The Aftershock: Tech’s Reaction to a Blue Wave 11/5/26

Anthropic

Across the shows: 102 speakers in other episodes, 130 bullish and 85 bearish calls, past 30 days

• Jessica Lessin · The InformationBearish

  Anthropic's rising compute costs are a bad sign for its margins ahead of its expected IPO.

  Lessin notes that Anthropic is a big buyer of compute and that its numbers are good enough to go public on its own timetable. But margins are under heavy scrutiny, and she does not know whether the compute squeeze will slow its growth. She concludes that the cost of compute going out for Anthropic is not a good sign.

  “And so cost of compute going out for Anthropic, not a good sign.”at 19:47

• Kai-Fu Lee · 01.AIBearish

  Anthropic's plan for democracies to set AI rules that China must accept is hubristic and won't work.

  Lee says the US should extend an olive branch and negotiate with China, since both see the dangers of AI. He calls Anthropic's proposal, that democratic countries make the rules and force China to accept them, a ridiculous idea based on hubris. He says China, or any country that is number two, will not accept such a deprecating premise, so the approach won't work. He describes China as a rational government acting in its own interest, and says it is as important for China to compete in the AI race as for the US.

  “What Anthropic is proposing, let the democratic country make the rules and force China to accept, is a ridiculous... idea based on hubris. Not gonna work.”at 40:43

NVIDIA

Across the shows: 73 speakers in other episodes, 68 bullish and 26 bearish calls, past 30 days

• Jessica Lessin · The InformationBearish

  Nvidia is vulnerable as a supplier, since harder compute access threatens demand and its valuation.

  Lessin says that if you are a supplier, the post-election environment is tough. Named examples are Nvidia and the neoclouds. The point is that if a company's existing business is already baked into its valuation, a slowdown in new compute build-out hurts it. Her case rests on the supply side of the compute squeeze she describes, and she gives no numbers.

  “But if you're a supplier, for example... Nvidia, tough. Obviously, the Neo Cloud's tough. If sort of all your existing business is baked into your valuation, tough.”at 18:57

OpenAI

Across the shows: 46 speakers in other episodes, 29 bullish and 39 bearish calls, past 30 days

• Jessica Lessin · The InformationUncertain

  OpenAI's enterprise sales recovery could support a solid showing, if it continues and margins improve.

  Lessin says OpenAI had a tough start to the year but turned it around as enterprise sales grew. If that continues, she thinks it will be in a good place for a solid showing. She is more cautious on margins: OpenAI has historically favored scale over margins. She also says its recent round was driven more by the new valuation, which funds acquisitions, than by a need for cash, and that there is clear nervousness about the company.

  “OpenAI had a really tough beginning of the year, and they did turn it around as their enterprise sales grew. If that continues, I think they'll be in a good place to have a solid showing.”at 25:01

AI infrastructure & data centers

• Jessica Lessin · The Information

  Post-election, scarce compute favors firms that already hold it, boosting their margins and value.

  Lessin argues that the midterm outcome handed big tech a gift. Democratic control of Congress makes it harder to get new compute in the short term, so anyone who already has compute becomes more valuable and enjoys better margins. The argument is a mechanism: scarcity of new capacity raises the value of existing capacity.

  “I think the voters handed at least the big tech companies a gift by basically saying in the short term, it's gonna be harder to get new compute. If you have it, it's way more valuable,”at 18:21

• Jessica Lessin · The Information

  Data center projects will still get approved, with big tech paying cities and towns more.

  Lessin argues that the industry is now big enough that its projects will keep moving forward. Companies will pay more to cities and towns, and firms that already have agreements with local authorities will be able to scale existing projects more easily. She admits the current political rhetoric is hostile, and that changing it would take a big shift.

  “Are these projects going to still get approved? Yes. Are the big tech companies going to pay more, pay more to the cities, pay more to the towns? Yes.”at 20:19

US–China AI & chip race

• Kai-Fu Lee · 01.AI

  The US leads China in AI by six to twelve months, a lead most companies can wait out or accept.

  Lee says the US currently leads by six to twelve months in any given scenario. He argues that for most companies, a lead of that size is either fine to wait for or fine to accept as a deficit. The argument is that a short lead does not decide the race, since most companies can operate with a lag.

