Nvidia's board just chose $13.6 billion over national security. Every director judging Jensen Huang's affair with a suspected Chinese spy gets poorer if they find something serious. In the same statement that opened the investigation, the board said it has confidence in Jensen and his leadership team. So I went through Nvidia's own proxy filing to see who's doing the judging. Here's who's sitting on that board: - Mark Stevens owns about 34.1 million Nvidia shares, worth roughly $6.2 billion. - Tench Coxe owns about 30.6 million shares, worth roughly $5.5 billion. He's been on the board since 1993, the year Jensen co-founded Nvidia. - Harvey Jones owns about 7 million shares, worth roughly $1.3 billion. He's also been there since 1993. At Thursday's close of $181.60, I count about $13.6 billion of Nvidia stock held by the 9 independent directors. Jensen himself owns about 871 million shares, or 3.58% of the company, worth roughly $158 billion. So picture the choice in front of them: If they find something serious and push Jensen out, Nvidia's stock probably takes a hit, and so does their own net worth. If they find nothing, everything stays exactly the way it is. And the board isn't the only referee with money on the line: The President called Jensen an honorable guy who probably has an excuse, and said he hopes Jensen keeps running Nvidia. Washington's own plan to let Nvidia sell H200 chips in China gives the US government a 25% cut of those sales. That plan is now reportedly frozen. So almost every referee in this story has a reason to hope there's nothing to find. The one exception might be Congress, which wants Jensen at a closed-door briefing on November 18. That's the day after Nvidia reports earnings. Jensen Huang reached the final stage of wealth: The Chinese Spy. He's way too influential to face any serious consequences, and this immunity goes beyond just an affair...

Nvidia's board just chose $13.6 billion over national security. Every director judging Jensen Huang's affair with a suspected Chinese spy gets poorer if they find something serious. In the same statement that opened the investigation, the board said it has confidence in Jensen and his leadership team. So I went through Nvidia's own proxy filing to see who's doing the judging. Here's who's sitting on that board: - Mark Stevens owns about 34.1 million Nvidia shares, worth roughly $6.2 billion. - Tench Coxe owns about 30.6 million shares, worth roughly $5.5 billion. He's been on the board since 1993, the year Jensen co-founded Nvidia. - Harvey Jones owns about 7 million shares, worth roughly $1.3 billion. He's also been there since 1993. At Thursday's close of $181.60, I count about $13.6 billion of Nvidia stock held by the 9 independent directors. Jensen himself owns about 871 million shares, or 3.58% of the company, worth roughly $158 billion. So picture the choice in front of them: If they find something serious and push Jensen out, Nvidia's stock probably takes a hit, and so does their own net worth. If they find nothing, everything stays exactly the way it is. And the board isn't the only referee with money on the line: The President called Jensen an honorable guy who probably has an excuse, and said he hopes Jensen keeps running Nvidia. Washington's own plan to let Nvidia sell H200 chips in China gives the US government a 25% cut of those sales. That plan is now reportedly frozen. So almost every referee in this story has a reason to hope there's nothing to find. The one exception might be Congress, which wants Jensen at a closed-door briefing on November 18. That's the day after Nvidia reports earnings. Jensen Huang reached the final stage of wealth: The Chinese Spy. He's way too influential to face any serious consequences, and this immunity goes beyond just an affair...

AI was the perfect excuse to fire people, and companies just stopped using it. In May, US employers blamed AI for 38,579 job cuts. In October, they blamed it for 1,265. That's a drop of about 97% in 5 months. But companies still announced 43,994 job cuts in October, according to Challenger, Gray & Christmas. AI was the reason given for fewer than 3% of them. I think that tells us what AI was really doing in all those layoff announcements... It let executives present a cost cut as a bet on the future. And Meta is the best example I've seen: In January 2025, Mark Zuckerberg told Joe Rogan that Meta would have an AI that could work like a mid-level engineer that year. Then Meta drew up a plan called Project OT that considered shrinking some teams by as much as 60% and handing their work to AI agents, according to Reuters. On May 20, it cut about 8,000 jobs, roughly 10% of its staff. The night before, Zuckerberg cancelled a second wave that was planned for November. So what happened when the agents took over more of the work? Here's what Meta's own internal numbers showed: - Code changes to internal systems rose 220% from a year earlier. - Changes that turned into new or better features for users rose just 36%. - Major technical and security incidents rose 40%. - Time spent fixing those incidents rose 70%. One internal post warned that the agents were taking huge, disruptive actions that a human would almost never take. In July, Zuckerberg told staff at a town hall that agent-driven work hadn't sped things up the way Meta expected. And this was a company with internal leaderboards, some of them unofficial, ranking employees by how many AI tokens they used. So I'd say Meta measured how much AI its people used, and the AI produced a mountain of activity and a lot more cleanup. Now The Guardian reports Meta is rehiring managers. Meta isn't the only one walking it back though: - In a Robert Half survey of 2,000 US hiring managers, 32% said their companies brought back the exact roles they had cut because of AI. - An Atlanta Fed paper surveying nearly 750 executives found little evidence of near-term job losses across the economy due to AI. Meanwhile, 121,401 job cuts this year were announced with AI as the reason. That's about 1 in 5 of all the cuts Challenger has counted in 2026. Those people paid for the AI story, whether or not the AI could do their jobs. Companies should report revenue per employee 12 months later. If it didn't go up, explain what the AI did instead. How many of this year's AI layoffs were ever really about AI?

