Midterm Blood Bath, $4T AI Sell Off, Data Center Freeze, What's Next for Trump

Alphabet

Across the shows: 65 speakers in other episodes, 59 bullish and 42 bearish calls, past 30 days

• Chamath Palihapitiya · Social CapitalBullish

  Google could give its new data center capacity to Gemini, letting Gemini catch up with frontier model rivals.

  Chamath says Google has a large amount of capacity coming online. Google could route it to its own Gemini model rather than renting it out. Constraining the compute available to leading labs would make them pay more and degrade their user experience. Gemini would then have time to catch up in quality with the frontier labs, which are on a 60-to-90-day model improvement loop.

  “So Google has a ton of capacity coming online. They could say, I'm gonna give it to Gemini. What does that do? It actually allows Gemini to catch up in quality relative to the foundational models.”at 58:29

• Chamath Palihapitiya · Social CapitalBullish

  Google's control of power lets it take over data center capacity from compute renters like Crusoe.

  Chamath cites the news that Google just took over a Crusoe data center from its renter. He says the reason is that Google has all the core critical ingredients for AI infrastructure, specifically power. He reads this as the big platforms co-opting renter capacity, which makes Google a stronger position in the compute race than the renters it displaces.

  “Google just took over a Crusoe data center. They co opted the renter... why is that happening? It's because they have all of the core critical ingredients, specifically the power.”at 1:11:40

Space Exploration Technologies

Across the shows: 50 speakers in other episodes, 70 bullish and 25 bearish calls, past 30 days

• Chamath Palihapitiya · Social CapitalBullish

  SpaceX could terminate its 90-day Anthropic compute deal and reprice to spot, roughly doubling or tripling revenue.

  Chamath says the SpaceX and Anthropic compute deal has a mutual 90-day termination clause. Anthropic needs a disproportionate share of that compute. SpaceX could give notice and reprice to the spot market rate, which he says is probably double the current rate and could be triple. He says Anthropic would have to accept the higher price. That would lift SpaceX's compute revenue two to three times. He calls it the king-making move and says it would force other hyperscalers and neoclouds to re-examine their compute strategy.

  “Anthropic can term in ninety days. SpaceX can term in ninety days, terminate. If SpaceX says I'm gonna terminate this compute in ninety days, That is the king making move.”at 59:28

• Jason Calacanis · LAUNCHBullish

  SpaceX is severely undervalued as a stock, and Calacanis has kept buying shares through the sell-off.

  Calacanis argues that SpaceX's foresight in building Starlink and its fiber-like connectivity makes it easy to place data centers offshore, and that the company is severely undervalued as a stock. He says he has been buying shares through the market's turbulence, so he has a position.

  “...let alone the prescience of SpaceX, which is severely undervalued as a stock... I've been buying shares through this whole insanity.”at 1:19:01

Anthropic

Across the shows: 109 speakers in other episodes, 135 bullish and 93 bearish calls, past 30 days

• Jason Calacanis · LAUNCHBearish

  Anthropic's IPO valuation could fall to about $1 trillion, below earlier $1.5-4 trillion expectations.

  After the market sell-off, which wiped out close to $4 trillion in market cap, Jason says the selloff puts a cloud over Anthropic's IPO. Polymarket odds of an OpenAI or Anthropic listing this year fell to 71% from 95% before the midterms. A Wall Street Journal report says the haircut could now be baked in. Jason says the valuation could settle near $1 trillion instead of the $1.5-2 trillion and later $3-4 trillion that had been expected.

  “Instead of 1.5 to 2,000,000,000,000 and then going to 3 or 4,000,000,000,000 as Gavin Baker said... we could be looking at a 1,000,000,000,000”at 54:17

OpenAI

Across the shows: 46 speakers in other episodes, 23 bullish and 34 bearish calls, past 30 days

• David Friedberg · The Production BoardBullish

  OpenAI's 175-page proof disproving the exponential time hypothesis shows AI can produce breakthrough research quickly.

