Anthropic S-1 Shocker, Greatest Business Ever? Google's $500B Wipeout, Midterm Predictions

Anthropic

Across the shows: 74 speakers in other episodes, 83 bullish and 63 bearish calls, past 30 days

• David Sacks · Craft VenturesBullish

  Anthropic's IPO could be the biggest venture outcome ever, the fastest value creation in venture history.

  Sacks argues the Anthropic IPO will likely be the biggest venture outcome of all time, given the company is only about six years old. He says the exit will rival SpaceX's IPO, which took more than twenty years to reach, making it the fastest value creation of any venture-backed company. He also points to the wealth it will create for employees, with hundreds of employees making hundreds of millions and thousands making $10M+.

  “It could very well be. I think this is probably gonna be the biggest venture outcome of all time, especially given the time in which this was created.”at 5:52

• David Sacks · Craft VenturesBullish

  Anthropic's reported $80B loss is a non-cash accounting artifact; the company is very capital efficient.

  Sacks says the $80B loss was a phantom number. It came from convertible notes held by infrastructure partners such as Amazon. As Anthropic's valuation rose, the notes were recorded as a loss that flowed through the income statement, but no cash was lost. Had Anthropic issued equity instead, the loss would not exist. The S-1 shows Anthropic has raised about $110B, spent only about $9-10B, and still holds about $100B in cash, making it an extraordinarily capital-efficient business.

  “That narrative basically just got completely refuted because it turns out that that $80,000,000,000 loss is a phantom number.”at 7:00

• Jason Calacanis · LAUNCHBullish

  Anthropic is only about 1% penetrated, and AI is creating large value across the stack.

  Calacanis estimates that about 1 million businesses globally could spend $100K or more on Anthropic while that spend remains under 1% of their total spend. Anthropic has about 11,900 such customers, so it has only about 1% penetration of its potential customer base, and far less of the spend potential. He concludes that everyone in the AI value chain is profitable, which signals real productivity gains, and that the addressable market is effectively all of global commerce.

  “There's so much value being created all the way through to the customers, and we're only 1% or Anthropic is only 1% penetrated on their customer potential.”at 14:26

• Jason Calacanis · LAUNCHBullish

  Anthropic's high margins give it room to cut token prices for cheap tasks and charge premiums for advanced ones.

  Calacanis argues that Anthropic's high incremental gross margin gives it good pricing elasticity. It could push prices down to compete for cheaper, easier tasks, while charging higher prices for premium tasks to stratify the value of intelligence. He expects the whole industry to move toward this tiered pricing.

  “I think what Claude is showing is that they have such a high margin. They actually have good elasticity on pricing. So they could push price down, and you could start to see them start to segment the market a little bit.”at 15:48

• Chamath Palihapitiya · Social CapitalBullish

  Compute, mainly data centers and power, is Anthropic's biggest risk, so it is the binding constraint.

  Chamath ran a word-share analysis of the S-1 risk factors. Compute is mentioned more than any other risk, with data centers, power, chips and the grid as the main components. Inside compute, data centers are the biggest concern, followed by financing commitments and power. He reads this as a company that understands its binding constraint and is on the 'forward foot.'

  “the share of words in the risk factors that are allocated to different risks are very important, and what this tells you is the single biggest risk to the ant business is compute. They mention it more than anything else.”at 19:09

• David Sacks · Craft VenturesUncertain

  Anthropic is racing toward full recursive self-improvement faster than any company, despite its safety rhetoric.

  Sacks cites reporting and NGO commentary arguing that, once Anthropic's rhetoric is separated from its actions, it leads the charge on recursive self-improvement. He points to an August Anthropic blog showing 26% of model-development tasks already used RSI, and says its cost cuts come from using its own models to find algorithmic and compute efficiencies. He reads Dario's pacing letter as an alibi, since Anthropic cannot keep its frontier lead without faster RSI. He says he does not fully believe the doomer RSI story, but thinks a doomer-Anthropic split is possible.

  “There's no single company in the world that is doing more to race towards full RSI than Anthropic.”at 43:53

• Chamath Palihapitiya · Social CapitalNeutral

  At a $1.04-1.06 IPO valuation, Anthropic looks fairly priced, so there is no obvious deal.

