Come now, Jensen
Welcome to Cautious Optimism, a newsletter on tech, business and power.
Thursday. Apple intends to use a Google AI model to power Siri, at least until it can get its home-grown tech up to snuff. Apple will therefore depend on Google for search and AI, making it a hardware company that increasingly feels like a Trojan Horse for Mountain View code. Which is funny, because the two companies are also arch-rivals in mobile operating system market share. Alphabet is betting that selling AI to Apple is better than its mobile rival buying from another company. To which, sure, but it’s still an odd bedfellows moment. To work! — Alex
📈 Trending Up: Sierra, heading towards $100M ARR … layoffs … union drama … no shit … bipartisan concern about AI job loss? … no shit … venture capital in Africa …Chinese economic influence? … Nexperia drama … Figma, after earnings … diffusion models … tarrifs on Russia …
Anthropic’s valuation, as Google discusses a new round and a $350 billion valuation for the AI giant. Which competes with Google, and uses a lot of Amazon cloud. And Google cloud. And has raised money from both hyperscalers. Well, that’s one way to avoid paying your own capex!
📉 Trending Down: Fintech regulation, after a truckload of Andreessen cash …Doordash, after earnings … Duolingo, after earnings … lower-decile consumer spending … Microsoft-OpenAI dependencies … self-driving stocks … launch videos …
Things That Matter
Tariffs are probably cooked: Arguing before the Supreme Court yesterday, the Trump administration contended that tariffs enacted by the President via executive fiat were legal. Most justices appeared less than impressed with the White House’s reasoning. Regarding whether a president can invoke the International Emergency Economic Powers Act (IEEPA) to enact tariffs, even some of the more administration-friendly justices were skeptical. As a result, folks are betting that SCOTUS will unwind at least a portion of the present-day tariff regime.
A loss before the Supreme Court will not halt the POTUS tariff barrage; Treasury Secretary Scott Bessent made it clear that the White House believes it has other mechanisms it can employ. Which will, naturally, create more legal disputes.
Still, it’s encouraging from a free-trade perspective that in the United States, there will be some guardrails on how much a single person can upset global commerce.
Which is good news for the economy: Tariffs are harming e.l.f Beauty’s gross margins, Pinterest’s revenue growth, and profitability at Axon. How about freer trade to allow for national competitive advantage to shine?
IPOs are probably cooked: As October closed, Morgan Stanley bought EquityZen. The transaction got less attention than it deserved; here was an investment bank, a player in the IPO game, buying a secondary marketplace for private-company shares. The transaction was a blinking light that banks are betting that the IPO dearth is set to continue for a very, very long time.
And then today, Charles Schwab announced plans to buy Forge Global, a competitor to EquityZen. Secondary marketplaces allow for the sale of shares in startups. But as secondary sales have become an increasingly key way for venture firms to return cash to their own backers. So sure, it’s great that we’re seeing investment into the pathways that early employees can take to turn some of their shares into cash when executive leadership decides to simply never list. But ever-more sophisticated and institutionalized secondary activity will also help the LP→GP→Startup→GP→LP capital recycling flow, too.
Why is this a good idea: Apple and Google are not the only folks cooking up a major AI deal. But instead of the AI provider (Google) getting paid by the AI user (Apple), Perplexity (AI provider) intends to pay the AI user (Snap) for access to its customer base?
Starting in early 2026 […] Perplexity’s AI-powered answer engine will let Snapchatters ask questions and get clear, conversational answers drawn from verifiable sources […] Perplexity will pay Snap $400 million over one year, through a combination of cash and equity, as we achieve global rollout.
That’s a lot of spend for a company that recently raised $200 million at $20 billion valuation. Framing the deal from a different perspective, Perplexity is set to pay about twice its current ARR (“approaching $200 million” in September 2025) to a partner for distribution next year. And that doesn’t count the cost of supporting all those Snap users.
Perplexity has plans to yank revenue out of the deal, of course. It’s not suicidal.