  Friedberg says OpenAI published a 175-page paper that disproves the exponential time hypothesis, a question that stood for 27 years. The team did the work end to end in under ten days, and some reports say a few hours. He notes the proof is still being evaluated and is not yet fully verified by third parties, but says it is holding so far. He uses it to show the scale of value AI can deliver, and calls it a moment of humility for mathematicians and computer scientists.

  “They did the whole thing end to end in less than ten days. Some reports have argued they did it in a few hours.”at 1:28:02

AI infrastructure & data centers

• Jason Calacanis · LAUNCH

  Data center opposition is largely posturing to win better deals; the governors will not stop buildout.

  Jason argues the Texas governor-elect and President Trump both look at data centers the same way: as a way to get better community deals. He says that for both, the bottom line is to avoid disturbing neighbors, avoid pollution and share some of the funds. He says much of the data center backlash is performative. The governors just want a better deal, and they will not block the buildout.

  “I think a lot of the data center stuff is performative. They just wanna get a better deal.”at 45:28

• Chamath Palihapitiya · Social Capital

  Data center backlash is emotional rather than fact-based, but the big-company tax breaks behind it are gratuitous.

  Chamath watched a SpaceX presentation to Cochise County, Arizona, on a six-gigawatt data center, where SpaceX claimed it would cut electricity costs 40%, avoid water use and cut property taxes 20%. The audience was hostile. He calls the backlash an emotional issue not tethered to facts that stands for oligarchy and economic inequality. But he thinks the deals are gratuitous. He questions why the seven or eight most powerful companies need tax breaks to invest hundreds of billions of dollars and generate trillions in revenue, and he expects incoming governors to take these incentives off the table.

  “should it really be the case that the most powerful seven or eight companies need a tax break to invest... Probably not.”at 56:50

• Chamath Palihapitiya · Social Capital

  A data center slowdown would favor hyperscalers with existing capacity and hurt neoclouds, which are renters.

  Chamath expects local governments to block or slow data center projects in the coming period and says this will cut data center buildout. He says hyperscalers would benefit because they already own capacity. Neoclouds would be hurt because they rent capacity and need new supply to grow. Frontier model labs would also be hurt, because they are so short of capacity and need it to grow.

  “I think what it means is we're gonna see a slowdown. And I think what that does is it benefits the hyperscalers because they have existing capacity. I think it's very bad for the Neo Clouds because they're renters.”at 57:11

• Jason Calacanis · LAUNCH

  If US states restrict data centers, compute moves offshore and the US becomes a net importer of tokens.

  Jason argues that US governors who try to slow data centers would not stop the AI buildout. Data centers could move to Canada, Mexico, Singapore, Malaysia and other locations. The Middle East is currently off the table because of the war, but he expects it to become a site once the war ends. Next-generation models would then be trained on non-US soil. The US would become a net importer of tokens, and the construction jobs and local taxes would go to other countries.

  “these data centers could be put in Canada, in Mexico, in Singapore, in Malaysia... We're going to be a net importer of tokens”at 1:01:21

• Chamath Palihapitiya · Social Capital

  Regulating data centers would constrain supply, leaving Meta, Amazon, Google, Microsoft and SpaceX in control.

  Chamath expects any AI legislation to be broad and to cover data centers and AI factories. He concedes it may bring some regulatory capture for the closed model companies, but argues the supply constraint would dominate. The result would be that the companies with the power, capital and sites (Meta, Amazon, Google, Microsoft and SpaceX) end up controlling compute. He adds that politicians on both sides would not want that outcome.

  “maybe you'll get some regulatory capture on the one end, but you're gonna constrain the supply so severely on the other that the real people that will be in control of it are Meta, Amazon, Google, Microsoft, and SpaceX.”at 1:10:55

• Chamath Palihapitiya · Social Capital

  Foundation model labs lack fortress balance sheets and are compute renters, so their data center capacity is at risk.