  Chamath argues the proposed IPO price range is about fair value, so buying at the IPO does not offer a mispricing. The IPO price is the entry point into the future, so an investor must have a view on Anthropic's future risks. He frames those risks as the enterprise ROI ramp and the broader market risk from AI borrowing.

  “And I think at the $1.04, $1.05, $1.06, it's kind of priced fairly, so there's no deal.”at 46:50

• Chamath Palihapitiya · Social CapitalUncertain

  Anthropic's enterprise customer ramp is the key checkable risk, measured quarter by quarter.

  Chamath says the first risk is how long it takes for Anthropic's roughly 10,000 customers to move from under $1 million of spend to single-digit millions and then tens of millions. Growth depends on defensible use cases and white papers that let executive teams sign off on multimillion-dollar AI spend. He says this risk can be priced and tracked quarter after quarter as the numbers swell, while not disputing revenue traction or business quality.

  “The first risk was what I showed you on their revenue composition, which is how long does it take for Ant to show those 10,000 customers the ROI math that they will need to move from less than a million to single digit millions to tens of millions.”at 47:06

• Jason Calacanis · LAUNCHBullish

  Anthropic could out-earn the top three historical franchises combined within its first seven years.

  Calacanis compares Anthropic's revenue path with the cumulative first-seven-year revenue of iPhone, Google search cost-per-click and Amazon retail, excluding AWS. He says Anthropic is at about $100 billion this year and, if it hits its targets, will reach $300 billion and eventually $600 billion. He argues it is possible the business ends up bigger than those three franchises together, which he says has never been seen before.

  “If Anthropic hits the targets they've set out, you know, ballpark here is getting to a 100 and then 300,000,000,000 in revenue, eventually 600,000,000,000 in revenue.”at 51:38

• Jason Calacanis · LAUNCHBullish

  Open models are not slowing Anthropic yet; enterprise demand is big enough to absorb lower prices.

  Calacanis says that in light of Anthropic's numbers, the enterprise opportunity is large enough that it is not being dramatically impacted by open alternatives. Startups and enterprises are anecdotally exploring open models, but not enough to slow Anthropic. He adds that enterprises are satisfied with cheaper second and third tier models, so Anthropic's price cuts widen usage rather than cannibalize it.

  “you would have to say that they are lowering the prices and the opportunity in the enterprise is so great that they're not being dramatically impacted by it.”at 52:56

• David Sacks · Craft VenturesBullish

  Anthropic's token price cuts raise margins, driven by its own algorithmic efficiency gains.

  Sacks says Anthropic is offering the same capability at about one-tenth the price, but this is not margin sacrifice. Its algorithmic improvements, increasingly generated with its own models, lower its cost to serve, so it is gaining margin. He says a Q3 slowdown partly reflected a deliberate crackdown on distillation that cost revenue, and that demand reaccelerated after the recent token cost cuts.

  “It's not just because they're giving up pricing power. Right? It's not that they're losing margin. They're actually gaining margin.”at 55:19

• David Sacks · Craft VenturesBearish

  Open-source models are Anthropic's biggest risk, commoditizing intelligence at low prices in coding.

  Sacks says the biggest risk to Anthropic is open-source models saturating intelligence at a low price point, commoditizing coding and other large chunks of the market. He adds a second-order risk: to stay ahead of open models, Anthropic must keep using recursive self-improvement, which it calls risky. He wonders how sustainable that is and whether open models catch up if RSI is slowed.

  “I think the biggest risk factor to Anthropic, I do think it is some version of the open source models, maybe not catching them, but sort of saturating intelligence at a low price point”at 57:15

• David Sacks · Craft VenturesBearish

  Anthropic is seeking regulatory capture and, as a dominant player, should be treated with suspicion.

  Sacks says Anthropic prioritized government affairs from its startup days, hired former Biden and later Trump administration officials, and has funded PACs, nonprofits and NGOs. He argues its growing capital will buy more Washington influence. He calls it the most powerful monopoly or duopoly in history and says that as a public big-tech company it should be treated with suspicion, not as an innocuous startup. He says it is unseemly for a company so dominant to seek regulatory capture.

