The Great Nvidia Escape
Two of AI's biggest names spent the day making the same quiet bet: that the future shouldn't run entirely on someone else's silicon. Anthropic, the company behind the Claude models, has begun early-stage work on its own custom AI chip and held preliminary talks with Samsung Electronics about manufacturing it, according to The Information. It's a striking move for a firm best known for software — designing your own chip is expensive, slow, and hard, but it's also the surest way to stop depending on scarce, costly Nvidia GPUs. Turning to Samsung, rather than the usual foundry of choice, hints that Anthropic wants leverage as much as capacity.
They're not alone. Amazon confirmed it is building custom AI chips to power consumer hardware — Echo speakers, Fire TV, and, per CNET, a future generation of mobile gadgets. Devices chief Panos Panay confirmed the effort to CNBC. Amazon already designs data-center silicon; pushing AI chips into home gadgets is about controlling cost and performance on devices that ship by the millions.
The throughline: when compute is the bottleneck, the biggest players would rather own it than rent it.
Nvidia's Answer: Finance the Whole Thing
Nvidia isn't standing still. The company unveiled an optional financing model that lets AI startups get its coveted GPUs now and pay later — partly by handing back a slice of their future revenue. Coverage ranged from neutral (Tom's Hardware, AI: Reset to Zero framing it as expanding "revenue-share deals" and "computing power loans") to pointed: The Register called the datacenter-funding version a "double-dipping" scheme, noting Nvidia stands to profit both from selling the chips and from the revenue they generate. It's a clever way to keep demand flowing even as customers stretch for cash — and a reminder of how much of the AI boom now runs on creative financing.
Nvidia is building out its commercial muscle in other ways too, poaching Microsoft sales veteran Nick Parker after 26 years in a package reported by GeekWire at over $40M — a sign of how aggressively the chipmaker is scaling its go-to-market operation.
Foundries, Smuggling, and a Cooling Market
The supply side had its own drama. Taiwanese authorities are investigating server maker Supermicro over alleged violations tied to the smuggling of Nvidia AI chips, with the fallout already hitting its stock. Meanwhile, Intel's contract-manufacturing ambitions are facing what one report calls their first real test, as a broader chill settles over semiconductor stocks in 2026. Intel bet big on becoming a foundry-for-hire; a cooling market is an unforgiving place to prove it.
Quantum Keeps Raising
Beyond the AI-chip scramble, quantum computing kept its momentum. Qolab raised $54.2M in Series B financing and commitments led by UC Investments to chase superconducting quantum chips. And Vancouver's BTQ Technologies moved a step closer to acquiring French quantum firm QPerfect, per BetaKit — a small but telling sign that consolidation is beginning in a field still years from maturity.
The takeaway: the industry spent today trying to loosen Nvidia's grip — through custom silicon, creative financing, and bets on what comes next — even as Nvidia tightened it.