Works in a world that was designed for humans right and the thing about computer use you don't have to build all these new rails and everything which takes a really long time to negotiate all of that stuff instead you just give it ubiquitous use of the world that already exists um you know uh straight out of the gate so I think this this is something I really wanted to make a point on”
Mega-fund investors function as 'free rolls'—GPs get paid regardless of outcomes, treating the carry (20% of profits) as a call option, so they win if any one of their portfolio companies becomes a 10x+ power law outcome, but don't lose anything if the other investments fail.
“and by the way I mean rather than say they only care about the management fee they May view the other one there's a phrase in poker called a free roll they may view it as a free roll I'm getting rich no matter a loty ticket it's exactly call option but this other it's a call option right and if this company happens to be you know the next Google the next meta we've seen how those compound over 20 years and it'll work out right it'll work out in that or maybe it'll work out for one of the 10 that I'm putting 400 million in and then I'll be okay right”
Even if meta-cap companies are concerned about capex guidance and spending levels, the competitive environment creates a Pascal's Wager dynamic: if AI upside is potentially infinite and all competitors are investing heavily, no company can afford to pull back capital investment, even if returns are uncertain, because falling behind in AI capability would be catastrophic to their competitive position.
“Pascal's his French philosopher he says I don't know whether God exists or not but if he does exist and I don't believe then it's going to be a really bad outcome for me and if he does exist I do believe I have eternal life so if you believe that the upside of AI right is infinite is eternal right is some really big outcome and all your competitors are doing it what else are you going to invest your money in you have we said we said here several weeks ago you can buy back your shares you can issue a dividend or you can buy gpus right do you think any one of these Founders or leaders you can do license deals with companies you want to acquire you can you know so I think but I think most of these people are in this game for this moment but I think it it it it it raises a really interesting alternative question which is it's pretty clear to me now that there are very few new entrance that are going to be able to play in that game”
The transition to mega-funds and overcapitalization emerged from a combination of multiple factors (near-zero interest rates from 2008-2020, the rise of AI as a perceived infinite-upside opportunity, and the globalization of fundraising bringing in sovereign wealth and pension funds) rather than any single willful decision, suggesting the system is emergent and actors are stuck in a game they cannot easily escape.
“I think it emerged out of a combination of what was happening with interest rates in Zer and then the ra rise of AI and maybe the globalization of of fundraising that brought Sovereign wealth in and it all kind of combined and created this this mix that we're living with and for the most part I think we're all actors stuck in the game right you know I'm I'm not sure you can escape it”
The election is likely 70-80% priced into markets as a Trump victory, though not 100%, meaning there is 20-30% downside scenario if Harris wins, in which case the trades that have been unwinding based on Trump probability would sharply reverse (democratic spread trade is down 20-30% in 30 days), creating significant portfolio implications for investors betting on Trump policies.
“you know there's 70 to 80% baked into the cake you can see that a lot of stocks have moved a lot and I peeled this back a little bit this is unbelievable I'd love to see what's in this Democratic spread trade bucket it's it's down 20% 30 days”
Nvidia Jenson Huang estimates that 9 trillion dollars of cumulative capex on 200 million GPUs is 'reasonable' and possibly even 'too small,' and the NPV math suggests that if AI achieves 'artificial superintelligence' and replaces 5% of the global workforce (worth ~$9 trillion annually in wages), then 9 trillion in cumulative capex over seven years would be NPV positive and justifiable as a single-entity investment.
“Masa at fii today said 9 trillion of cumulative capex on 200 million gpus is very reasonable he said in fact I think it may be too small and he went on to say you know that if you took the most um critical estimate so of the value of AI so he said the lowest estimate that he's seen on the value of AI is that it could do the the you know what he calls artificial superintelligence which is what he thinks we'll have if we spend n trillion is that it would replace 5% of the global Workforce and it just so happens if if you do the math on that 5% of the global Workforce cost about $9 trillion a year so he said imagine you were a single entity single Enterprise the global Workforce was your Workforce and you could spend 9 trillion cumulatively over the next seven years that you would then recoup in a single year through the efficiencies gain in that Workforce it would obviously be an npv positive investment”
Endowments historically provided capital discipline to venture markets through legendary allocators like Phil Rotnei and others who would directly call Silicon Valley VCs and tell them 'you're getting over your skis, back it off,' but this disciplinary ecosystem has disappeared, replaced by pensions and sovereign wealth funds that lack the same willingness or expertise to push back on mega-fund sizes.
“I do think that pensions and Sovereign wealth funds have you know stepped in to replace some of the money that came out of the endowment e ecosystem right because endowments you know I think about the legends of of that business who were allocating those dollars you know the Phil rotner of the world or or the Swinson of the world right like they would phone up everybody on Sano Road and they would say you're getting over your skis you need to back it off a little bit right that industry is long gone”
A new browser company called Browserbase and conceptually similar approaches are exploring whether to build a new browser designed for agents (with composable APIs instead of buttons) rather than humans, representing an alternative architecture to the 'computer use' approach of scraping human-designed websites and clicking buttons, but this would require rebuilding all underlying transaction infrastructure (like Stripe) to work with agents.
“this new browser right it's company called browser base and I think this is a really interesting concept to think about kind of transitionary States versus end States right the transitionary state today is we're going to have agents that essentially can scrape websites and view it as if I'm doing it they're going to recognize the button and click CLI the button they're going to go here what if we reinvented the browser and built it in a way where it's not a button for a human eye but everything is some composable API that an agent can recognize and an agent can go interact with and so you're not scraping the website where you would lose the eyeballs you're building a new browser for agents versus Humans is that the end State there's lots I don't know the answer but it's a it could you have to rebuild all the rails too I mean everything has to start to look like strip”
Concerns about mega-cap capex were overblown because operating income and margin expansion have outpaced capex increases in recent periods; unless a recession occurs with declining earnings while capex remains elevated, high capex is sustainable with rising topline growth and operating leverage.
“I I I think that you know there is I mean remember these capex um expenses have gone fairly parabolic in the last couple years but operating income so margin expansion and Topline growth have outpaced the increase in capex so people have tolerated it what won't work well right let's say we had a recession next year and that you know earnings were came in a lot lower than people expected or Topline growth came in lower but that there was uh you were unhinged on capex I don't think that would be a particularly good equation for the stocks”
Brad uses ChatGPT rather than Google to help his son Lincoln study for AP History, as it's a better knowledge-discovery tool, but this use case isn't monetized by Google anyway (knowledge graph searches), so it doesn't cannibalize Google's high-value searches.
“I mean a lot of people think of these as the knowledge graph searches um you know I've been using you know all last night I was using chat GPT to study help help Lincoln study for his AP History exam there's not a lot of monetization of that in Google anyway that page could be the same right”