Wall Street has already made billions betting on the companies selling the picks and shovels of the AI boom. The next opportunity may be figuring out how to trade the thing everyone is digging for: Compute.
That may sound like semantics, but now, after Nvidia CEO Jensen Huang convened partners spanning BlackRock, Apollo, and Blackstone for a new $500 billion initiative — it's much, much more.
Sure, Nvidia sells chips. And cloud providers rent computing power. Data centers house the machines. None of this exactly new. But now that Huang opened the door this week by selling Wall Street on his chips being an "investable asset," something has now categorically changed.
The answer is that compute is beginning to behave less like a piece of technology and more like a scarce economic input — something that can be financed, contracted, priced and, eventually, hedged. In other words, compute is starting to look like a commodity.
"The computer is now part of the infrastructure, like electricity, like the internet," Huang said during the CNBC roundtable this week. "And so you have to think about it like it's infrastructure and build it out accordingly."
That is a much bigger idea than simply saying the world needs more GPUs.
Bitcoin Holds $63,000 Despite Another Strategy Sale
Why Trump's Crypto Project Is Launching Its Own Dollar
For most of the history of computing, chips have been depreciating technology. Buy the newest processor today and eventually a faster, more efficient one comes along. Yesterday's cutting-edge hardware becomes tomorrow's obsolete equipment. The AI boom has complicated that familiar curve.
As Coinage has covered for years now, demand for Nvidia's H100 chips became so intense that older hardware retained unusually strong pricing power. What ordinarily would have been aging technology instead became access to compute capacity the market desperately needed.
Oil is not valuable because barrels are technologically sophisticated. It is valuable because enormous parts of the global economy need energy, and access to that scarce input has to be financed, transported, contracted and hedged. AI compute is moving in the same direction.
The companies building frontier models need staggering amounts of processing power. That is why Huang's choice of words was so notable.
"We used to build chips that we sell," he said. "But now Nvidia's AI factory platform is really an investable asset, an infrastructure asset." That is the bridge between Silicon Valley and Wall Street.
If AI infrastructure can produce durable revenue streams, then banks can finance it and investors can own it. If access to compute becomes standardized enough, companies can sign longer-term contracts around it. And if the price of compute becomes an important enough input cost, markets can eventually develop ways to hedge it.
That last step may sound futuristic, but crypto markets are already experimenting with it.
Lighter founder Vlad Novakovski told Coinage that his exchange had introduced an H100 compute index. The point was not merely to give traders something else to speculate on.
"If you're building a fully autonomous agent, you can actually hedge the cost of your own compute onchain," Novakovski said.
That is where the commodity analogy becomes much more than a metaphor. Airlines hedge fuel because they cannot run their businesses without it. Manufacturers hedge metals. Farmers hedge crops.
If AI companies eventually become just as dependent on compute, why wouldn't they want to protect themselves against rising compute costs? And if autonomous AI agents eventually buy their own computing resources, the market could become stranger still: Software managing its own exposure to the price of the processing power it needs to exist.
It's a leap investors should be paying attention to. The first phase of the AI trade was primarily about selling more chips. Nvidia became the clearest expression of that thesis. The second phase may increasingly be about monetizing access to those chips.
Financial firms can provide the capital needed to build more of it. And markets may eventually emerge to trade the price of the compute itself. Not every company claiming exposure to that buildout will benefit equally.
CoreWeave offers perhaps the most obvious public-market example. Its business gives investors exposure to the infrastructure layer that sits between chipmakers like Nvidia and the companies actually consuming enormous amounts of compute.
The financial industry has spent centuries finding ways to turn economically important resources into investable markets. It finances oil wells, power plants, pipelines, warehouses and telecommunications networks. It creates futures, swaps and other derivatives so companies can manage their exposure to volatile input costs.
There is no reason to assume AI infrastructure will remain exempt from the same process.
Subscribe to the free Coinage newsletter to stay up-to-date on all things crypto and finance