  Chamath says the frontier model companies are not safe. Being safe would require a fortress balance sheet, a profitable business model and quasi-infinite compute of their own. Instead they are renters of data center capacity, and that renter class is 'very much at risk' as the data center market tightens and the bigger players take control of power and sites.

  “What does home free mean? That they have a fortress balance sheet, a profitable business model, and quasi infinite compute for their own ambitions. And right now, they are renters.”at 1:11:12

• Jason Calacanis · LAUNCH

  Offshore data centers and desktop local models will be the trend over the next two years.

  Calacanis argues that agent usage will grow by 500 to 1,000 times, and that the compute will go wherever it can be placed. Local models are a fallback so users are not throttled, which he calls unmetered intelligence. If permitting makes US sites hard to build, capacity moves to Canada or Mexico, and Starlink and cheap fiber make that easy to connect. He makes this prediction explicitly.

  “So offshore data centers will be the trend of the next two years. That's what you're gonna see. And desktop usage of local models. I'll just make that prediction here.”at 1:14:24

• David Sacks · Craft Ventures

  Fully permitted data centers finish in 2027, but new large projects stall and the pipeline shrinks by 2028.

  Sacks's base case is that fully permitted projects get built in 2027. He expects few major new projects under the new state regime, since politicians now face campaign attacks for approving data centers and have no incentive to sign off. Under-construction projects may lack final permits, which could strand half-finished builds. Gigawatt sites cost around $50 billion and need certainty. Governors in big data center states favor moratoriums, so he expects the pipeline to shrink by 2028 and to hurt the economy, with Democrats blaming Trump.

  “My base case is that the projects that are fully permitted will get done in 2027, but I don't think there's gonna be major new projects in this new regime for a long time”at 1:16:46

• David Sacks · Craft Ventures

  If the US doesn't build data centers, they'll be built abroad and America will pay to import AI tokens.

  Sacks says AI superintelligence will bring many benefits, and that if the US doesn't act, China will benefit. He agrees with Calacanis that if the US does not build data centers, they will be built in other countries. The US would then pay to import the tokens, which he calls a very dangerous situation, citing Mexico and Canada as likely hosts.

  “And frankly, if we don't do it, China's going to... They're gonna be the ones that benefit. J. Cal, you're right that if The US doesn't build the data centers, they'll get built in other countries and we'll be paying to import the tokens.”at 1:29:52

US elections & politics

• Chamath Palihapitiya · Social Capital

  The midterm loss repudiated the economy, not ideology; ideology and money didn't determine outcomes.

  Chamath argues the vote was about economic reality, not socialism or the far left or far right. He offers three pieces of evidence. First, heavy spending in races like Texas and the Ohio Senate did not win; the candidate with less money would have been the better bet. Second, ideologically drawn gerrymanders backfired: Democrats won 11 of 15 Republican seats that were redrawn. Third, the decisive voters were centrists focused on affordability who went from Biden to Trump to Democrats. He concludes that Iran, Israel and AI had little to do with it.

  “This was a complete and total repudiation about the state of the economy. ... The economy is broken.”at 20:48

• David Sacks · Craft Ventures

  Losing young male podcast audiences to the war protest vote was a major factor in the midterm loss.

  Sacks argues the Republican coalition fractured when podcast-audience figures such as Tucker Carlson, Joe Rogan and Theo Von opposed the war. Young males, who are a core podcast audience, flipped their vote by 36 points in this midterm compared with 2024. He says many of these voters showed up and cast protest votes to signal they wanted out of the war. He says the GOP depends on this alternative media circuit and should have courted it more.

  “you lose the young males who are, again, the fans of all these podcasts. ... Young males flip their vote by 36 points in this midterm relative to the twenty twenty four election.”at 27:32

• Jason Calacanis · LAUNCH

  Polls and prediction markets were accurate; Trump's net approval fell from +5 to -28 over his term.