AI was the perfect excuse to fire people, and companies just stopped using it. In May, US employers blamed AI for 38,579 job cuts. In October, they blamed it for 1,265. That's a drop of about 97% in 5 months. But companies still announced 43,994 job cuts in October, according to Challenger, Gray & Christmas. AI was the reason given for fewer than 3% of them. I think that tells us what AI was really doing in all those layoff announcements... It let executives present a cost cut as a bet on the future. And Meta is the best example I've seen: In January 2025, Mark Zuckerberg told Joe Rogan that Meta would have an AI that could work like a mid-level engineer that year. Then Meta drew up a plan called Project OT that considered shrinking some teams by as much as 60% and handing their work to AI agents, according to Reuters. On May 20, it cut about 8,000 jobs, roughly 10% of its staff. The night before, Zuckerberg cancelled a second wave that was planned for November. So what happened when the agents took over more of the work? Here's what Meta's own internal numbers showed: - Code changes to internal systems rose 220% from a year earlier. - Changes that turned into new or better features for users rose just 36%. - Major technical and security incidents rose 40%. - Time spent fixing those incidents rose 70%. One internal post warned that the agents were taking huge, disruptive actions that a human would almost never take. In July, Zuckerberg told staff at a town hall that agent-driven work hadn't sped things up the way Meta expected. And this was a company with internal leaderboards, some of them unofficial, ranking employees by how many AI tokens they used. So I'd say Meta measured how much AI its people used, and the AI produced a mountain of activity and a lot more cleanup. Now The Guardian reports Meta is rehiring managers. Meta isn't the only one walking it back though: - In a Robert Half survey of 2,000 US hiring managers, 32% said their companies brought back the exact roles they had cut because of AI. - An Atlanta Fed paper surveying nearly 750 executives found little evidence of near-term job losses across the economy due to AI. Meanwhile, 121,401 job cuts this year were announced with AI as the reason. That's about 1 in 5 of all the cuts Challenger has counted in 2026. Those people paid for the AI story, whether or not the AI could do their jobs. Companies should report revenue per employee 12 months later. If it didn't go up, explain what the AI did instead. How many of this year's AI layoffs were ever really about AI?

Amazon's profits are about to be held hostage by Anthropic's stock price. Millions of retirement accounts own Amazon through index funds. And this year, a huge chunk of the profit those funds are paying for came from one private company getting more expensive on paper. That company is Anthropic. And its $1.5 trillion IPO price will likely have to come down, maybe closer to $1 trillion. Amazon depends more on that number than almost anyone. In the second quarter, Amazon made $27.5 billion of operating income. Its net income was $62.6 billion. Most of that gap was Anthropic. After Anthropic's May round valued it at $965 billion, Amazon marked up its Anthropic shares by about $50.5 billion. The first quarter had already added about $16.8 billion of Anthropic gains. That's roughly $67 billion of pre-tax gains in 6 months from revaluing one stake. On June 30, Amazon carried the whole position at about $190 billion. Its first $8 billion went in between 2023 and 2024. And the money runs in a circle: In April, Amazon put another $5 billion into Anthropic, with up to $20 billion more tied to milestones. In the same deal, Anthropic agreed to spend more than $100 billion on AWS over the next 10 years. So the loop goes: Amazon invests in Anthropic -> Anthropic commits to buying Amazon's cloud -> Anthropic raises money at a higher price -> Amazon books the higher price as profit None of this is illegal. The accounting rules tell Amazon to update the value when a new funding round sets a new price. But the third quarter shows what Amazon looks like with no new round to mark: Net income fell to $20.7 billion. Free cash flow over the last 12 months came in at minus $42.4 billion, with $77.3 billion of capex in the quarter alone. And the IPO changes the game completely. Once Anthropic trades publicly, Amazon's stake gets valued off a stock price that moves every single day. That price could start moving Amazon's reported profit far more directly. Investing .com ran an illustration: If Anthropic trades at $750 billion, Amazon's stake would be worth about $42 billion less than its June value. That's about TWICE what Amazon earned in the entire third quarter. To be clear, a $1 trillion price still sits above the $965 billion Amazon marked to in May. So the real risk starts after the first trade. Alphabet is in a similar spot. It carried $124.3 billion of private company stakes on June 30, and reports say most of that is Anthropic. If Anthropic falls it takes the big guys with it.

Amazon's profits are about to be held hostage by Anthropic's stock price. Millions of retirement accounts own Amazon through index funds. And this year, a huge chunk of the profit those funds are paying for came from one private company getting more expensive on paper. That company is Anthropic. And its $1.5 trillion IPO price will likely have to come down, maybe closer to $1 trillion. Amazon depends more on that number than almost anyone. In the second quarter, Amazon made $27.5 billion of operating income. Its net income was $62.6 billion. Most of that gap was Anthropic. After Anthropic's May round valued it at $965 billion, Amazon marked up its Anthropic shares by about $50.5 billion. The first quarter had already added about $16.8 billion of Anthropic gains. That's roughly $67 billion of pre-tax gains in 6 months from revaluing one stake. On June 30, Amazon carried the whole position at about $190 billion. Its first $8 billion went in between 2023 and 2024. And the money runs in a circle: In April, Amazon put another $5 billion into Anthropic, with up to $20 billion more tied to milestones. In the same deal, Anthropic agreed to spend more than $100 billion on AWS over the next 10 years. So the loop goes: Amazon invests in Anthropic -> Anthropic commits to buying Amazon's cloud -> Anthropic raises money at a higher price -> Amazon books the higher price as profit None of this is illegal. The accounting rules tell Amazon to update the value when a new funding round sets a new price. But the third quarter shows what Amazon looks like with no new round to mark: Net income fell to $20.7 billion. Free cash flow over the last 12 months came in at minus $42.4 billion, with $77.3 billion of capex in the quarter alone. And the IPO changes the game completely. Once Anthropic trades publicly, Amazon's stake gets valued off a stock price that moves every single day. That price could start moving Amazon's reported profit far more directly. Investing .com ran an illustration: If Anthropic trades at $750 billion, Amazon's stake would be worth about $42 billion less than its June value. That's about TWICE what Amazon earned in the entire third quarter. To be clear, a $1 trillion price still sits above the $965 billion Amazon marked to in May. So the real risk starts after the first trade. Alphabet is in a similar spot. It carried $124.3 billion of private company stakes on June 30, and reports say most of that is Anthropic. If Anthropic falls it takes the big guys with it.

