Decoding the business of technology.
examnity.

The Download: the future of chipmaking and Anthropic’s government clash

Two stories this week crystallize the fault lines running through tech's most consequential sectors: the hardware that makes AI possible, and the policy scrambles trying to contain it. ASML's lithography dominance faces its first real challengers.

Grace Linwood, Silicon Valley Culture & Venture Chronicler · updated June 23, 2026

The Download: the future of chipmaking and Anthropic’s government clash

The Download: the future of chipmaking and Anthropic's government clash

Anthropic just learned what happens when you flag your own model as a cybersecurity risk and Washington responds before you can blink.

The $400 million machine everyone depends on—and everyone wants to replace

ASML's newest EUV lithography system weighs over 150 tons, costs $400 million, and stands roughly the size of a double-decker bus. You climb to its top to see where chip futures get written. The machine patterns features using extreme-ultraviolet radiation—light outside the visible spectrum, generated by blasting lasers at molten tin droplets tens of thousands of times per second.

ASML commands roughly 90% of all chip-lithography tools globally. That dominance has become a geopolitical chess piece and a business vulnerability. Would-be competitors are now gunning for its territory, sensing that any single point of failure in the AI supply chain invites disruption. For founders building on advanced silicon, the question isn't theoretical: what happens when your chip supply bottleneck has a single Dutch gatekeeper?

Anthropic's self-imposed wound

In April, Anthropic disclosed that Mythos—a coding model it built—could pose a cybersecurity risk. The company released Fable, a safer variant. Days later, the US government slapped export controls on Fable. Within hours, Anthropic pulled access to both models entirely.

What's striking here is the sequence: Anthropic raised its own hand, flagged concern, and still got blindsided by a policy response that looks reactive rather than coordinated. This wasn't a rogue bioweapon scenario. It was a coding tool. The intervention signals that Washington is willing to move fast on AI controls—even when the threat profile doesn't match the "doomer" narrative that's dominated safety discourse.

What founders should actually watch

These aren't isolated headlines. They're connected threads pulling at the same sweater.

For hardware builders: ASML's dominance is real but not permanent. The $400 million price tag and concentrated supply chain mean any challenger with credible EUV alternatives—whether from Japan, China, or a well-funded startup—could reshape procurement strategies within a decade. Track who's investing in next-gen lithography outside the Netherlands.

For AI companies: Anthropic's experience is a warning shot. Self-reporting model risks doesn't guarantee a predictable regulatory response. If the government can move faster than a company can course-correct, the incentive structure around transparency gets complicated fast. Founders need to think about export control exposure before they ship, not after.

The common thread? Both stories reveal how infrastructure bottlenecks—whether physical or regulatory—can hemorrhage a company's momentum overnight. The founders who survive this era won't just build great products. They'll build resilient supply chains and regulatory playbooks before they need them.