  “Second, they're going for regulatory capture, and we should definitely not allow that to happen, especially in light of number one.”at 1:00:16

Alphabet

Across the shows: 38 speakers in other episodes, 29 bullish and 18 bearish calls, past 30 days

• Jason Calacanis · LAUNCHBearish

  Google's core search ad franchise faces AI disruption and decelerating growth; a real challenge.

  Calacanis notes Google missed on search revenue for two straight quarters even though searches are growing. Search revenue growth slowed from 17% to 12%. AI assistants like ChatGPT and Anthropic answer questions directly or solve agent tasks, which threatens cost-per-click ad clicks. Part of the recent growth may come from an ad pricing change, but advertisers will eventually notice they are overpaying and cut spend. The key question is whether a consumer AI product can ramp revenue fast enough to replace search.

  “I think the core services, the core Google ad revenue is a real challenge right now in terms of can they keep the momentum up.”at 1:03:32

• Jason Calacanis · LAUNCHBullish

  Google Cloud is tremendous: operating income jumped to over $11B, and owning models lifts margins.

  Calacanis says Google Cloud was not making money not long ago. Operating income was about $3.5 billion a year earlier and is now over $11 billion, and cloud revenue doubled year over year. He compares it to the SpaceX AI story, where the value of built compute infrastructure is huge. Because Google also owns the models on top of the compute, it can capture a higher margin on that part of the stack, so he sees several years of progressive growth.

  “I think the Google Cloud business is looking, you know, really tremendous... Last year, the Google Cloud business for this quarter was generating 3,500,000,000 of operating income. This quarter, it's over 11,000,000,000 of operating income.”at 1:03:58

• Jason Calacanis · LAUNCHBearish

  Google's model builders may be losing resources to compute rental, fuelling internal turmoil over AI strategy.

  Calacanis points to Jeff Dean and others leaving Google as evidence of an internal battle. His theory is that the scientists building models are not getting the resources they wanted, because Google can make more money renting compute to others. He says Google lacks a clear answer on what the company is, whether search, AI or cloud, and that this did not come across well in the earnings report. He also says Alphabet was a way to let bets like Waymo fend for themselves. He hedges that he does not know this for sure.

  “I don't know this for sure, but the theory is that the computer scientists that have been building models at Google maybe aren't getting the resources that they had hoped to get because Google can make so much more money renting that compute out to others.”at 1:06:17

• Chamath Palihapitiya · Social CapitalUncertain

  If Google dropped Vickrey auction pricing it is over-earning on ads, a risk once markets notice.

  Chamath discusses a widely circulated post by an advertiser claiming Google changed its auction model away from a Vickrey (second-price) auction. Under a Vickrey auction, the winner pays the underbidder's price, which is good price discovery. If Google has abandoned it, advertisers are overpaying, which is what Friedberg suggested. Chamath says the claims have circulated since July, that he does not know if they are true, and that Google IR has not nipped them in the bud. If the market smells this out, investors will wonder what is happening under the hood.

  “But if now what they're doing is they've abandoned the Vickrey auction, [Friedberg's] right, you're over earning.”at 1:08:51

• Chamath Palihapitiya · Social CapitalUncertain

  Search's falling revenue share could look like robbing future earnings, and the market may react badly.

  Chamath notes that Search fell below 50% of Alphabet's revenue. If the narrative builds that this is 'robbing Peter to pay Paul', pulling money from the future into the present to smooth earnings, he expects the market to react badly. He says Google IR should refute it before analysts settle on that view. He also expects that once ad revenue falls below a threshold, Alphabet looks more like a structured conglomerate with YouTube, GCP, Other Bets and Search blocks. He sees that as better: more diversified and less levered to any one business. But he warns that switching the shareholder base from search or ad bettors to a conglomerate buyer pool is complicated and may leave some holders out.

  “I think if this momentum builds, this idea that this franchise is essentially robbing Peter to pay Paul, like pulling money from the future into the present to kind of smooth out earnings, the market will react badly to it.”at 1:09:39

• David Sacks · Craft VenturesBullish

  Google search still growing about 12%; Alphabet's stock is oversold on the falling-share headline.