This little auction matters more than everything the Nasdaq 100 lost the same day. Stocks trade every second, so everybody knows what Nvidia's worth at any moment. Private data center loans almost never get priced in public like this. And for 2 years, the whole AI boom has been funded on the assumption that this paper's safe. In 13 days, real buyers have to name a real price. ING is seeking offers on a dozen loans through a bids wanted in competition process, with bids due November 19 and a $10 million minimum. The pile includes debt tied to Vantage Data Centers and CyrusOne. And it includes a $25 million piece tied to the New Mexico site in Stargate, the venture backed by OpenAI, SoftBank and Oracle. Bloomberg calls this a rare auction of only data center debt. No price has been reported, and ING declined to comment. $25 million is pocket change on a campus reported at $165 billion, and I won't pretend otherwise. But just look at which campus it is: Oracle sent a force majeure notice to the developer of that site in September. That's the legal notice saying events beyond a company's control may excuse it from its obligations. Oracle says the schedule hasn't changed, and it calls these notices commonplace on projects this size. New Mexico regulators have until November 23 to decide the air permit for the site's power plant. That's 4 days AFTER the bids are due. The FT has already reported the project's roughly $18 billion construction loan trading at stressed levels. So anybody bidding on that slice is putting a price on a permit nobody has yet. And ING isn't alone: Startup Fortune, citing Bloomberg, says JPMorgan, Morgan Stanley, SMBC and MUFG have been shopping pieces of AI debt they helped underwrite. Andrzej Skiba of RBC's BlueBay unit told Bloomberg he's seeing a "massive hedge fund capitulation" in the space. CoreWeave's 9.625% notes due 2032 last traded around 76 cents on the dollar, for a yield of about 16%. The cost of insuring CoreWeave's debt was reported around 1,164 basis points this week, close to its October record. SoftBank's shares fell more than 9% in Tokyo on Friday, and its default insurance hit a 4-year high. Meanwhile Apple went out for about $7 billion of bonds this week and drew about $31 billion of orders. That's the tell for me, because lenders still have plenty of money. They're just getting pickier about AI paper. And the AI trade has needed the opposite. The money behind these campuses has to keep rolling from one lender to the next, and that only works if someone's always willing to take the paper at a good price. If the next buyer wants a big discount, the bank thinks twice about the next loan. Then the next campus costs more to finance, and eventually somebody's forecast for 2028 falls apart. Now, a smart person would say I'm making a mountain out of a molehill. $500 million is a rounding error in a market this big. Banks sell loans every day to manage their books. ING may have decided this months ago for reasons that have nothing to do with AI. Investment grade credit held up fine on Thursday, with LQD up 0.15%. And Oracle reportedly can't terminate the New Mexico lease, which should comfort the lenders. If ING gets close to full price, I'm wrong and the panic was overdone. But I think that's the point of watching this, and I'd bet the result leaks. CoreWeave reports November 10. Nvidia reports November 17. Then the ING bids land on November 19. I'd pay more attention to the third one. So who wins this one: ING getting close to full price for those Stargate loans, or a buyer demanding a serious discount?

This little auction matters more than everything the Nasdaq 100 lost the same day. Stocks trade every second, so everybody knows what Nvidia's worth at any moment. Private data center loans almost never get priced in public like this. And for 2 years, the whole AI boom has been funded on the assumption that this paper's safe. In 13 days, real buyers have to name a real price. ING is seeking offers on a dozen loans through a bids wanted in competition process, with bids due November 19 and a $10 million minimum. The pile includes debt tied to Vantage Data Centers and CyrusOne. And it includes a $25 million piece tied to the New Mexico site in Stargate, the venture backed by OpenAI, SoftBank and Oracle. Bloomberg calls this a rare auction of only data center debt. No price has been reported, and ING declined to comment. $25 million is pocket change on a campus reported at $165 billion, and I won't pretend otherwise. But just look at which campus it is: Oracle sent a force majeure notice to the developer of that site in September. That's the legal notice saying events beyond a company's control may excuse it from its obligations. Oracle says the schedule hasn't changed, and it calls these notices commonplace on projects this size. New Mexico regulators have until November 23 to decide the air permit for the site's power plant. That's 4 days AFTER the bids are due. The FT has already reported the project's roughly $18 billion construction loan trading at stressed levels. So anybody bidding on that slice is putting a price on a permit nobody has yet. And ING isn't alone: Startup Fortune, citing Bloomberg, says JPMorgan, Morgan Stanley, SMBC and MUFG have been shopping pieces of AI debt they helped underwrite. Andrzej Skiba of RBC's BlueBay unit told Bloomberg he's seeing a "massive hedge fund capitulation" in the space. CoreWeave's 9.625% notes due 2032 last traded around 76 cents on the dollar, for a yield of about 16%. The cost of insuring CoreWeave's debt was reported around 1,164 basis points this week, close to its October record. SoftBank's shares fell more than 9% in Tokyo on Friday, and its default insurance hit a 4-year high. Meanwhile Apple went out for about $7 billion of bonds this week and drew about $31 billion of orders. That's the tell for me, because lenders still have plenty of money. They're just getting pickier about AI paper. And the AI trade has needed the opposite. The money behind these campuses has to keep rolling from one lender to the next, and that only works if someone's always willing to take the paper at a good price. If the next buyer wants a big discount, the bank thinks twice about the next loan. Then the next campus costs more to finance, and eventually somebody's forecast for 2028 falls apart. Now, a smart person would say I'm making a mountain out of a molehill. $500 million is a rounding error in a market this big. Banks sell loans every day to manage their books. ING may have decided this months ago for reasons that have nothing to do with AI. Investment grade credit held up fine on Thursday, with LQD up 0.15%. And Oracle reportedly can't terminate the New Mexico lease, which should comfort the lenders. If ING gets close to full price, I'm wrong and the panic was overdone. But I think that's the point of watching this, and I'd bet the result leaks. CoreWeave reports November 10. Nvidia reports November 17. Then the ING bids land on November 19. I'd pay more attention to the third one. So who wins this one: ING getting close to full price for those Stargate loans, or a buyer demanding a serious discount?

