Silicon Data's Carmen Li thinks AI computing power should trade like oil. Futures contracts. Price discovery. Hedging strategies. The whole circus.
She believes this market could rival the world's largest commodity markets. Which means she thinks people will eventually panic-sell GPU hours at 2 a.m. because some analyst in Frankfurt sneezed during an earnings call.
Oil makes sense as a commodity. You dig it up. You ship it. You burn it. It's gone. Supply and demand actually mean something when the product stops existing after you use it.
Compute doesn't work that way. A server farm in Iowa doesn't run dry. NVIDIA doesn't need to explore for new deposits of graphics cards in the Arctic. The comparison falls apart the second you think about it for more than thirty seconds, which is approximately twenty-nine seconds longer than anyone pitching this idea thought about it.
But sure, let's create a derivatives market for something that's already virtualized, already metered, already priced dynamically, and already traded between cloud providers. What could make that system better? Letting a bunch of commodity traders in Chicago speculate on it while some kid's margin account gets liquidated because he went long on H100 futures and didn't understand contango.
The retail traders salivating over this are the same people who bought USO in April 2020 and learned what physical delivery means the hard way. Now they get to learn what happens when their AI compute contract expires and somebody delivers sixteen pallets of server racks to their studio apartment.
Li is right about one thing. This market will get huge. Not because it solves a problem, but because Wall Street has never met a functioning market it didn't want to financialize into a casino. The new oil isn't computing power. It's whatever lets brokers charge commissions on something that used to just be a f*cking AWS bill.
Leave a Comment