  Sacks argues the headline that Search fell below 50% of revenue misreads the numbers. Search revenue rose about 12% year over year, from $56B to $63B, a roughly $7B increase, so the business is not deteriorating. Search share fell because other Alphabet businesses grew faster, mainly Google Cloud, which roughly doubled year over year to $30.7B. He sees Alphabet's diversification as a good thing, with YouTube and Workspace subscriptions growing. Search is growing slower than other segments but is not in decline, so he thinks the stock sold off too hard on the news.

  “So in any event, I tend to think Google's a little bit oversold on this news today.”at 1:13:16

• Jason Calacanis · LAUNCHBullish

  Alphabet is a long: search volume still rising and other businesses now half of revenue.

  Calacanis says search query counts are still rising, so the franchise is still growing. The concern is revenue per search, which he calls the issue people focus on. He agrees with Sacks that Alphabet's other businesses are now only about 50% dependent on search, and that this diversification will continue. He points to Alphabet's stakes in SpaceX, Anthropic and Waymo as additional upside. On that basis he says he is long and bought more Alphabet shares during the show.

  “So I'm long. I just bought more shares of Google while doing the show.”at 1:13:51

NVIDIA

Across the shows: 49 speakers in other episodes, 39 bullish and 16 bearish calls, past 30 days

• Jason Calacanis · LAUNCHBullish

  Nvidia earns about 75% gross margin on chips, a large share of AI's value chain profit.

  Calacanis maps the AI value chain. Nvidia spends $1 on cost of goods and sells the chips for about $4, a 75% gross margin. SpaceX then earns an 83% margin on that chip spend, depreciated over five years, and Anthropic earns about 85% incremental gross margin on each new dollar. A dollar of Nvidia cost becomes about $18 of Anthropic revenue, showing that every layer of the stack is profitable.

  “start with NVIDIA. How much does NVIDIA spend to make the chips, and then how much do they sell the chips for? So for every dollar of COGS, every dollar that NVIDIA spends, that turns into $4 of revenue for NVIDIA, 75% margin.”at 12:49

Microsoft

Across the shows: 33 speakers in other episodes, 31 bullish and 4 bearish calls, past 30 days

• Chamath Palihapitiya · Social CapitalBullish

  Microsoft's revenue nearly doubled with flat headcount, showing AI productivity gains, not job loss.

  Chamath cites Microsoft's earnings data. Revenue rose from $198B to $332B in four years, while staff increased by only 2,000. Before AI, he says, nearly doubling revenue at that scale would have required roughly doubling headcount. He sees the headcount outcome as productivity gains at the same level of human input, not job loss. He also notes that Microsoft's voluntary buyout offer could cut its workforce further.

  “this idea that it creates job loss, don't think it's true. It's more that you're seeing productivity gains at the same number of human input, and I think that's interesting.”at 17:50

US growth & recession risk

• David Sacks · Craft Ventures

  AI capex at about 3% of GDP is driving essentially all US economic growth right now.

  Sacks cites Q3 GDP growth of 3.7%, up from 2.5% in the first half of the year. He notes that AI data center build-out capex is roughly 3% of GDP. From this he concludes that all of US economic growth is coming from the superintelligence build-out.

  “And you could also look at all the CapEx that's being spent right now on this build out and it's roughly 3% of GDP. So you could say that all of the economic growth in The United States right now is coming from this superintelligence build out.”at 24:47

• Chamath Palihapitiya · Social Capital

  The US economy is levered to about seven or eight AI companies, which set the tone for the whole country.

  Chamath argues that a handful of the most important companies are in a massive spending binge, after having been highly cash-generative. Their spending drives GDP and growth. He ties this to politics: slowing power and data center build-out would be a way for Democrats to produce a recession going into 2028, while Republicans must get power and electricity right to win.

  “The entire United States economy is levered to the AI complex. So very narrowly, we're talking about seven or eight companies that sets the tone for the entire country.”at 32:46

• David Sacks · Craft Ventures

  The US economy is not a disaster; growth is strong and real household incomes are rising.