Jensen Huang says engineers should burn $250,000 of AI a year, and Apple just cut its engineers' AI budget to about $50,000. And Apple plans to cut it again. Last month, some Apple engineers saw their daily AI spending cap drop from $300 to $200, Bloomberg reported. Some engineering teams could be cut to $100 a day next. Many groups outside engineering already sit at $100, and Apple has considered going as low as $25 for some of them. In March, Jensen said on the All-In podcast that he'd be deeply alarmed if a $500,000 engineer used less than $250,000 of AI tokens in a year. He also said Dario Amodei's forecast of hundreds of billions in AI revenue was too conservative. So Jensen's rule is simple: AI spending should run about half of what an engineer gets paid. By my math, at about 250 workdays a year, $200 a day comes to roughly $50,000. The planned $100 cap comes to roughly $25,000. By Jensen's rule, a $100 daily cap only makes sense for an engineer paid about $50,000 a year. And Apple still wants its people using MORE AI. It's steering teams toward cheaper models and AI that runs on the device, and Bloomberg's Mark Gurman has said Apple's internal work runs heavily on Anthropic's models. I think Apple just showed where the real limit on the AI boom sits now: The finance department. And that's just ONE company: - Accenture capped Claude at $29 a month per employee on October 1, after some staff had been spending up to $1,000 a month, 404 Media reported. - Uber burned through its entire 2026 AI budget in about 4 months, then capped AI coding tools at $1,500 a month per tool. - Microsoft has cut allowances too, and its AI chief Mustafa Suleyman told Bloomberg that Anthropic is extremely expensive and that Microsoft wants to eventually get rid of that cost. - Meta had internal leaderboards ranking staff by how many tokens they used, and now it has introduced token budgets. The usage is clearly there. Uber's engineers used the tools so much they spent a full year's budget in 4 months. So the people saying stop are the ones paying the bill. That's a huge problem for the AI labs, for one reason: Their spending is already locked in. Anthropic has committed to $895 billion, mostly for computing power. OpenAI has signed more than $1.4 trillion of compute deals. And paying for all of it depends on companies spending more on AI per worker every year. Apple, Accenture, Uber and Microsoft all pay for Anthropic's models, and Anthropic could start its IPO roadshow as early as next week. Its own filing shows revenue grew about 47% from the second quarter to the third, down from about 144% the quarter before, and price cuts took about $600 million out of revenue. Plenty of people will say this is just housekeeping, and they have numbers: Bloomberg says few Apple engineers ever used the full $300, so the savings may be small. Cheaper models can do the same work for less. Apple's R&D still rose 32% to $45.8 billion in fiscal 2026, partly because of AI. And more companies start using AI every month, so total spending can keep growing even with a cap on each worker. But Jensen's pitch depends on spending per worker going UP. Apple, Microsoft, Meta, Uber and Accenture are all writing rules that push it down. So the first question for Anthropic's roadshow should be simple: How much are its biggest customers spending per employee compared with a year ago, and how many of them have put caps in place? Who's right about what an engineer's AI is worth, Jensen Huang or Apple's finance team?

Jensen Huang says engineers should burn $250,000 of AI a year, and Apple just cut its engineers' AI budget to about $50,000. And Apple plans to cut it again. Last month, some Apple engineers saw their daily AI spending cap drop from $300 to $200, Bloomberg reported. Some engineering teams could be cut to $100 a day next. Many groups outside engineering already sit at $100, and Apple has considered going as low as $25 for some of them. In March, Jensen said on the All-In podcast that he'd be deeply alarmed if a $500,000 engineer used less than $250,000 of AI tokens in a year. He also said Dario Amodei's forecast of hundreds of billions in AI revenue was too conservative. So Jensen's rule is simple: AI spending should run about half of what an engineer gets paid. By my math, at about 250 workdays a year, $200 a day comes to roughly $50,000. The planned $100 cap comes to roughly $25,000. By Jensen's rule, a $100 daily cap only makes sense for an engineer paid about $50,000 a year. And Apple still wants its people using MORE AI. It's steering teams toward cheaper models and AI that runs on the device, and Bloomberg's Mark Gurman has said Apple's internal work runs heavily on Anthropic's models. I think Apple just showed where the real limit on the AI boom sits now: The finance department. And that's just ONE company: - Accenture capped Claude at $29 a month per employee on October 1, after some staff had been spending up to $1,000 a month, 404 Media reported. - Uber burned through its entire 2026 AI budget in about 4 months, then capped AI coding tools at $1,500 a month per tool. - Microsoft has cut allowances too, and its AI chief Mustafa Suleyman told Bloomberg that Anthropic is extremely expensive and that Microsoft wants to eventually get rid of that cost. - Meta had internal leaderboards ranking staff by how many tokens they used, and now it has introduced token budgets. The usage is clearly there. Uber's engineers used the tools so much they spent a full year's budget in 4 months. So the people saying stop are the ones paying the bill. That's a huge problem for the AI labs, for one reason: Their spending is already locked in. Anthropic has committed to $895 billion, mostly for computing power. OpenAI has signed more than $1.4 trillion of compute deals. And paying for all of it depends on companies spending more on AI per worker every year. Apple, Accenture, Uber and Microsoft all pay for Anthropic's models, and Anthropic could start its IPO roadshow as early as next week. Its own filing shows revenue grew about 47% from the second quarter to the third, down from about 144% the quarter before, and price cuts took about $600 million out of revenue. Plenty of people will say this is just housekeeping, and they have numbers: Bloomberg says few Apple engineers ever used the full $300, so the savings may be small. Cheaper models can do the same work for less. Apple's R&D still rose 32% to $45.8 billion in fiscal 2026, partly because of AI. And more companies start using AI every month, so total spending can keep growing even with a cap on each worker. But Jensen's pitch depends on spending per worker going UP. Apple, Microsoft, Meta, Uber and Accenture are all writing rules that push it down. So the first question for Anthropic's roadshow should be simple: How much are its biggest customers spending per employee compared with a year ago, and how many of them have put caps in place? Who's right about what an engineer's AI is worth, Jensen Huang or Apple's finance team?

