All-In Podcast Recap: Midterm Results, the AI Sell-Off and Data Center Uncertainty

The latest full episode of the All-In Podcast, "Midterm Blood Bath, $4T AI Sell Off, Data Center Freeze, What’s Next for Trump", came out on November 7, 2026 and runs about an hour and 33 minutes. It was recorded Friday morning, November 6. Jason Calacanis hosted with Chamath Palihapitiya, David Friedberg and David Sacks, who had missed the Tuesday election-night livestream. The panel walked through why they think Republicans lost so badly, what the next two years of the Trump presidency might look like, a sharp drop in AI-related stocks, and a closing science segment on an OpenAI research paper. Here is a plain-English summary, plus a timestamp guide so you can jump to the parts you care about.

This is an independent summary written by GNOUP. It is not affiliated with, sponsored by or endorsed by the All-In Podcast, OpenAI, Anthropic, any host, party or candidate, and GNOUP does not endorse any company, party or candidate mentioned. The views described below belong to the speakers. Nothing here is financial, investment or political advice.

A note on results: the hosts described the midterms as a large Democratic win and gave seat counts that they said were still being finalized, including outstanding Senate races in Alaska and Mississippi. Those figures are as stated on the show. Check official election results for final outcomes.

Key takeaways

• The hosts described a historic Democratic wave. Jason said Democrats swept both chambers of Congress, won at least 29 governor races, and were on track for about 54 to 56 Senate seats and at least 250 House seats, possibly 260. He said Republicans lost ground with every voter group, including white men, and lost longtime red states such as Texas, Kansas, Iowa and South Carolina that he said President Trump had won by double digits two years earlier. Sacks called it the biggest midterm defeat since the Watergate-era midterms. These are the hosts' recounting of the results, not a GNOUP tally.
• They disagreed on the main reason. Sacks pointed to the war in Iran and the higher gas, diesel, food and borrowing costs he said followed it. Friedberg argued the deeper problem was that the cost of living was never fixed, including after an effort to cut government spending faded in 2025. Chamath said he no longer thinks party labels explain the result and called it a protest against an economy that works for asset owners but not for wage earners. Jason said he had expected a Democratic sweep and pointed to a longer slide in the president's approval, which he said began before the war.
• Sacks's advice for a recovery was to end the war. He said the president's approval tracked the price of gasoline, that prediction markets only turned against Republicans after the war began on February 28, 2026, and that 74% of voters in a survey he cited called the war either crucial or a major factor in their vote for Congress. He argued that if the war ended, oil and gas prices would fall quickly, inflation would ease and the Federal Reserve would have more room to cut interest rates.
• AI stocks had a brutal week, in the hosts' telling. Jason said a broad group of companies, including Nvidia, SpaceX, Tesla, AMD and Micron, saw close to $4 trillion in market value wiped out, which he called the worst single day in stock-market history by total dollars (not by percentage). He said that cast a shadow over Anthropic's planned IPO: a Polymarket contract he cited had fallen from about a 95% chance of going public this year before the midterms to about 71%, and he said a Wall Street Journal report suggested the company might now be valued closer to $1 trillion instead of the much higher figures discussed earlier. He added that markets had rebounded a bit on the day they recorded.
• Data centers were the market flashpoint. The hosts tied the sell-off to fears that incoming Democratic governors, especially Texas governor-elect Gina Hinojosa, would slow or rework data-center deals. Chamath thought communities would demand clearer local benefits and fewer tax breaks. Sacks warned that even projects already under construction might still need later permits, and that politicians now have little incentive to approve new ones.
• They also argued about future AI rules. Friedberg and Sacks worried that a new Congress could try to tightly control who gets to build and use advanced AI. Chamath thought voters' real demand is economic fairness, and that heavy regulation could accidentally hand more power to the largest tech companies that already own computing capacity.
• Science Corner: an OpenAI paper on a long-standing computer-science puzzle. Friedberg said OpenAI published a 175-page paper claiming to disprove the Exponential Time Hypothesis, a 1999 idea about how long certain hard problems take to solve. He said the work is still being reviewed by others.

Newsworthy insights

These are claims and opinions voiced on the show, not independently verified findings by GNOUP, unless noted.

