Nativism is going to wreck the tech economy
Welcome to Cautious Optimism, a newsletter on tech, business and power.
📈 Trending Up: Chip restrictions … Anthropic … Sure, Jan … Samsung profit margins … neolabs? … flight pain … you don’t say … Blue Origin … IPOs in India … trendy tequilla …
📉 Trending Down: Baidu, after Ernie 5.0 … StubHub … AI-free browsing … AI politics in the White House … climate progress in Europe …
OpenAI, the next social giant?
In late October, we dug into the awe-inspiring growth that Anthropic reported in 2025. From a single billion dollars’ worth of annualized revenue to start the year, Anthropic could reach a staggering $9 billion run rate by the time we drop the ball and kickstart 2026.
Anthropic’s ripping growth proves there is more than one way to build a foundation AI model company. While OpenAI is the consumer Kleenex of AI services today — evidenced by its 800 million weekly active users of ChatGPT — Anthropic is generating huge incomes by selling its AI products to other companies. OpenAI sells to companies, too, and Anthropic has prosumer users, but the two companies are fundamentally different.
With that context in mind, I present you with the following headline:
Mix in the stand-alone Sora app and it’s pretty clear that OpenAI is betting its future on being the leading consumer AI company in the world, requiring it to monetize enough regular users to fund the inference cost of freeloaders. (Hence its ChatGPT Go product in India, for example.)
I, for one, cannot wait to see what idiotic, hilarious, and degenerate nonsense my group chats generate with ChatGPT in the sidecar. You’re welcome in advance for the training data, Sam. And the jokes.
More recently, CO noted that Anthropic was running an asset-light business, leaning on hyperscalers to handle its compute needs. Not so any more — Anthropic announced a “$50 billion investment in American computing infrastructure, building data centers with Fluidstack in Texas and New York” earlier this week.
AI, jobs, visas, and closed borders
There’s a split in the President’s coalition that could eventually divide the modern Republican party in two. On one side, the technologist-right and its long-held belief that high-skill immigration is critical to our domestic economy and national technology leadership. On the other, complaint from the nativist-right that H-1B visas are mere instruments of corporate abuse, allowing large companies to import lower-cost talent over which they have more sway than American workers, depriving citizens from well-paying gigs.
No matter which side of this particular debate you sit on, I think we’re heading for a national argument about labor:
The labor market is softening: While we lack official data, surveys from private firms indicate that the US labor market in October was poor to the point of being net-negative. Concerns about a K-shaped recovery — one in which the wealthy do well and the lower classes get annihilated — are partially driven by this economic weakness.
AI is likely harming early-career employment: New grad hiring in the spring is expected to be worse than what we saw this year. That means more educated, indebted students staring down the barrel of rejection letters and living at home than we endured in 2025. There are a host of factors impacting new grad hiring and the labor market in general, but I think it’s fair to say that entry-level white collar jobs are the warehouse gigs of blue-collar-land; canaries in the automation coalmine.
Tech companies are lowering domestic labor demand: Tech companies racing to build ever-more-intelligent AI are partially to blame for shifts in the labor market that could push unemployment rates higher for a time. Add in the fact that the same tech companies are laying off staff at home, and it’s all a bit dark.
Firing staff while replacing human labor and you still need H-1B visas? Thus said VP Vance in July. From a surface-level perspective, he has a point.
Despite POTUS’s occasional huarrumphs in favor of some immigration — recall the green card fiasco — highlighting the need to bring in some high-skill workers, vibes amongst American conservative political leaders and pundits are incredibly negative when it comes to letting folks in. No matter their skill. Even at the university level.
This is in direct contrast with how the tech economy has long functioned. Tech companies in the US have a history of being built by immigrants. Many of our leading CEOs today are immigrants, too, with names like Nadella and Pichai springing to mind. And yet.
POTUS it term-limited, and the VP wants to run in his stead in the 2028 elections. Will tech folks who favor immigration for talented folks stick with the Vice President if he stays critical of tech companies and their labor demands?
The answer is probably yes, but it does increasingly appear that we’re looking at a future in which at least one of the two major political parties of the United States wants to zero out immigration. And that will mean a lot more offshore founding and hiring. A bummer, in my view.
No one can stop AI, tied up as it is in national security. That means we’re going to get an AI-ified economy regardless of what your reservations may be. Halting our ability to brain-drain the world at the same time feels self-defeating.
Are we in a risk-off moment? Maybe not
Tech stocks took a whacking this week as investors cooled on the AI trade. The consumer looks weak. China’s economy is not performing at peak vitality. So, are we in entering risk-aversion territory? I don’t think so. Not yet, at least in the startup game.
Two data points:
Cursor just raised $2.3 billion, pushing its valuation to nearly $30 billion.
Thining Machines Lab is looking to raise fresh capital at a valuation that could surpass $50 billion.
The Cursor story is interesting, if understandable. The AI IDE provider announced, as part of its funding news, that it has “crossed $1B in annualized revenue, counting millions of developers and many of the world’s most accomplished engineering organizations as our customers.”
If you are not impressed by that result — a doubling of its Summer-era revenue scale — I bet I know why. You want to know about Cursor’s margins. After all, the company reportedly had negative gross margins earlier this year, likely driving its pricing change that caused many of its customers to kick up sand.
Cursor could have raised a bunch of money without getting its profitability house in order; it’s growing so quickly that it can afford to put off certain traditional business tasks. But with the launch of its own, in-house AI model in late October is probably part of the company’s plan to stop paying Anthropic and OpenAI’s own gross margins; if it can get enough of its users to prefer its homemade Composer model and it’s cheaper to run, then the company’s economics may flip into the black.
Overall profitability is a different challenge, of course, but with fewer than 400 staff today, Cursor is hardly bloated. Still, the company is not entirely derisked in economic terms and faces bruising competition from rival startups, AI model companies, and hyperscalers alike. That investors just affixed a $29.3 billion price tag to the startup is indicative of persistent risk appetite amongst private-market investors.
Just as the Thinking Machines Lab round that is forming is a signal that risk tolerance remains high amongst the startup backing set.
The vibes are off, and they're deteriorating overall. But for AI hotspots, the sky is still full of sunshine and rainbows. Perhaps the K-shaped economy will also apply to tech companies, too?