Galaxy Research explores emerging onchain markets for AI compute pricing, tokenization, and GPU hardware financing.
AI & Agents ·
A new financial infrastructure is forming around AI inference—the execution of trained models on new data—as decentralized networks and onchain protocols coordinate GPU operators through crypto incentives and settlement. Galaxy Research describes this layer as an "inference capital market," combining GPU marketplaces, payment systems, tokenization mechanisms, and liquidity provision into a single integrated system. Users can route prompts to distributed networks rather than relying solely on centralized providers like OpenAI or Anthropic, with some configurations offering cryptographic or economic guarantees on output accuracy and privacy.
Several shifts are driving this convergence. Inference has surpassed training as the primary driver of global GPU consumption, while open-weight models have narrowed capability gaps with proprietary frontier models for many use cases, making cheaper alternatives economically viable. A decline in token spending tracked by the Silicon Data LLM Index reflects this shift toward lower-cost model usage. Simultaneously, companies including Coinbase, Microsoft, and AirBnB have moved toward open models—predominantly Chinese variants—and platforms offering diverse model access, as supply constraints increase the marginal cost of inference provision.
The second force is financialization: efforts to render AI compute a tradable commodity within broader financial markets. The convergence of these primitives into a functioning capital market remains experimental, though demand now originates from genuine productive activity and uses beyond crypto itself, including autonomous agents that programmatically consume inference without human intermediation.