  Calacanis argues that polls and prediction markets were correct, while right-leaning commentators said they were wrong or manipulated. He points to the Silver Bulletin tracker, which he says nailed the cycle. He shows Trump's net approval starting above +5 and ending at -28, a swing of more than 30 points. He ties the decline to the Liberation Day tariffs, the summer 2025 ICE enforcement, the Epstein files controversy, and then the Iran war with its inflation.

  “All of the polls were correct. Everybody on the right said the polls were wrong.”at 28:44

AI regulation & safety

• David Friedberg · The Production Board

  Regulators will target compute and data centers like utilities, not chips or models; AI rules are likely.

  Friedberg expects governments to say they want to regulate compute. Chips and models are too hard to police, and software development and deployment are hard to regulate. But the use of software is limited by access to compute, so regulators will treat data centers like utilities. He says the chance of a major restriction is much higher than 10%, pointing to hundreds of state laws regulating model development, deployment and use, including state audits of how models operate.

  “They're not gonna regulate the chips or the models because those are too hard to kinda catch up on... I think the data centers end up being treated like utilities in the future”at 44:57

• David Sacks · Craft Ventures

  Democrats will push major AI legislation, likely a federal review regime for frontier models.

  Sacks argues Democrats want a surveillance and censorship state, supercharged by superintelligence. He expects a federal department of AI that would require all frontier models to be reviewed and approved before release. Regulators would raise harms such as disinformation and dissident thought, the same concerns they pressed on social media during COVID. The push comes in the next Congress, which begins in 2027.

  “it's a 100% that Democrats are gonna make a major push for AI legislation in the next congress.”at 1:07:42

• David Sacks · Craft Ventures

  A federal AI safety push and a new AI department risk censorship, surveillance and government control.

  Sacks splits AI regulation into two buckets. Federal action will focus on AI safety and a proposed federal department of AI, and he calls the risks of that path Orwellian, including censorship, surveillance and control. Infrastructure rules will be decided at the state and local level, which is a separate issue.

  “What the feds are gonna deal with is the AI safety and creating a federal department of AI and all of those Orwellian sorts of risk around what could turn into censorship and surveillance and control.”at 1:14:49

US deficit & government debt

• David Friedberg · The Production Board

  Runaway government spending drives the cost of living; the end of DOGE and a roughly $2T deficit confirm it.

  Friedberg argues the fundamental driver of cost of living is government spending: the more the government spends, the more expensive things become. He says DOGE's goal of making each government dollar earn a positive return was good, but Musk struggled, Congress showed no real interest in efficiency, and DOGE ended in June 2025. The July 2025 one big beautiful bill cut some entitlement costs but did not reduce overall spending, and the government still runs about a $2 trillion annual deficit. He argues this structural problem is what left Americans feeling they can't afford things.

  “I do think it is the fundamental driver on the cost of living is government spending. The more the government spends, the more expensive things get.”at 12:51

• David Friedberg · The Production Board

  A $2 trillion annual deficit will keep the cost of living high and worsen it; the bond market will signal the path.

  Friedberg argues that doubling government spending during COVID inflated asset prices, benefited asset owners and hurt wage earners. He expects no clean budget to pass in 2027 with this Congress and the sitting president. As a result, spending will continue as is. If the roughly $2 trillion-a-year deficit continues, costs will stay high and get worse. He says the bond market will show where this is headed.

  “I think the bond market will tell you where they think that's headed. Because if we continue this $2,000,000,000,000 a year deficit spending, things are gonna remain expensive, they're gonna get worse”at 42:39

Tariffs & trade policy

• Jason Calacanis · LAUNCH

  Liberation Day tariffs spiked inflation and failed to balance the budget as promised.

  Calacanis says supporters argued the Liberation Day tariffs would be great, that no prices would rise, that Americans wouldn't pay for them, and that they would balance the budget. None of that happened. The tariffs did not balance the budget and did spike inflation. He adds that the Supreme Court ruled against the administration on the tariffs. He argues the chaotic rollout was a major reason Trump's net approval fell sharply early in the term.

