Blockchain firms, DePIN networks, and major exchanges are exploring compute trading as AI demand grows, though experts remain divided on how quickly the market can mature.
The race to build the AI economy is no longer just about GPUs. It’s increasingly about creating financial markets around them. From blockchain infrastructure firms and DePIN (Decentralized Physical Infrastructure Networks) projects to traditional exchanges and Wall Street institutions, a growing ecosystem is working toward turning compute power into a tradable asset class.
While some industry leaders believe compute futures could become the next major derivatives market, others argue the industry is still years away from reaching that point.
Former FTX US President and Architect CEO Brett Harrison said compute is following the same path as commodities like oil, electricity, and natural gas.
According to Harrison, demand is coming from a wide range of participants, including GPU cloud providers, neoclouds, data center operators, AI training companies, energy firms, investment banks, brokerages, ETF issuers, venture capital firms, and blockchain infrastructure companies.
His view is that as AI adoption accelerates, businesses will increasingly need tools to hedge fluctuating GPU costs, creating demand for regulated compute derivatives.
Several major US derivatives exchanges are already preparing themselves to facilitate this new market.
Architect, a derivatives exchange group, recently acquired a CFTC-regulated derivatives exchange. It plans to launch dedicated AI compute futures and options trading using pricing data from Compute Desk.
Meanwhile, CME Group has partnered with Silicon Data to introduce compute futures later this year, while ICE Futures US has announced similar plans using pricing benchmarks from Ornn, pending regulatory approval.
These exchanges aim to provide standardized contracts that allow businesses and investors to hedge or speculate on GPU pricing while improving overall price discovery.
The companies most closely aligned with this trend are compute marketplaces and blockchain-powered infrastructure platforms.
A growing list of firms, including Akash Network, Aethir, Spheron, Prime Intellect, Hyperbolic, RunPod, Vast.ai, Compute Exchange, Compute Desk, HydraHost, VoltagePark, Shadeform, SaladCloud, Clore.ai, Cudo Compute, DigitalOcean, SF Compute, Ornn, and NVIDIA DGX Cloud, Lepton, are already matching GPU buyers with suppliers.
Many of these platforms also collect GPU pricing data that could eventually serve as benchmarks for regulated futures markets.
For blockchain-based DePIN projects, this represents a major opportunity, as decentralized GPU networks could become an important source of liquidity for future compute markets.
Despite the excitement, not everyone is convinced. Investor Jonah Van Bourg argues that financializing compute will be much harder than many expect. He compares today's market to failed attempts to build liquid derivatives around LNG.
His biggest concern is that GPU technology evolves too quickly. Contracts based on NVIDIA's H100 chips could become obsolete as newer hardware like H200, B200, or Rubin systems take over.
He also said the market will fragment across NVIDIA, AMD, Google TPUs, Amazon Trainium, hyperscaler ASICs, inference chips, and sovereign AI hardware, making it difficult to establish a single benchmark.
Analyst Van Bourg also pointed out that compute isn't standardized enough for traditional futures markets.
Pricing depends on factors such as chip type, networking, cluster size, location, uptime, storage, security, and support, making each compute contract slightly different.
He believes sustainable compute trading will eventually emerge, but likely through narrow OTC contracts tied to specific chips, regions, and delivery terms rather than broad standardized futures.
For now, the debate highlights one thing clearly: as AI infrastructure demand grows at its current rapid rate, blockchain firms, DePIN networks, exchanges, and traditional financial institutions are all competing to define how compute will be bought, sold, and traded in the years ahead.
No comments yet
Be the first to share your take when accounts launch.
