The Meltdown — March 15-21
Trillion-dollar projections, a co-founder in handcuffs, and the enterprise AI scoreboard flipping overnight.
01 — Jensen Huang sees $1 trillion in AI chip orders
NVIDIA's GTC keynote put a number on the AI hardware boom: $1 trillion in combined Blackwell and Vera Rubin purchase orders through 2027, double last year's projection. Vera Rubin promises 10x more performance per watt than Grace Blackwell, and Microsoft has already powered on the first system in its labs. Huang also unveiled the Groq 3 LPU from NVIDIA's $20 billion licensing and talent deal, showed a Kyber rack prototype, and declared OpenClaw "the next ChatGPT" and "the largest open-source project in the history of humanity." The agentic AI theme dominated: NemoClaw for enterprises, autonomous driving with Uber planned across 28 cities by 2028, and enough robotics partnerships to fill a separate newsletter.
02 — Anthropic flipped the enterprise scoreboard
Anthropic now captures 73% of all spending among companies buying AI tools for the first time, according to Ramp data, a swing from 60/40 in OpenAI's favor just weeks ago. Claude Code alone is running at $2.5 billion in annualized revenue, and the company now has more than 500 clients spending over a million dollars a year. Total annualized revenue hit $19 billion against OpenAI's $25 billion, and Anthropic's growth rate has been roughly double OpenAI's, which puts a mid-2026 crossover in play. The shift seems less about model quality and more about product: Claude Code fits developer workflows in a way ChatGPT Enterprise hasn't matched.
03 — OpenAI bought the Python toolchain
OpenAI is acquiring Astral, the company behind uv, Ruff, and ty, three tools that have quietly become foundational to modern Python development. Millions of developers use them daily, and the Astral team will fold into Codex, which has hit 2 million weekly active users with 5x usage growth since January. The tools stay open source, but the developer community is justifiably nervous about what "deeper integrations" between a package manager and an AI coding agent actually means in practice.
04 — The White House wants to override 27 states on AI
The Trump administration released a national AI legislative framework urging Congress to preempt state AI laws it views as too burdensome. The four-page document proposes no new federal regulatory body, sector-specific rules through existing agencies, regulatory "sandboxes" with waivers lasting up to 10 years, and shifts the child safety burden to parents rather than platforms. Democrats introduced legislation to repeal the underlying executive order. There are dozens of AI bills alive across 27 states right now, and whether this framework survives Congress or not, the signal is clear: the federal government wants one set of AI rules, and it wants them loose.
05 — A Supermicro co-founder was arrested for chip smuggling
Federal agents arrested Supermicro co-founder Wally Liaw for allegedly diverting $2.5 billion in Nvidia GPU-packed servers to China through Southeast Asian intermediaries, violating U.S. export controls. The stock lost a third of its value in a single session. Chip restrictions keep getting tighter, but so far the main effect has been pushing supply chains underground rather than cutting off access, and this case shows what that looks like at industrial scale.
06 — The UK backed down on AI copyright
The UK government dropped its plan to let AI companies train on copyrighted works without permission, after sustained pushback from creative industries. Elton John had called the government "absolute losers" over the plans. Dua Lipa, Thom Yorke, and Abba's Björn Ulvaeus weighed in alongside thousands of smaller artists arguing that opt-out was functionally no protection at all. The government now says it has "no preferred option," which in practice means the licensing market develops on its own terms.