  “Not only did it not balance the budget, it did spike inflation.”at 29:53

• Chamath Palihapitiya · Social Capital

  Tariffs will look positive if committed investment follows through and revives domestic manufacturing.

  Chamath argues that tariffs paired with the investment commitments companies have made could revive US manufacturing. If that money follows through and manufacturing comes back, the policy will look positive as its effects compound over time. He says history will judge the presidency more favorably on the policies that endure, including border closure and tariffs, than on the war.

  “Then you have tariffs, which if you partner it with the amount of money that people have committed, if that money follows through and it actually revives domestic manufacturing... that's gonna look positive.”at 49:32

Middle East & Iran

• David Sacks · Craft Ventures

  The Iran war and its fuel-price fallout drove the GOP midterm loss more than anything else.

  Sacks argues the war with Iran was the main cause of the midterm rout through its economic consequences. Gasoline rose from about $3 to over $4 a gallon after the war began, diesel rose further, and that flowed into food, transport and fertilizer costs. Inflation rose from 2.4% in January to 3.7%, which, combined with the affordability narrative, produced the 'tidal wave.' He cites a chart showing Trump's net approval tracking gas prices, and a chart showing the Senate forecast shifting after the war began on 02/28/2026. He says 74% of voters called the Iran war a crucial or major factor, and those voters broke heavily for Democrats, including independents and younger Republicans.

  “It was this war and the economic consequences that follow from the war.”at 5:12

US inflation

• David Sacks · Craft Ventures

  Energy-driven inflation jumped from 2.4% to 3.7%, wiping out the gains from falling prices.

  Sacks says US inflation was 2.4% in January before the war, down from around 9% under Biden. It then rose to 3.7%, mostly because of gasoline and diesel costs, which raise transportation costs for all goods. That inflation stopped the expected rate cuts and produced what he calls a rate hike, pushing mortgage rates to roughly 7.5%. He argues this inflation, rather than the economy's overall health, is what hurt households.

  “you got inflation going from 2.4 to 3.7%. ... mostly driven by the cost of gas and diesel and all these other products where you have to pay a higher cost of transportation.”at 6:27

US growth & recession risk

• David Sacks · Craft Ventures

  Underlying US economy is strong: jobs robust, unemployment near 4.1%, Q3 GDP tracking 3.7%.

  Sacks argues the economy is not the problem. The new jobs report was robust, prior job creation was revised up, and unemployment fell to 4.1%, almost a record low. First-half GDP grew about 2.1% to 2.5%, and third-quarter tracking was 3.7%. He uses this to argue that the electorate reacted to inflation and affordability, not to a weak economy.

  “So just today, we had a new jobs report. It was a very robust jobs report. ... You've got unemployment tick down to almost record low, 4.1%.”at 7:29

AI: overall outlook

• David Friedberg · The Production Board

  US AI pessimism makes the AI-oligarchy narrative an easy political attack on tech.

  Friedberg cites a Gallup study on AI optimism. 93% of Chinese respondents think AI will mostly help the country and make their daily lives better. Only 36% of Americans say the same, placing the US fourth from last. He argues this low optimism, together with the perception that AI builders are the ones getting ahead while most people are not, made it easy for politicians to frame AI as an oligarchy issue. He expects AI to play a major role in how politics unfolds over the next year.

  “The most optimistic country for AI is China. 93% of Chinese citizens think AI will mostly help people in the country ... We're fourth on the list of countries at 36% of Americans believing that.”at 15:53

Wealth, inequality & billionaires

• Chamath Palihapitiya · Social Capital

  Asset owners profit in every rate and tariff regime while wage earners' economic capacity is structurally broken.

  Chamath says he and his peers, the asset owners of America, benefited in every economic condition. When rates were zero they used cheap capital to build businesses, and when rates were high they had pricing power. They gained under tariffs and without tariffs, and with or without subsidies. Wage earners, however, are left behind because the structural capacity of the US economy to support and grow them is broken. He says he got this wrong by assuming top-down economics would eventually reach everyone.