Morgan Stanley's data shows hedge funds are underweight AI for the first time since 2023, while dip buyers poured a record $6.1 billion into a single 3X chip fund. This is the most dangerous trade in the market right now, with 2 groups of people going in opposite directions. That Morgan Stanley data, as reported, shows the pros have cut their total stock exposure by more than 8% since June. That's their biggest retreat since at least 2017. Goldman's data is even uglier. Hedge funds sold AI stocks for 5 straight sessions in October. And those 5 days erased ALL of their net AI buying since the start of 2025. Goldman's read is that this unwind is closer to its start than its end, and hedge fund AI holdings still rank in the top 17% of the last 3 years. So who stepped in right after that? Retail tech buying from October 15 to 21 hit its highest level since July. The favorites were SanDisk, TSMC and Nebius. Tech ETFs took in their second largest weekly inflow on record. Then retail investors put record money into leveraged equity ETFs over 2 weeks. And the Direxion 3X semiconductor fund pulled in that $6.1 billion in one week. That fund then fell 20.5% in a single day, and by October 23 it was on pace to lose more than 43% for the month. That product is built to move 3 times as hard as the chip index every day, in both directions. And in that same stretch, 47 AI and chip stocks lost about $5 trillion in market value from October 6, according to Yahoo Finance data. Nvidia and Alphabet lost about $835 billion each. But it gets even worse... I think the real danger is the supply about to hit the market: Anthropic is weighing 15% to 20% of its IPO for individual investors. JPMorgan is running US retail sales, and Jamie Dimon is expected to help pitch it. Citi is handling retail in Japan, Britain, Australia, Canada and parts of Europe. Investors are pushing that valuation down to around $1.5 trillion from about $2 trillion, according to the Wall Street Journal. OpenAI's CFO said in April that it'll reserve IPO shares for retail too. And SpaceX showed all of them what's possible, because its $86.2 billion IPO drew more than $100 billion of demand from individuals. The most sophisticated money on Earth is reducing its AI risk. And at the same time, the biggest AI companies ever built are lining up to sell shares directly to regular people. I'm not claiming anyone planned that. But when the professionals step back and the deals still need trillions in valuation, the banks go where the buyers are. Households are already stretched too. Federal Reserve data shows stocks made up a record 48% of household and nonprofit financial assets in Q2. At the dot-com peak that figure was 39%. You'd rather be early with the hedge funds than be the person buying these IPOs from the banks that just watched their best clients sell. So who wins this trade: The hedge funds or the retail crowd?

Morgan Stanley's data shows hedge funds are underweight AI for the first time since 2023, while dip buyers poured a record $6.1 billion into a single 3X chip fund. This is the most dangerous trade in the market right now, with 2 groups of people going in opposite directions. That Morgan Stanley data, as reported, shows the pros have cut their total stock exposure by more than 8% since June. That's their biggest retreat since at least 2017. Goldman's data is even uglier. Hedge funds sold AI stocks for 5 straight sessions in October. And those 5 days erased ALL of their net AI buying since the start of 2025. Goldman's read is that this unwind is closer to its start than its end, and hedge fund AI holdings still rank in the top 17% of the last 3 years. So who stepped in right after that? Retail tech buying from October 15 to 21 hit its highest level since July. The favorites were SanDisk, TSMC and Nebius. Tech ETFs took in their second largest weekly inflow on record. Then retail investors put record money into leveraged equity ETFs over 2 weeks. And the Direxion 3X semiconductor fund pulled in that $6.1 billion in one week. That fund then fell 20.5% in a single day, and by October 23 it was on pace to lose more than 43% for the month. That product is built to move 3 times as hard as the chip index every day, in both directions. And in that same stretch, 47 AI and chip stocks lost about $5 trillion in market value from October 6, according to Yahoo Finance data. Nvidia and Alphabet lost about $835 billion each. But it gets even worse... I think the real danger is the supply about to hit the market: Anthropic is weighing 15% to 20% of its IPO for individual investors. JPMorgan is running US retail sales, and Jamie Dimon is expected to help pitch it. Citi is handling retail in Japan, Britain, Australia, Canada and parts of Europe. Investors are pushing that valuation down to around $1.5 trillion from about $2 trillion, according to the Wall Street Journal. OpenAI's CFO said in April that it'll reserve IPO shares for retail too. And SpaceX showed all of them what's possible, because its $86.2 billion IPO drew more than $100 billion of demand from individuals. The most sophisticated money on Earth is reducing its AI risk. And at the same time, the biggest AI companies ever built are lining up to sell shares directly to regular people. I'm not claiming anyone planned that. But when the professionals step back and the deals still need trillions in valuation, the banks go where the buyers are. Households are already stretched too. Federal Reserve data shows stocks made up a record 48% of household and nonprofit financial assets in Q2. At the dot-com peak that figure was 39%. You'd rather be early with the hedge funds than be the person buying these IPOs from the banks that just watched their best clients sell. So who wins this trade: The hedge funds or the retail crowd?