Why they think Republicans lost

• Sacks said inflation had been near 2.4% in January, before the war, after what he described as a drop from about 9% during the Biden years, and that it had since risen to about 3.7%. He said gasoline went from around $3 a gallon to more than $4, with diesel even higher, and that a rate increase had pushed mortgage costs to roughly 7.5%. He thought the election would be only a modest setback, because a jobs report released the morning they recorded looked strong, unemployment was about 4.1%, and economic growth was still solid (about 2.1% to 2.5% in the first half of the year, with third-quarter tracking around 3.7%). He said he was wrong: people vote on how affordable life feels, and he argued that when prices rise faster than wages, households lose purchasing power. He also said the war split the Republican coalition and cost support among popular independent media figures whom he thinks Republicans rely on, because much of traditional media is hostile to them.
• Jason quoted reporting that Vice President JD Vance had named four rebuild tasks: ending the war, focusing on the cost of living, putting "America first," and reviewing the country's relationship with Israel. Sacks said he largely agreed with Vance that the war was the driving force.
• Friedberg replayed a prediction he made after the 2024 election: that economic growth would be uneven, with big winners and losers, and that socialist-style politics would gain ground. He said an effort known as DOGE, aimed at making government spending more efficient, had effectively ended by June 5, 2025, and that Congress was never seriously on board. He said the government is still running a budget of about $7 trillion a year with a deficit of about $2 trillion a year. In his view, many Americans still cannot get ahead on health care and everyday costs, and they notice that stock-market and AI gains seem to flow to a small group of "tech oligarchs." He cited a Gallup survey in which 93% of people in China said AI would mostly help their country, while only 36% of Americans said the same, ranking the United States near the bottom of the list.
• Chamath said his own blind spot was looking at politics through his life as an asset owner. He argued that since the COVID-19 years, a large group of swing voters has punished whoever is in power when the economy feels broken, first backing Biden, then Trump, then Democrats again. He said campaign spending and redrawn election maps, which usually track party strategy, both failed: in his telling, a great deal of political money was wasted, and of 15 Republican-redrawn seats meant to help the GOP, Democrats won 11 and Republicans won only one. He does not think the war, Israel or AI explain most of the result. His conclusion was that an "economic populist" message could win the presidency in 2028 for either party, and that Democrats now have to choose between trying to pass economic measures and spending their energy on investigations.
• Jason said a tracker he cited, Nate Silver's Silver Bulletin, showed the president's net approval starting the term a little above plus 5% and ending near minus 28%. He argued the slide began with a chaotic tariff rollout, continued through summer immigration-enforcement controversies, and then deepened after the war. Sacks replied that Jason's prediction only lined up once the war began, that the average household was paying about $200 a month more for gasoline, and that young men's votes flipped by about 36 points compared with 2024. Jason, who described himself as an independent, said the administration's ideas were often better than its follow-through, and that many voters stopped believing the president was working for them.

What they think should happen next

• Sacks said ending the war was his first, second and third recommendation. He noted that the price of Brent crude fell about 10% on Election Day on the mere chance the war might end sooner, and he guessed that oil could fall from around $110 a barrel (he also mentioned a peak near $117) toward about $70 if the fighting stopped. He compared past midterms: Bill Clinton in 1994 and Barack Obama in 2010 both lost badly and later recovered after changing course, while George W. Bush after 2006 doubled down on the Iraq war and Republicans lost again in 2008. Sacks warned against advice to escalate now.
• Chamath said some parts of the Trump record, such as a tighter border, may look better with time, and that tariffs could too if promised manufacturing investment actually arrives. He suggested the president protect the parts of his agenda most likely to last, and trade for economic measures that help wage earners, such as ending federal income taxes below a band he put around $125,000 to $150,000.
• Jason urged the president to "go into deal mode." Among his ideas: use pardons early so investigations would not dominate the next two years; change how very large investment gains and loans against stock holdings are taxed; work on health care; raise the federal minimum wage to $15 an hour over four years; ban members of Congress from trading stocks and from certain foreign business dealings after leaving office; and encourage more home building. These were his personal suggestions, not recommendations from GNOUP.
• Friedberg worried that a divided government will have trouble passing a clean budget for 2027 and 2028. If large deficits continue, he said, prices will stay high. He argued that "socialist" outcomes can appear even when voters are not choosing that label, if people ask government to do more in order to lower costs.