  “But we are few and far between. And then there's everybody else who are wage earners, and the structural capacity of the American economy to support and grow wage earners is broken.”at 20:26

US tax & spending policy

• Chamath Palihapitiya · Social Capital

  Proposes ending federal loan underwriting, cancelling student debt, and zeroing federal taxes under $150k.

  After concluding that the economy is broken for wage earners, Chamath sets out a policy framework to give that group more ownership of the US economy. It includes stopping federal underwriting of loans, combining that with cancelling student debt, and taking all federal taxes to zero for incomes below about $150,000. He presents this as the economic populist platform that could win the presidency in 2028.

  “stop the federal underwriting of loans, combine that with getting rid of student debt, take all federal taxes beneath a certain band, probably 150 k to zero.”at 24:32

US rates & the Fed

• David Sacks · Craft Ventures

  Once the war ends and energy prices fall, the Fed can cut rates and mortgage costs will drop.

  Sacks argues the inflation problem comes from oil, gasoline and diesel prices. If the war ends quickly, those prices should drop quickly and inflation should follow. That would give the Fed room to cut, bringing mortgage costs down and putting the economy back on track. He treats the Fed's cuts as dependent on the war ending.

  “If the war has ended quickly, those prices will drop very quickly. Inflation will come back down. That will give the Fed permission to cut rates, so mortgage costs come back down”at 28:19

US immigration

• David Sacks · Craft Ventures

  Trump deserves enormous credit for completely sealing the border and solving the immigration issue.

  Sacks says that while not every issue was executed perfectly, Trump deserves enormous credit for completely sealing the border and solving the immigration issue. He presents this as a clear achievement that the critics are overlooking when they look at the administration's record.

  “I do think that president Trump deserves enormous credit for completely sealing the border and solving that immigration issue”at 33:12

Oil & fuels

• David Sacks · Craft Ventures

  Ending the Iran war would push Brent crude toward $70 from about $110, lowering gas, diesel and food prices.

  Sacks argues that ending the war is the top fix for the administration. Brent fell 10% on election day when markets priced in only the possibility of an earlier end to the war, with no change in the Strait of Hormuz. He expects Brent to fall from about $110 (it has peaked at $117) to $70 if the war ends. That would cut gas and diesel prices and then food prices, letting strong GDP and employment data come into focus.

  “Because what will happen? The price of Brent crude will drop precipitously... It'll go from [$110] to 70. If this war ends, gas prices will drop, diesel will drop, the price of food will drop”at 35:49

Enterprise AI adoption & AI software

• Jason Calacanis · LAUNCH

  AI agent subscriptions return roughly $2M in value at his firm, while most corporate users still don't use AI daily.

  Calacanis runs a venture firm with about 20 people that invests in roughly 100 companies a year. He uses Claude and Grok subscriptions and estimates the firm gets about $2 million in value from them. Automation has raised investment velocity and media output. He says only 10 to 20% of corporate America uses the technology daily, so usage could rise five to ten times as adoption spreads to people who find agents easy to use.

  “I estimate we're getting $2,000,000 in value from that... only 10 or 20% of people in corporate America are using this technology every day on a regular basis.”at 1:13:00

AI pricing & unit economics

• Chamath Palihapitiya · Social Capital

  If data center buildout slows, compute capacity tightens and GPU-hour and megawatt prices spike.

  Chamath argues that if many data centers are slowed by constraints, there will be far less capacity than people expect. The cost per megawatt per month and the cost of a GPU hour would rise sharply. Only buyers with huge balance sheets would pay the premium, so a firm holding long-term capacity could resell it to Google or Meta instead of OpenAI.

  “if a bunch of data centers get slowed down, a bunch of constraint happens, there's gonna be a lot less capacity than people think. The dollar cost per megawatt per month is gonna go through the roof. The dollar cost of a GPU hour is gonna go through the roof”at 1:21:46