  “So I think US now leads six to twelve months in any given scenario, but I would say that for most companies, that six to twelve month lead is either okay to wait for or okay to accept the deficit.”at 32:41

• Kai-Fu Lee · 01.AI

  US leads in AI breakthroughs by six to twelve months, while China follows closely at a tenth of the price.

  Lee says the US clearly leads in AI breakthroughs, and China is a very capable strong follower that delivers similar capability at about one-tenth the price. He puts the US lead at six to twelve months in any given scenario. For most companies, he argues, that gap is either acceptable to wait out or acceptable to accept as a deficit.

  “There's no doubt that US leads in the AI breakthroughs, and China is very clever in being a strong follower at one tenth the price.”at 39:10

Enterprise AI adoption & AI software

• Kai-Fu Lee · 01.AI

  Most CEOs misuse AI by chasing headcount cuts, when its real value is as a top strategic adviser.

  Lee says he sees many CEOs talking about AI transformation, and that 99% of them are doing it wrong. They look for cost-cutting use cases in customer service, legal and finance, measured by agent counts and headcount reductions. Lee argues AI matters more as an IQ-300 employee who helps run the company, so it should not be deployed only on routine tasks.

  “I see so many CEOs every year, and they all talk about AI transformation and 99% of them are doing it all wrong. ... They're looking for numbers. I have thousand agents and I'm saving. I've cut 200 headcount. But that is not how AI matters.”at 35:21

• Kai-Fu Lee · 01.AI

  A company-specific harness on an on-premise open-weight model beats frontier models for CEO-level AI transformation.

  Lee recommends fully on-premise open-source models for enterprise deployments, so no company data goes to Anthropic or OpenAI or to China. For general questions, frontier models are better. For helping a CEO run AI transformation, his approach is better. He concedes open-weight models are six to twelve months behind the best closed models, but says a harness adds two things no model has: knowledge of the company's confidential details (org chart, decision-making, workflows), and a reasoning process that makes answers explainable so hallucinations can be spotted.

  “The open weight model is always six-twelve months behind the best closed weight model, but our approach of adding the harness adds two additional knowledge that any model does not have and cannot have.”at 38:16

Open-source AI models

• Kai-Fu Lee · 01.AI

  Closed and open AI models can each have a leader, like iPhone and Android.

  Lee uses the smartphone analogy: the closed model is like the iPhone and the open model is like Android. He argues there is room for everyone to play, and that a leader can emerge in each approach. The claim is that the AI race is not winner-take-all, so open and closed models can both succeed.

  “There's room for everyone to play. And, also, closed model is like iPhone. ... Open model is like Android. There is room for a leader in each place.”at 2:17

AI regulation & safety

• Jessica Lessin · The Information

  Democrats' anti-AI and anti-data-center rhetoric is a losing position that will not last.

  Lessin says Democrats who keep talking negatively about AI and data centers are in a losing camp, and that no one can stay in it. She argues that voters will move on. The framing is that opposition is popular now but cannot hold, and she says the tech companies must also prove that AI changes people's lives in positive ways.

  “I think that's a losing... least unpopular, least popular... There's no way you can stay in that camp.”at 20:43

Antitrust & tech regulation

• Jessica Lessin · The Information

  Sweeping tech regulation is unlikely, but Democratic probes could be tough for big tech legislatively.

  Lessin says no one expects sweeping regulation of any kind. She argues that if Democrats carry hearings and probes into the next administration, the legislative environment could become very tough for these companies. She points to the Biden-Musk relationship as a precedent for how such scrutiny can shift the political landscape.

  “no one thinks that there should be any sweeping regulation of any sort. But if the democrats continue this into the next administration... there's a lot legislatively that could be very tough for these companies.”at 23:20

Frontier AI labs & model race

• Kai-Fu Lee · 01.AI

  No winner-take-all: OpenAI and Anthropic's six-to-twelve-month leads will be matched, leaving room for all.

  Lee rejects the idea that a frontier model reaching recursive, AGI-like capability would give its maker a hockey-stick advantage and own the whole space. Even if OpenAI and Anthropic are six to twelve months ahead of everyone else, including Google, Meta and Chinese companies, the others will catch up in six to twelve months. He uses an analogy: closed models are like the iPhone and open models are like Android, so there is room for everyone to play and a leader in each category.

  “I absolutely do not believe that. If... OpenAI and Anthropic is six to twelve months up... Ahead of everyone else, including Google, Meta, and the Chinese companies. They'll catch up in six to twelve months. So there's room for everyone to play.”at 41:23