Compute Becomes Trillion-Dollar Asset Class as CME Plans Futures

The recognition of compute as a formal asset class, coupled with the introduction of futures trading, signals a significant maturation of the AI infrastructure market, offering new avenues for price discovery and risk management. For streaming services and platforms reliant on AI for content delivery, personalization, and operational efficiency, this means increased financial complexity and new hedging strategies for compute costs. The industry should monitor the adoption and liquidity of these new financial instruments, as they will influence the stability and predictability of AI infrastructure pricing.

The surge in compute demand is driving substantial investment, with Morgan Stanley (May 2026) projecting nearly $3 trillion in AI-related infrastructure spending through 2028, with over 80% still to be deployed. This build-out is now a macro variable influencing GDP and earnings, shifting from speculative tech spending to industrial expansion, with data center construction alone projected to hit $2.9 trillion by 2028. Deloitte (2026) notes that while there's speculation about a shift towards less data center computing for inference, most AI computation will still occur in large data centers using expensive, powerful AI chips. Inference workloads are expected to account for roughly two-thirds of all compute in 2026, with the market for inference-optimized chips growing to over $50 billion. However, even with efficiency gains, demand for compute is rising faster, leading to a potential 'compute crunch,' especially for long-context agentic AI workloads, as indicated by Epoch AI (May 2026). This could drive up the price of access to frontier AI capabilities. Vontobel Asset Management (May 2026) reinforces that demand for AI computing will remain resilient through 2026 and likely into 2027, driven by broadening AI usage beyond training to day-to-day operations and a shift from experimentation to revenue-generating production. Broadcom is gaining significant traction with custom accelerators and networking chips, securing a long-term agreement with Google through 2031, which was highlighted at Google Cloud Next (April 2026). Zylos Research (April 2026) reports that inference now accounts for 85% of enterprise AI budgets and two-thirds of global AI compute spend, representing a structural inversion in AI economics. Despite falling per-token costs due to hardware advancements like NVIDIA's Blackwell architecture, total organizational spend on inference is accelerating due to the multiplication of agentic workloads.