but commodity pricing also means that tokens are subject to supply and demand.
even the availability of finetuned models are dependent on the availability of the compute to experiment, finetune, and then run inference. in a typical sense, the economic value will flow towards the inference providers. but enterprises that understand this and exert control over the full stack will win.
it all comes back to compute
Our H100 spot rental price index has been trending up much further before mid-Nov. There are lots of potential reasons, but a bigger reason for spot + short-term reserved prices spiking is buyers paying for immediate supply because future supply availability is uncertain.
"This turns compute from a fixed constraint into an actively managed portfolio." - CFO OpenAI, today
"All clouds will need new forms of portfolio optimization that fits their risk profile, capital structure, and financial engineering strategies." - Squaretower Research, three days ago
With @StorkOracle we are able to bring critical datasets to 72+ organizations, enabling Squaretower’s compute data to power new products that underwrite hundreds of billions of dollars in growing compute risk.
Compute is the most critical commodity of the next century. More than ever, we need transparent data and tools to understand how this commodity will shape and form economies of the future. This starts with Squaretower’s proprietary datasets, benchmarks, and research.
If you’re a team interested in data access — whether for understanding compute pricing or for new products — reach out.
Bringing GPU compute pricing on-chain.
@squaretower_ is building the financial infrastructure for the compute economy, aggregating GPU rental pricing across cloud providers and normalizing fragmented configurations into standardized, comparable metrics.
Now, Squaretower joins Stork as a data publisher, delivering high-fidelity GPU pricing directly to on-chain applications, starting with an NVIDIA H100 index.
We've reached a milestone of $20.2 million notional volume traded on our compute index.
AI compute markets are still missing the basics: transparent benchmarks, rapid price discovery, and risk transfer tools, despite GPUs becoming the single biggest and most volatile line item for today's most successful companies. Pricing + hedging exposure to the most important asset of the next century should be straightforward.
Squaretower in @MessariCrypto 2026 thesis for launching the first ever compute derivative — the next wave of startups, AI labs, and infrastructure providers need novel markets & infrastructure to unlock the trillion-dollar opportunity in compute commodities
H100 rental prices have been surprisingly volatile in the last few months, according to Squaretower data. The demand floor is rising, but so is uncertainty.
And H100 demand is likely strengthening A100s, as buyers slide down the curve when access is low or costs are high.
today on @tbpn, @BernsteinRasgon says
- compute demand side shows no signs of weakening
- hyperscalers are clearly offloading risk to the neoclouds
- most of capex is still being funded by free cash flow ($1.4t), NOT debt ($800b)
- nvidia basically reporting low sales on its $500b order pipeline for blackwell & rubin in 2026
- is "probably not" AGI-pilled
Great piece by @jongall45 covering the rapidly growing compute markets space. Squaretower provides the data, tools, and insights to make these new markets possible.
@MauiBoyMacro the key here is that the tail is very, very long
biggest players will keep winning. everyone else trying to get a slice of the pie will keep losing to the prices they set