  Sacks rejects the claim that the economy is in trouble. Real GDP grew 2.5% last quarter and is tracking at 3.7%. Median household income is almost $90,000, an all-time high. Poverty is at its lowest level, and the share below the poverty line is around 12%. He argues this is not a K-shaped recovery: the average household is seeing real wage growth, which he says did not happen during the Biden years.

  “The economy is not a disaster. That is total fake news. We know what the growth rates are. It was two and a half percent in the last quarter, and it's tracking at 3.7% right now.”at 1:17:19

Enterprise AI adoption & AI software

• Chamath Palihapitiya · Social Capital

  Anthropic's enterprise revenue could reach hundreds of billions within three to seven years.

  Chamath says Anthropic has barely scratched the surface of the commercial working world. He argues it does not need to capture 100% of the addressable market, only 25-30%, with a reasonable distribution of negative churn, to become a very large business. He sees roughly 1,900 companies that can already show ROI for large deals, though he calls the ROI case still speculative. The main constraint is budgets: once companies move from million-dollar to multimillion-dollar deals, buying becomes more thoughtful and cross-functional.

  “And what you start to see is that this can be a very large revenue business in the hundreds of billions of dollars in sort of this three to seven year timeframe.”at 27:59

• Jason Calacanis · LAUNCH

  Only 1-2% of corporate workers use AI tools, so usage is about to rise sharply.

  Calacanis argues that only about 1-2% of people in corporate America are using these AI tools today. He points to OpenAI's and Anthropic's recent product launches as still-small use cases. He concludes that observers are greatly underestimating the coming uptick, given the 100-200 million white-collar workers who could adopt them.

  “the fact that only one or 2% of people in corporate America are even using these tools. We are greatly underestimating the uptick here.”at 34:59

• Jason Calacanis · LAUNCH

  Enterprise AI use is nascent and could jump from about 10% daily to 60-80% of workers.

  Gallup data show only 10-20% of workers use AI daily or frequently, and large enterprises restrict employees out of cybersecurity fears. Once those fears are mitigated and agents can connect to email, knowledge bases and Slack, Calacanis expects daily use to rise to 60-80%. He points to Uber buying ChatGPT for every employee as the next wave. He also argues only 1-2% of corporate America uses these tools today, so the uptick is being underestimated.

  “this is gonna go from 10% of people being daily users to 60, 70, 80%. There's so much growth left here when every single employee has an agent.”at 59:33

AI infrastructure & data centers

• David Sacks · Craft Ventures

  AI firms' huge capex is justified because these are smart companies that expect returns.

  Sacks responds to the view that AI companies' massive debt raising and capex is a warning sign and is pulling down Treasuries and stocks. He sees it the opposite way: these are smart companies, and they would not make such investments unless they expected a return. He also argues that Anthropic's $100B-plus cash and roughly $80B debt facility mean it can back its compute commitments through 2030.

  “I don't think they'd be making those kinds of investments unless they thought there was ROI there.”at 34:46

• Jason Calacanis · LAUNCH

  Compute demand will far outstrip supply because the industry's buildout is badly underestimated.

  Calacanis argues that corporate AI use will explode as white-collar workers run multiple agents, similar to the Office 365 rollout. He says utilization could rise about 500x, with only 1-2% of corporate America using these tools today, and that token usage will grow far faster than expected. He concludes there will not be enough compute and that the industry has the buildout completely wrong, which people will look back on as a mistake.

  “I think the utilization of these things is going 500 x... That means we're gonna have... Not enough compute, and everybody's got the build out completely wrong. There won't be enough.”at 35:20

• Chamath Palihapitiya · Social Capital

  US data center buildout is clogged by approvals, utility staffing and equipment shortages.

  Chamath argues the constraint on AI buildout is not demand or willingness but physical and regulatory bottlenecks. Permitting and approvals are slow, utilities lack the staff to connect new load, and the capital equipment needed is not available. As a result, he says the country is clogged and cannot build out even if it wanted to.

  “approvals and the ability for utilities to find the staff and the ability to actually turn these things on and then the ability to get the capital equipment you need, it just isn't there right now. So we as a country are sort of clogged up.”at 35:52

US elections & politics

• Chamath Palihapitiya · Social Capital

  Republicans likely keep the Senate and Democrats win a narrow House; Democrats may use subpoenas to create FUD.

