Let us start with what is not in dispute. Everything we are being told about the coming economy — the data centres, the models, the automation, the entire architecture of the next decade — is possible only because of Nvidia chips. That is not a marketing line: the company shipped more than 90 percent of the GPU-equipped servers sold worldwide in the last quarter of 2024, on IDC's count. It is a fact of industrial geography, and Jensen Huang deserves the credit he gets for having built it.
And then came the lawsuit.
Shareholders are suing the man who multiplied their share price fifteen times in three years, because the stock fell five percent. Read that again. The same investors enriched by the AI story are now suing the author of the story over a single bad session. This is not ingratitude, and it is not greed in any ordinary sense. It is the clearest signal yet that the AI chip boom has changed phase.
The reason is valuation. At three trillion dollars — one and a half times the GDP of Italy — Nvidia is no longer a company whose price is set by what it sells. It is a claim on the future, and it got there by decupling in eighteen months and overtaking Apple as the most valuable company in the world. Someone who put ten thousand dollars in at the end of the last century would be worth sixty million today. Numbers like that do not describe a business. They describe an expectation, and expectations have owners who want them defended.
Here the dotcom comparison is instructive. After the internet bubble, two stocks told two stories: the company that made the switches and hardware, and the company that used those switches to sell books and everything else to ordinary people. The lesson is not that hardware is bad. The lesson is that the firm which sells the future and the firm which uses the future are priced on entirely different logic — and shareholders eventually discover, usually at the worst moment, which one they actually own.
On top of that sits politics. Nvidia's demand is not set in a market; it is set in a chancellery that decides how much of China — 13 percent of Nvidia's revenue in the year to January 2025 — the company is permitted to sell to. Washington bans exports to China and the stock drops seven percent. A later round of restrictions sends the shares down again just as they were recovering. Then a robust quarter arrives and the price still falls after hours, because from China come loud anti-Nvidia signals. Beijing finally answers the export bans by prohibiting imports of advanced chips and announcing that Chinese firms will use their own — which they have been building, Huawei's HiSilicon among others, exactly as the boomerang logic of sanctions predicted. Meanwhile, when Gulf states want to buy, we are invited to worry about what "autocratic" governments might do with the hardware. The customer list is vetted by politics and the supplier is lectured by politics. That is not a stable revenue base. It is a diplomatic variable with a ticker symbol.
Which brings us to governance, the real subject of that lawsuit. When a company's price depends on export licences, on a decision in Beijing and on the mood of a US administration, the shareholders' complaint is not really about one quarter's guidance. It is that they were sold a growth curve as though it were a law of nature. Making the shares rise fifteenfold is not a crime. But in a litigious market, success at that scale stops being a defence and becomes an exhibit.
And the American model of risk has already been settled: in the banking scare, where 94 percent of Silicon Valley Bank's deposits sat above the insurance limit at the end of 2022 and were made whole anyway, and in the Evergrande parallel, the small savers were protected while shareholders and bondholders were not. That is a defensible way to settle accounts. But it teaches every equity holder that he is first in line to lose — and therefore first in line to sue. Expect more litigation, not less. Expect disclosure, guidance and executive pay to become weapons in the next round.
Nvidia's chips will keep powering the new data economy, politics permitting. But the next front in the tech war is not the fabrication plant. It is the courtroom and the quarterly call. Building a factory that never produces a single chip is one kind of risk; pricing in a future that must be delivered every ninety days is another. And Europe, which enters this story only through diplomatic courtesies and holds roughly 10 percent of the global semiconductor market, is not competing on either front.
Leave a Reply