The US and a Dutch Chip Giant Disagree on Where the World’s Most Important Machine Is
The three biggest stories in AI today all trace back to the same question: who controls the building blocks of artificial intelligence? A missing machine that may or may not be in China, a $1.5 billion bet on AI infrastructure, and Snap quietly offloading a project it could no longer afford — together, they reveal just how much pressure the AI industry is under right now.
The US Government Says a Critical Chip Machine May Be in China. The Maker Says That’s Wrong.
EUV lithography — extreme ultraviolet lithography — is the process used to etch the tiniest, most powerful circuits onto computer chips. ASML, a Dutch company, makes the only machines in the world capable of doing this at the highest level. No EUV machine, no cutting-edge chip. That’s why this story matters so much.
The US Commerce Department has told ASML it believes one of these machines may have ended up in China, which would violate strict export restrictions designed to keep advanced chip technology out of Chinese hands. ASML has pushed back firmly, saying it has never shipped an EUV machine to China. That’s a direct factual conflict between a government and a major corporation — and neither side is backing down, as TechCrunch reports.
Think of an EUV machine like a master key. Whoever holds it can make the most advanced chips in the world. The US has spent years trying to ensure China doesn’t get that key. If one did make it across, it would be a significant hole in a carefully constructed wall.
For everyday people, this matters because the chips in your phone, your laptop, and the AI tools you use daily depend on a global supply chain that is increasingly becoming a geopolitical battleground. Disruptions here eventually reach your pocket.
Why this matters: If a machine did end up in China, it could accelerate Chinese chip development by years. Even the uncertainty is enough to rattle markets and trigger new restrictions.
“US concerns that ASML’s EUV chipmaking tool may be in China; ASML denies the shipment occurred.”
A Startup That Runs AI Models Just Raised $1.5 Billion — Months After Its Last Giant Round
AI inference — the process of actually running an AI model to generate a response — is expensive, technically demanding work. It’s the difference between building a car and operating a taxi fleet. Baseten does the operating.
Baseten is reportedly raising $1.5 billion at a valuation of $13 billion, according to TechCrunch. What makes this striking is the timing: the company raised a large round just months ago. Investors are clearly doubling down fast.
The reason is straightforward. Every company building an AI-powered product needs somewhere to run their models. Most don’t want to manage that complexity themselves — they’d rather pay someone else to handle it reliably and at scale. Baseten has positioned itself as that someone. As demand for AI products grows, demand for inference infrastructure grows with it.
For anyone using an AI-powered app at work or at home, companies like Baseten are part of why those tools actually respond quickly and don’t crash. The infrastructure is invisible, but without it, nothing works.
Why this matters: The AI gold rush isn’t just about the models themselves — it’s about the plumbing underneath. Investors are pouring money into that plumbing at a remarkable pace.
Snap Couldn’t Afford Its AI Video Team, So It Made Them Their Own Company
Generative AI video — software that creates video clips from text prompts or images — is one of the most resource-hungry areas of AI development. The computing costs alone can be staggering.
Snap has spun its internal AI video team out into a new, independent company called Dotmo. Snap retains a significant ownership stake, and Snap’s CTO Bobby Murphy is personally backing the venture. But the core reason for the split, as TechCrunch explains, is cost. Keeping the project inside Snap had simply become too expensive.
This is a creative corporate maneuver. Rather than shutting the team down, Snap handed them independence and a chance to raise their own funding. The technology stays alive. Snap keeps upside through its ownership stake without carrying the full financial burden.
For regular Snap users, nothing changes immediately. But Dotmo now has room to build products beyond Snapchat, which could eventually mean AI video tools showing up in places you wouldn’t expect.
Why this matters: Big tech companies are quietly offloading their most expensive AI experiments rather than killing them. This pattern is likely to repeat across the industry.
Also Happening in AI
On the developer side of things, Anthropic shipped two consecutive updates to its Python SDK — versions 0.110.0 and 0.111.0 — adding a code execution tool and better request labeling, making it easier for developers to build and track AI-powered applications. LangChain, a popular toolkit for connecting AI models to outside data and tools, released two maintenance updates as well, tightening up security and keeping its dependencies current. And in an unusual celebrity-tech crossover, Queer Eye life coach Karamo Brown launched a wellness app called Kē, built around an AI version of himself designed to offer guidance and support between sessions.
What to Watch
The ASML story is worth following closely — if US investigators find evidence that an EUV machine actually crossed into China, expect a swift and significant policy response that could reshape global chip supply chains. Meanwhile, watch how many other large tech companies follow Snap’s lead and spin out expensive AI projects rather than fund them internally. That trend, if it accelerates, tells you something important about which AI bets even the biggest players aren’t willing to carry alone.