The AI market is becoming calcified. OpenAI has snapped up huge consumer market share and is touting rising business adoption. Anthropic has a good chunk of the enterprise AI market sewn up. Google has its own distribution channels (browser, mobile OS, etc) for its AI efforts. Microsoft is still mostly an OpenAI shop but also delivers AI through its existing products — so where does that leave Perplexity? Coughing up for distribution that many of its don’t have to pay for, or already earned.
Meanwhile, Perplexity is being sued by Amazon for sending its user AI agents as normal Chrome instances; the ecommerce giant alleges that Perplexity is dodging its attempts to block the bots, breaking the CFAA in the process. Cloudflare has also complained that Perplexity is not following the rules of the Internet, among other criticisms. (Perplexity’s response to Amazon can be read here.)
How about no: OpenAI CFO Sarah Friar argued recently that guarantees for its investments in data center capacity — from “an ecosystem of banks, private equity, maybe even governmental, the ways governments can come to bear” — would lower its borrowing costs, and allow it to spend even more.
Should the government backstop OpenAI’s infra spend? No.
Should other private companies handle the work of non-public finance? Yes
I suppose this is why AI lobbying is soaring.
What say you, Jensen?
In terms of surprising headlines, the FT reporting that Nvidia CEO Jensen Huang expects the United States to lose the AI race was a banger. Here’s the key excerpt from the financial paper:
In the starkest comments yet from the head of the world’s most valuable company, Huang told the FT: “China is going to win the AI race.” […]
The Nvidia chief said that the west, including the US and UK, was being held back by “cynicism”. “We need more optimism,” Huang said on Wednesday on the sidelines of the Financial Times’ Future of AI Summit.
Huang singled out new rules on AI by US states that could result in “50 new regulations”. He contrasted that approach with Chinese energy subsidies that made it more affordable for local tech companies to run Chinese alternatives to Nvidia’s AI chips. “Power is free,” he said.
Jensen makes three points. First, that the United States’s AI push is being held back by cynicism. Second, that China’s subsidies of its domestic AI industry is an edge. And third, that the United States is set to enact a confusing web of state-level laws that impede AI progress at home. As a result of the three intersecting issues, we’re going to lose the AI race.
I’m not so skeptical. But let’s take the points seriously:
Is the United States AI-skeptical? No. Domestic investment in data centers is world-leading — US data center market share is peerless — just as capital inflows for American AI startups are now worth nearly two-thirds of total domestic venture deal value. That means US-based venture investment into AI was worth north of $50 billion in Q3 2025 (leaning on a second data set), more than all other startups in the world raised from VCs for all types of company building.
Is China subsidizing domestic AI? Yep. Sticking strictly to recent news, China is moving to provide cut-rate power to domestic data centers that use homegrown chips. This is not merely a boon; China is crimping the ability of its local companies to purchase US GPUs. But local chips consume more power. So, the state is making up the difference. Perhaps it will prove a net-win for Chinese companies to use dated chips with free juice, but I doubt that any cutting-edge company wants to work with silicon that is less than SOTA.
Are state-based AI regulations coming? No, they are already here. Florida, Texas, California, and other states have put laws on the books. So far? AI investment continues to soar here at home. More states will put laws into effect. Our expectation? That AI investment will continue to soar at home.
In a perfect world, Congress would come up with a set of AI rules for the nation so perfect that no state would feel the need to add its own strictures to the mix. The chance of that is zero. So, states will get a say.
This is not a failure of democracy, but a feature. If you want to avoid the messy democratic muddle that is our form of government, you can do so. China’s system is very different, very centrally controlled, and able to set national policy with greater rigor and speed. All it costs is freedom of speech, religion, and assembly, along with property rights.
Your call!
Jensen’s comments are a great underlining of our argument yesteday that the vibes are bad. Mr $5 Trillion is annoyed that democracy is going to debate. Jensen, take a weekend off. Read a novel. Drink some really good coffee. Things are going to be fine! And the open, free, democratic model will win. As it always has.