For most of the past decade, an investor who wanted exposure to computing power bought it the obvious way: shares in the companies that build and rent it out. A newer argument is gaining adherents — that AI compute itself, the electricity and the GPUs that turn a data center into billable inference, is becoming an asset class in its own right, something to own directly rather than hold at one remove through a mega-cap balance sheet.
The case starts with demand. The visible face of AI is the chatbot, but the spending that matters sits deeper, where companies wire models into the core of how they operate — pricing, logistics, research, fraud, design — and find that more compute converts fairly directly into more output. Microsoft's Satya Nadella has spoken of "tokens" processed as a measure of that activity, a way of framing capacity as the raw material of the AI economy. Where that holds, demand for inference does not saturate the way a one-time software rollout does; it compounds with each new use.
That has drawn attention to the plumbing. A wave of "neocloud" providers is trying to supply inference more cheaply than the hyperscalers by owning the stack end to end — power, modular data centers, orchestration software. Among them is Antimatter, launched in April 2026 by the French entrepreneur David Gurlé — the founder of Symphony and a Chevalier of the Légion d'Honneur — which describes a distributed, energy-first network of modular units. In a further sign of the appetite for direct exposure, a related vehicle called Promessia was publicly announced through the French media channel Grand Angle, proposing to let individual investors help finance a compute farm rather than buy it through listed equity.
None of this settles whether "asset class" is the right words. It is a claim, and the counter-arguments are not small. Listed AI-infrastructure valuations already price in years of growth, which leaves room for disappointment if the build-out runs ahead of real demand and tips into overcapacity — the classic fate of capital-intensive booms. And gaining direct exposure, especially through private, early-stage vehicles, means trading the liquidity and disclosure of public markets for bets that can be hard to value and harder to exit. The reminder from Bitcoin's own history is apt: a genuinely new asset class is obvious only in hindsight, and the road there is rarely smooth.
This is market analysis, not a recommendation to buy or sell any security.