The worldwide chip industry is being repriced in real time somewhere between a quiet symposium hall in Shanghai and a converted trading pit in Chicago. Other than setting the pace of a market that hardly anyone fully comprehends, the two cities have very little in common. Investors appear to think it is more important than anything else in technology at the moment. It’s difficult to disagree.
A few years ago, Huawei’s announcement on a Shanghai stage on a Monday morning in May would have seemed almost ridiculous. According to the company, it plans to manufacture chips with a transistor density of 1.4 nanometers by 2031, which puts it, at least theoretically, close to the global frontier. Nearly seven nanometers is where China’s most sophisticated, proven chipmaking is located. The timeline is ambitious, the leap is steep, and it is impossible to ignore the politics. It seems like Huawei isn’t actually recruiting engineers. Beijing, Washington, and all investors following the export-control narrative are being pitched.
The company refers to its methodology as the Tau Scaling Law, which is a fancy term for a reasonably grounded concept. Huawei wants to accelerate the flow of signals within the chip rather than further shrink transistors, which has become physically taxing because some features are now only a few atoms wide. shorter connections. reduced latency. improved data flow. It’s the kind of turnabout that businesses typically take when the traditional route is blocked, and in Huawei’s case, the path has been purposefully blocked by lithography restrictions in the United States.
As this develops, it’s difficult to ignore how much of the chip story now sounds like geopolitics wrapped in engineering jargon. Every announcement—ASML’s Dutch machines, Taiwan’s TSMC’s push toward two-nanometer production, Washington’s export restrictions, Beijing’s drive for self-sufficiency—carries the weight of something more significant than a product roadmap. Riding the same AI wave that has hyperscalers squabbling over wafer-processing tools they didn’t realize they needed two years ago, ASM International recently reported first-quarter sales of about $1 billion.
And there’s Chicago’s peculiar spectacle. CME Group intends to start a futures market for compute itself in collaboration with Silicon Data, a smaller company. Soon, traders will be able to wager on GPU rental prices in the same manner that they used to wager on crude oil or soybeans. Although regulatory review of the contracts is still pending, the symbolism is already very evident. Compute is now a commodity. Or at least that’s how the financial community has chosen to handle it.
It’s important to keep in mind that this concept is not wholly novel. One of the most well-known corporate collapses in American history resulted from Enron’s attempt to sell unused fiber-optic capacity using a similar method during the late 1990s broadband boom. Even though no one at CME wants to bring it up, that memory looms over the announcement. This time, the structure is different and the demand is clear, but the underlying reasoning—converting infrastructure into a tradable abstraction—is uncannily familiar.
Future AI systems are already being described by Morgan Stanley’s analysts as distributed grids of GPU and CPU racks operating side by side, memory chip prices skyrocketed in the first quarter, and executives continue to warn about bottlenecks. Silicon is no longer a major factor in the bidding war. It concerns who has the authority to determine the cost of thinking. It’s still unclear if that price will stabilize or act like the wild commodity it’s turning into. Even though most people are unaware that they are part of the auction, one thing is certain: it has already begun.