  Chamath still thinks the midterms are set up for Republicans to keep the Senate and Democrats to win a narrow House majority, despite a viral clip predicting a bloodbath. He argues the Democrats' mandate then becomes creating FUD (fear, uncertainty and doubt). Their tool would be subpoenas and document releases aimed at going into 2028. He also argues an economic slowdown benefits Democrats heading into the 2028 cycle, so they would try to influence power permitting and data centers, which he calls the levered bet.

  “I think it's still set up to have the Republicans keep the Senate and the Democrats to have a narrow majority.”at 1:15:07

• David Sacks · Craft Ventures

  Midterm losses for the president's party are normal; polls are biased and the media is overhyping a bloodbath.

  Sacks argues the out-of-power party usually does well in midterms, citing Clinton's 1994 loss and Obama's 'shellacking', so he expects a more normal result than the media's tsunami framing. He says Republicans should not expect a miraculous win. He also says polls are unreliable. Some polling firms have admitted their polls were fake, a study of the last three elections found about three-quarters of polls favoured Democrats, and in 2024 the Selzer Iowa poll badly missed.

  “it's quite customary for the out of power party to do very well in the midterms. I'm just saying that the media is already portraying this as a tsunami ... and I think it might be a more normal midterm result.”at 1:19:16

• Jason Calacanis · LAUNCH

  Democrats are very likely to sweep both chambers in the midterms.

  Calacanis trusts Polymarket, which gives Democrats a 68% chance of winning the Senate and 92% for the House. He backs this with Trump's net approval rating, which is at an all-time low. He expects Democratic-leaning voters to turn out against Trump, while moderates who were open to Trump in his second term are not showing up. He also says Republican voters are upset because inflation and the wars were not solved, so he sees 'a very easy sweep' for Democrats.

  “I think polymarket is gonna get it right. I think Democrats are gonna win the senate, 68% chance right now on polymarket.”at 1:19:45

AI pricing & unit economics

• David Sacks · Craft Ventures

  Anthropic's token costs fall roughly 10x a year while its gross margins rise, not just cut into profit.

  Sacks says Anthropic's gross margin is now around 85%. For a given task, cost-effectiveness improves roughly 10x year over year, so the company can cut the price of a task by about half. The customer captures part of the savings, but Anthropic captures a larger share through margin expansion, because its models keep getting more efficient ('compute multipliers' such as algorithmic gains, better data curation and MoE architectures). Sacks reads this as a refutation of the view that the business is commoditizing toward zero gross margin or depends on venture subsidies.

  “they're cutting the cost of their tokens significantly, but their margins are going up even faster because they're able to produce these models that are more and more efficient.”at 8:56

• David Sacks · Craft Ventures

  Falling token prices expand total AI demand, since agent and cyber uses consume far more tokens.

  Sacks argues Jevons paradox plays out at every level: as token costs fall and more use cases open, enterprises use more AI and total demand rises. New applications such as agents and continuous cyber monitoring consume far more tokens than prompting or annual pen tests. He concludes prices will keep falling while the market size keeps growing.

  “the total demand just keeps going up. And the total revenue and market size keeps growing.”at 56:29

AI regulation & safety

• David Sacks · Craft Ventures

  Democratic proposals to ban superintelligence or require federal model approval would be draconian.

  Sacks lays out what he sees as the Democratic AI agenda. It is either an outright ban, as Sanders, Warren and AOC are said to want, with prison terms for developers and corporate 'death penalty' for companies that build superintelligence, or a draconian federal review regime in which a federal department would approve every model before release, which he calls a 'DMV for models.' He thinks the party is moving toward that agenda and would use it to sabotage the Trump economy.

  “Either to ban it outright as Bernie Sanders and Elizabeth Warren and AOC want to do. They basically want twenty year jail sentences for developers of superintelligence.”at 23:02

• Jason Calacanis · LAUNCH

  Governments will regulate and allocate compute, and may try to stop data centers to curb AI's power.

  Calacanis argues AI companies are attracting more capital than governments, and that this creates a conflict. Governments on both left and right will try to subsume or control AI and will pass laws that assign compute capacity and decide who gets how much. He expects this to play out in the near term through capital, with governments trying to stop data centers.