Markets, data centers and who controls AI computing

• Chamath said opposition to data centers is emotional and not always tied to the details of a given project. He described a local meeting in an Arizona county where SpaceX staff presented plans for a large computing site, including claims that it would cut electricity costs, use no water and lower property taxes, and were shouted down. He thinks incoming governors will demand clearer local benefits and fewer tax breaks for companies that, in his view, do not need them. A slowdown, he said, would help the "hyperscalers" that already have capacity (he named Amazon, Meta, Microsoft, Google and SpaceX), and would hurt "neocloud" companies that rent capacity as well as the leading AI model labs that need more of it to keep improving.
• He sketched a more aggressive possibility: those large computing owners could keep new capacity for their own products, for example to help Google's Gemini catch up. He also said a computing contract between SpaceX and Anthropic has a 90-day exit for either side. If SpaceX ended it, he argued, it could resell that capacity at a much higher short-term "spot" price, perhaps two or three times the current rate. He framed that as a spreadsheet decision, not a judgment about either company. Jason thought Elon's working relationship with Anthropic might lead to a middle-ground price instead.
• Jason warned that if US states make it too hard to build, the next large AI models could be trained in places such as Canada, Mexico, Singapore or Malaysia (and in the Middle East if the war ends), and the United States could end up paying to import AI output. He said hundreds of thousands of well-paid construction jobs and the related local taxes would go with those projects.
• Sacks was less optimistic about Texas than some market commentators. After watching interviews with Hinojosa, he said she had not promised to protect existing projects and had talked about issuing no new permits until new rules are passed. He noted that a finished building can still need later approvals, such as an air-pollution permit. His base case was that fully permitted projects can still be completed in 2027, but that major new projects will be hard to start because both parties ran ads attacking politicians who approved data centers. He estimated that a single gigawatt-scale data center can cost around $50 billion, an amount no one will spend without more certainty. If building slows sharply by 2028, he expects Democrats to blame the president even though, in his view, state and local permitting would be the cause.
• Chamath said Google had just taken over a Crusoe data center, which he read as a sign that companies that control power and sites can absorb projects from smaller players.
• Friedberg compared advanced AI to a kind of ultimate power that governments will want to control. He thinks officials may try to treat data centers like utilities and regulate access to computing itself, because software is hard to police once it exists. He put the chance of a far-reaching clampdown higher than the 10% Jason suggested. Sacks said he is certain Democrats will push major federal AI legislation, and he predicted a new federal AI office that would review frontier models before release, which he compared to earlier government pressure on social media companies. Chamath doubted that full government control of the best models is the real priority, and said a political bargain that protects a few closed-model companies could backfire by making computing even scarcer and leaving more power with the largest platform companies.
• Jason said his venture firm is spending about $40,000 to $50,000 a year on AI tools and that he estimates they return a few million dollars in value. He argued that only a small share of corporate America uses these tools every day, so demand for computing could grow many times over, including on local desktop machines if cloud access is limited. That was his firm's experience, not a general rule.

Science Corner, as Friedberg explained it

• Friedberg said a core question in computer science is which problems a computer can solve, and how long that takes. Some problems can be solved in a fairly predictable amount of extra time as they get bigger. Others are easy to check once you have an answer, but may take much longer to solve. A 1999 idea called the Exponential Time Hypothesis said a particular family of those hard problems gets more difficult at an explosive, exponential rate.
• He said OpenAI's paper claims to show that this family can be solved faster than that hypothesis allowed. If the proof holds, he said, it would keep open the possibility that a much wider set of problems is solvable by computer. He emphasized that the 175-page paper uses novel methods, that outside experts are still examining it, and that the team produced it in less than 10 days, with some reports saying hours. These are Friedberg's account of a technical claim that has not been independently checked by GNOUP.

Watch the episode

Midterm Blood Bath, $4T AI Sell Off, Data Center Freeze, What’s Next for Trump on YouTube (All-In Podcast's official channel, published November 7, 2026)

Episode guide

Chapter times are taken from the episode's own YouTube description. Select a time to jump to that point on YouTube.

• 0:00: Bestie intros
• 3:37: Midterms postmortem: what led to a generational GOP collapse?
• 32:26: How can Trump salvage his presidency?
• 53:13: Market reaction: data center fears, about $4T in market cap shed, AI-ban talk
• 1:24:21: Science Corner: OpenAI and the Exponential Time Hypothesis

About this summary

This recap was prepared from the episode's publicly available transcript text and chapter list. Transcripts can contain errors, so figures and wording may differ slightly from the audio; watch the episode for the exact words. Election results, economic figures, valuations and deal details are as stated by the hosts and have not been verified by GNOUP. Mentions of companies, stocks or prediction markets are for general interest only. This is not financial or investment advice: do your own research or talk to a licensed professional before making investment decisions. Political views belong to the speakers, and GNOUP does not endorse any party or candidate.