Apple just bought back more of its own stock last quarter than Microsoft, Alphabet, Amazon, Meta and Oracle COMBINED. And that's genuinely embarrassing. For 3 years, Apple was the company everybody laughed at. It had the broken Siri and the tiny AI budget, and every analyst on TV said Tim Cook had slept through the biggest technology shift since the iPhone. So let's check the scoreboard after Monday night: Apple spent $24 billion buying back stock in the September quarter. A Reuters analysis of LSEG data puts the buybacks of those 5 AI giants at $13.6 billion for the same quarter. That's down 83% from $81.8 billion a year earlier. Alphabet and Meta haven't repurchased a single share in 2026. Amazon hasn't bought back stock since 2022. These are some of the most profitable companies on Earth, and they've nearly stopped buying back stock so they can pour concrete. Now this is where it gets really insane: Amazon spent $77.3 billion on capex in its latest quarter. It brought in $42.4 billion of operating cash and raised $48.1 billion in long-term debt. Apple spent $2.24 billion. That's under 3% of what Amazon spent, and Apple still pulled in $28 billion of operating cash and handed basically all of it back to shareholders. And Wall Street noticed too. Apple closed last Friday at $5.03 trillion, above Nvidia at $4.93 trillion. So Apple ended the week worth more than the company selling chips to everybody else. Meanwhile, Meta, Alphabet, Microsoft and Amazon lost about $665 billion in market value in the first session after each one reported. Alphabet alone lost about $496.5 billion, in a quarter where Google Cloud revenue grew 103%. So Google delivered the growth the entire AI trade has been begging for, and investors still looked at the bill and ran. But the part I find hilarious is the Siri deal: Apple reportedly pays Google around $1 billion a year to run Gemini inside Siri. Neither company has confirmed the terms. And Google has been paying Apple around $20 billion a year to stay the default search engine on the iPhone. So Alphabet burned through its cash flow to build the model, posted negative $9.1 billion of free cash flow last quarter, and then rented that model to Apple for a rumored 5% of what it pays Apple for search. That's the whole AI bull case in one contract, and I think it points in a very uncomfortable direction. The money was always supposed to flow to whoever owned the smartest model. But Apple owns the iPhone, and I don't think the people holding one care whose model is answering them. So the builder takes the debt, the depreciation and the power bill, and the owner of the customer takes the toll. If I'm right about that, a lot of this capex never earns its money back. So who looks smarter 2 years from now: Apple, or the 5 giants who stopped paying their shareholders to build the future?

Apple just bought back more of its own stock last quarter than Microsoft, Alphabet, Amazon, Meta and Oracle COMBINED. And that's genuinely embarrassing. For 3 years, Apple was the company everybody laughed at. It had the broken Siri and the tiny AI budget, and every analyst on TV said Tim Cook had slept through the biggest technology shift since the iPhone. So let's check the scoreboard after Monday night: Apple spent $24 billion buying back stock in the September quarter. A Reuters analysis of LSEG data puts the buybacks of those 5 AI giants at $13.6 billion for the same quarter. That's down 83% from $81.8 billion a year earlier. Alphabet and Meta haven't repurchased a single share in 2026. Amazon hasn't bought back stock since 2022. These are some of the most profitable companies on Earth, and they've nearly stopped buying back stock so they can pour concrete. Now this is where it gets really insane: Amazon spent $77.3 billion on capex in its latest quarter. It brought in $42.4 billion of operating cash and raised $48.1 billion in long-term debt. Apple spent $2.24 billion. That's under 3% of what Amazon spent, and Apple still pulled in $28 billion of operating cash and handed basically all of it back to shareholders. And Wall Street noticed too. Apple closed last Friday at $5.03 trillion, above Nvidia at $4.93 trillion. So Apple ended the week worth more than the company selling chips to everybody else. Meanwhile, Meta, Alphabet, Microsoft and Amazon lost about $665 billion in market value in the first session after each one reported. Alphabet alone lost about $496.5 billion, in a quarter where Google Cloud revenue grew 103%. So Google delivered the growth the entire AI trade has been begging for, and investors still looked at the bill and ran. But the part I find hilarious is the Siri deal: Apple reportedly pays Google around $1 billion a year to run Gemini inside Siri. Neither company has confirmed the terms. And Google has been paying Apple around $20 billion a year to stay the default search engine on the iPhone. So Alphabet burned through its cash flow to build the model, posted negative $9.1 billion of free cash flow last quarter, and then rented that model to Apple for a rumored 5% of what it pays Apple for search. That's the whole AI bull case in one contract, and I think it points in a very uncomfortable direction. The money was always supposed to flow to whoever owned the smartest model. But Apple owns the iPhone, and I don't think the people holding one care whose model is answering them. So the builder takes the debt, the depreciation and the power bill, and the owner of the customer takes the toll. If I'm right about that, a lot of this capex never earns its money back. So who looks smarter 2 years from now: Apple, or the 5 giants who stopped paying their shareholders to build the future?

Anthropic literally kept its biggest number out of its IPO filing for a week, and it turned out to be $895 BILLION in bills. If you were thinking about buying this stock, that's literally the number you should care about most. When Anthropic filed publicly on October 26, the line for its total purchase commitments had an empty space where the dollar figure should be. The amended filing out today fills it in: $895.4 billion, almost all of it for computing power and infrastructure. About $448 billion comes due by the end of 2030, and roughly $70 billion over the next 12 months. Anthropic brought in about $17 billion last quarter. That means it has signed up for bills worth almost 13 times its yearly revenue. And the same filing shows the squeeze has already started: - Revenue grew about 47% from the second quarter to the third, down from about 144% the quarter before. - Operating costs jumped to $17.5 billion from $10.6 billion. - Anthropic swung to a $567 million operating loss, after a $974 million operating profit in the second quarter. - Price cuts took about $600 million out of revenue between those 2 quarters. The filing gives the reason for the loss itself: A lot of Anthropic's compute was bought under short-term deals, and those currently cost significantly more than its multi-year contracts. So right now Anthropic is paying top dollar for computing power while charging its customers less. The cheaper long-term capacity it's waiting on is what most of the $895 billion pays for. That's the trade anyone buying at the $1.4 trillion to $1.5 trillion valuation being discussed is signing up for. Revenue has to grow fast enough to fill hundreds of billions of dollars of capacity, in the middle of a price war that's pushing down what Anthropic can charge. And the money behind all of it is getting harder to raise. The WSJ reported that OpenAI asked investors for $30 billion last week and had about $11 billion committed by Friday. The 2 biggest AI labs are now cutting prices against each other and chasing the same investors in the same month. Revenue is still up about 12 times from a year ago, Anthropic has $146.6 billion of liquidity, and on its adjusted numbers it still made a 15% operating margin. Most of the $23.3 billion net loss is an accounting charge on convertible notes, and about $2 billion of costs came from a non-cash charge tied to an employee donation matching program... But that adjusted margin fell from 19% to 15% in a single quarter, and the $895 billion stays exactly where it is no matter what the margin does. How much of the $895 billion can be canceled if demand slows? And in which quarter does the cheaper long-term compute finally replace the expensive short-term kind? Anthropic kept that number out of its filing for a week. I wonder why...