  “I think that you're gonna end up seeing compute get regulated in the future where compute capacity is going to be assigned by government, and they're gonna pass laws to try and do that.”at 38:38

US Treasuries & yields

• Chamath Palihapitiya · Social Capital

  AI firms' borrowing will crowd out Treasury demand, absorbing about 30% of issuance through 2027.

  Chamath says AI companies have moved from funding spending with cash to raising debt, with the AI complex owing roughly $344B. Their borrowing needs compete with the Treasury market, and he estimates they will need to raise about 30% of what the Treasury plans to issue through 2027. Chamath calls this crowding out the appetite for government paper, and he says these firms will start competing for demand. The debt figure is garbled in the transcript, so the $344B reading is approximate.

  “And what you can see here is that when you look at these AI firms, the amount of money that they need to raise through the end of twenty seventh, it's 30% of all of what treasury is about to issue in that same period.”at 34:19

• Chamath Palihapitiya · Social Capital

  AI borrowing now competes with Treasuries, raising borrowing costs economy-wide and default risk.

  Chamath says the AI complex now accounts for 40-50% of capital market equity and debt flows and 30-40% of Treasury issuance, so it competes directly with the government for capital. If hyperscalers must pay more, higher coupons cascade to neoclouds and others, pulling forward defaults in a negative loop. If Treasuries must pay more at auction, the cost of borrowing rises across an economy levered to the 2, 5 and 10-year yields. He calls this a market risk he has never seen before.

  “The AI complex has gotten so big that it now competes with US treasuries.”at 47:59

Frontier AI labs & model race

• David Sacks · Craft Ventures

  Anthropic has a big lead, and recursive self-improvement could widen it, refuting commoditization fears.

  Sacks argues that the S-1 refutes claims that Anthropic is being commoditized by open models, which would push its gross margin toward zero, or that it only works because of venture subsidies. Improving margins and falling costs show the business is strengthening. If recursive self-improvement is as far along as Anthropic claims, its lead could grow.

  “In fact, I think there's a lot of reasons to believe that Anthropic has a pretty big lead. And if you believe that recursive self improvement has kicked in as they are claiming it has, then their lead could actually grow.”at 10:10

Power demand & the grid

• Chamath Palihapitiya · Social Capital

  Data centers and power are the binding constraint on the AI buildout.

  Based on the Anthropic S-1 risk-factor analysis, Chamath concludes that data centers are the central constraint on AI. Within compute, the biggest issue is data centers, followed by financing commitments and power. He argues the whole AI complex therefore depends on data centers, which means it depends on power. This makes electricity policy the key political lever.

  “[symptomatic] of the entire AI complex, which is to say it's all about data centers, which means it's all about power.”at 21:15

IPOs, M&A & deal activity

• Chamath Palihapitiya · Social Capital

  Anthropic's smaller raise needs give it pricing power, a constructive setup for its IPO valuation.

  Chamath says Anthropic's CFO has built a large cash position, with about $125B entering June and a net burn of only about $9-10B. Because the company now needs to raise only about $80-120B instead of the $150-200B he had expected, the demands on capital markets are smaller. That gives Anthropic more pricing power and makes it easier to manage the float. He expects the IPO price to come in between $1.3T and $1.7T, with about $1.5T as the midpoint. The transcript's per-share figures are garbled, so the valuation range is taken as the speaker's intended reading.

  “And I think in order to do that, the price is probably now gonna come in between 1.3 on the low end to 1.7 on the high end. So call 1.5 the midpoint.”at 30:41

AI bubble & tech valuations

• Chamath Palihapitiya · Social Capital

  The AI complex is too big to fail, creating systemic risk, with more roads leading to downside.

  Chamath argues that when a few companies represent 40-50% of capital market equity and debt flows and 30-40% of Treasury issuance, they become too big to fail, which is a systemic problem. He says that for AI stocks, many roads lead to downside risk more than upside risk. He hedges that he is not saying a systemic problem exists, only that the structure is unprecedented and worth watching.

  “But I guess my structural takeaway maybe, Jason, is we're entering this very weird phase in The US economy... This complex is too big to fail.”at 50:21