Anthropic literally kept its biggest number out of its IPO filing for a week, and it turned out to be $895 BILLION in bills. If you were thinking about buying this stock, that's literally the number you should care about most. When Anthropic filed publicly on October 26, the line for its total purchase commitments had an empty space where the dollar figure should be. The amended filing out today fills it in: $895.4 billion, almost all of it for computing power and infrastructure. About $448 billion comes due by the end of 2030, and roughly $70 billion over the next 12 months. Anthropic brought in about $17 billion last quarter. That means it has signed up for bills worth almost 13 times its yearly revenue. And the same filing shows the squeeze has already started: - Revenue grew about 47% from the second quarter to the third, down from about 144% the quarter before. - Operating costs jumped to $17.5 billion from $10.6 billion. - Anthropic swung to a $567 million operating loss, after a $974 million operating profit in the second quarter. - Price cuts took about $600 million out of revenue between those 2 quarters. The filing gives the reason for the loss itself: A lot of Anthropic's compute was bought under short-term deals, and those currently cost significantly more than its multi-year contracts. So right now Anthropic is paying top dollar for computing power while charging its customers less. The cheaper long-term capacity it's waiting on is what most of the $895 billion pays for. That's the trade anyone buying at the $1.4 trillion to $1.5 trillion valuation being discussed is signing up for. Revenue has to grow fast enough to fill hundreds of billions of dollars of capacity, in the middle of a price war that's pushing down what Anthropic can charge. And the money behind all of it is getting harder to raise. The WSJ reported that OpenAI asked investors for $30 billion last week and had about $11 billion committed by Friday. The 2 biggest AI labs are now cutting prices against each other and chasing the same investors in the same month. Revenue is still up about 12 times from a year ago, Anthropic has $146.6 billion of liquidity, and on its adjusted numbers it still made a 15% operating margin. Most of the $23.3 billion net loss is an accounting charge on convertible notes, and about $2 billion of costs came from a non-cash charge tied to an employee donation matching program... But that adjusted margin fell from 19% to 15% in a single quarter, and the $895 billion stays exactly where it is no matter what the margin does. How much of the $895 billion can be canceled if demand slows? And in which quarter does the cheaper long-term compute finally replace the expensive short-term kind? Anthropic kept that number out of its filing for a week. I wonder why...

AI just crashed Intel, and Intel's own CEO said so. On Friday, Intel had its worst day since 1974. The stock fell 28.58%, and the US government's 10% stake lost about $12 billion in a single session. Most people will file this under another bad Intel quarter. But this is literally the first big company admitting that the AI boom is now eating the rest of tech. Here's what happened: Intel guided fourth-quarter revenue to $15 billion to $16 billion, against about $17 billion expected, and said its gross margin would drop to 37% from 43%. CEO Lip-Bu Tan gave the reason himself: Customers started signaling a PC slowdown because of memory prices and availability. That memory is going into AI data centers. AI servers need enormous amounts of it, and the AI labs locked it up early. In October 2025, Samsung and SK Hynix agreed preliminary deals to supply OpenAI's Stargate with up to 900,000 memory wafers a month, which Tom's Hardware estimated at about 40% of global output. SK Hynix now says it has long-term deals with just 14 buyers covering more than half of its 2027 and 2028 output, and it expects the shortage to last until at least 2030. So everyone else pays more for whatever is left: - Gartner expects memory and storage prices to rise 130% by the end of this year, and PC prices to rise 17%. - Global PC shipments fell 21% in the third quarter, according to Counterpoint. - The WSJ says flash memory prices have roughly quintupled, and Apple's new touchscreen MacBook Pro launched at $2,899, which is $400 above the current M5 Pro model. Meanwhile the memory makers are printing money. SK Hynix's operating profit jumped 632% last quarter to about $62.8 billion. Samsung's rose 783% to about $80.5 billion. That's the transfer happening underneath the headlines. Money is moving out of the companies that build computers and phones and into the companies that make the chips AI is buying up. And Intel was simply the first to say it on an earnings call. Apple reports Monday after the bell, in its first quarter under John Ternus. Apple is one of the biggest memory buyers on Earth, so its gross margin will show whether the AI memory grab has reached the most profitable hardware company in the world. AMD reports on Tuesday. SK Hynix's own forecast runs to 2030. That's a long time for every PC maker, phone maker, and console maker to keep paying more for the same chips. If I owned anything that sells devices, I'd check one number in every earnings report this month: Gross margin. Memory is the cost the AI build-out is pushing up, and gross margin is where it shows first. Intel is the first big company to blame its numbers on AI's appetite for memory, and I doubt it'll be the last.

AI just crashed Intel, and Intel's own CEO said so. On Friday, Intel had its worst day since 1974. The stock fell 28.58%, and the US government's 10% stake lost about $12 billion in a single session. Most people will file this under another bad Intel quarter. But this is literally the first big company admitting that the AI boom is now eating the rest of tech. Here's what happened: Intel guided fourth-quarter revenue to $15 billion to $16 billion, against about $17 billion expected, and said its gross margin would drop to 37% from 43%. CEO Lip-Bu Tan gave the reason himself: Customers started signaling a PC slowdown because of memory prices and availability. That memory is going into AI data centers. AI servers need enormous amounts of it, and the AI labs locked it up early. In October 2025, Samsung and SK Hynix agreed preliminary deals to supply OpenAI's Stargate with up to 900,000 memory wafers a month, which Tom's Hardware estimated at about 40% of global output. SK Hynix now says it has long-term deals with just 14 buyers covering more than half of its 2027 and 2028 output, and it expects the shortage to last until at least 2030. So everyone else pays more for whatever is left: - Gartner expects memory and storage prices to rise 130% by the end of this year, and PC prices to rise 17%. - Global PC shipments fell 21% in the third quarter, according to Counterpoint. - The WSJ says flash memory prices have roughly quintupled, and Apple's new touchscreen MacBook Pro launched at $2,899, which is $400 above the current M5 Pro model. Meanwhile the memory makers are printing money. SK Hynix's operating profit jumped 632% last quarter to about $62.8 billion. Samsung's rose 783% to about $80.5 billion. That's the transfer happening underneath the headlines. Money is moving out of the companies that build computers and phones and into the companies that make the chips AI is buying up. And Intel was simply the first to say it on an earnings call. Apple reports Monday after the bell, in its first quarter under John Ternus. Apple is one of the biggest memory buyers on Earth, so its gross margin will show whether the AI memory grab has reached the most profitable hardware company in the world. AMD reports on Tuesday. SK Hynix's own forecast runs to 2030. That's a long time for every PC maker, phone maker, and console maker to keep paying more for the same chips. If I owned anything that sells devices, I'd check one number in every earnings report this month: Gross margin. Memory is the cost the AI build-out is pushing up, and gross margin is where it shows first. Intel is the first big company to blame its numbers on AI's appetite for memory, and I doubt it'll be the last.

OpenAI plans to burn $278 billion by the end of 2030, and this week it couldn't even lock in $20 billion. It spends far more than it earns, so the gap gets paid by the next investor in line. And this week, that line got A LOT shorter. OpenAI went into the week asking for $30 billion at a $1.4 trillion valuation, with a Friday deadline. By Friday it had cut the ask to $20 billion at $1.25 trillion. And the FT says the first close is expected to come in below even that, because some investors cut what they were willing to put in. Now put that next to OpenAI's own numbers: - In the second quarter, OpenAI brought in $6.7 billion and posted a $12.3 billion operating loss, according to the WSJ. That's almost $2 lost for every $1 it earned. - In September, the FT reported a company presentation showing OpenAI expects $278 billion of negative free cash flow from 2026 through 2030. - By my math, that's about $55 billion a year. A $20 billion round covers roughly 4 months of it. The mechanism is simple but terrifying: OpenAI has signed more than $1.4 trillion in compute deals with Oracle, Microsoft, Amazon, and its Stargate partners. Those companies count those deals in their backlogs, and Wall Street prices their stocks on it. But OpenAI can only pay those bills with money investors keep handing it. So the whole AI build-out sits on one assumption: OpenAI can always raise the next round. In June, OpenAI's CFO, Sarah Friar, called an IPO just "another way to fundraise", and said the $122 billion from March was raised to give OpenAI maximum flexibility. She said fundraising would be a key part of building a lasting company. But this week, the fundraising came up a third short. Microsoft's own filing shows how tight this already is. Microsoft booked $10 billion of revenue from OpenAI last quarter, about 11% of its total. In the same 3 months, the bill OpenAI still owes Microsoft grew from $6.0 billion to $7.2 billion. And the timing couldn't be worse. On Friday the Nasdaq had its worst day since June, and it's now 7.8% below its October 6 record. The investors who trimmed their OpenAI checks are watching the AI stocks they already own fall at the same time. The FT also says OpenAI wanted this round closed before Anthropic lists in mid-November, because the two are chasing the same big investors. OpenAI's defenders say it raised $122 billion in March, so it isn't running out of cash next month. Its annualized revenue was about $50 billion at the end of September. And $1.25 trillion is still 47% above the $852 billion it was valued at in March. But a markup that shrinks in a single week tells me the private market just started asking the questions the public market asks every day. Any public company booking OpenAI contracts as future revenue should tell its shareholders how much of that backlog depends on OpenAI raising more money. And OpenAI should publish audited financials before another dollar of retail money flows in, because $3 billion of its March round came from individual investors through bank channels. OpenAI asked for $30 billion this week and investors came back with at least a third less. It's close...

OpenAI plans to burn $278 billion by the end of 2030, and this week it couldn't even lock in $20 billion. It spends far more than it earns, so the gap gets paid by the next investor in line. And this week, that line got A LOT shorter. OpenAI went into the week asking for $30 billion at a $1.4 trillion valuation, with a Friday deadline. By Friday it had cut the ask to $20 billion at $1.25 trillion. And the FT says the first close is expected to come in below even that, because some investors cut what they were willing to put in. Now put that next to OpenAI's own numbers: - In the second quarter, OpenAI brought in $6.7 billion and posted a $12.3 billion operating loss, according to the WSJ. That's almost $2 lost for every $1 it earned. - In September, the FT reported a company presentation showing OpenAI expects $278 billion of negative free cash flow from 2026 through 2030. - By my math, that's about $55 billion a year. A $20 billion round covers roughly 4 months of it. The mechanism is simple but terrifying: OpenAI has signed more than $1.4 trillion in compute deals with Oracle, Microsoft, Amazon, and its Stargate partners. Those companies count those deals in their backlogs, and Wall Street prices their stocks on it. But OpenAI can only pay those bills with money investors keep handing it. So the whole AI build-out sits on one assumption: OpenAI can always raise the next round. In June, OpenAI's CFO, Sarah Friar, called an IPO just "another way to fundraise", and said the $122 billion from March was raised to give OpenAI maximum flexibility. She said fundraising would be a key part of building a lasting company. But this week, the fundraising came up a third short. Microsoft's own filing shows how tight this already is. Microsoft booked $10 billion of revenue from OpenAI last quarter, about 11% of its total. In the same 3 months, the bill OpenAI still owes Microsoft grew from $6.0 billion to $7.2 billion. And the timing couldn't be worse. On Friday the Nasdaq had its worst day since June, and it's now 7.8% below its October 6 record. The investors who trimmed their OpenAI checks are watching the AI stocks they already own fall at the same time. The FT also says OpenAI wanted this round closed before Anthropic lists in mid-November, because the two are chasing the same big investors. OpenAI's defenders say it raised $122 billion in March, so it isn't running out of cash next month. Its annualized revenue was about $50 billion at the end of September. And $1.25 trillion is still 47% above the $852 billion it was valued at in March. But a markup that shrinks in a single week tells me the private market just started asking the questions the public market asks every day. Any public company booking OpenAI contracts as future revenue should tell its shareholders how much of that backlog depends on OpenAI raising more money. And OpenAI should publish audited financials before another dollar of retail money flows in, because $3 billion of its March round came from individual investors through bank channels. OpenAI asked for $30 billion this week and investors came back with at least a third less. It's close...