Research for AI hardware markets.
Analysis on GPU supply, deployment, power constraints, secondary markets, and price discovery.
Research
Stoa research
How the secondary GPU market works
Who buys and sells used enterprise GPUs, how the market clears today, and why residual value has become a financing question.
Read
What GPU-backed lending is
How financing secured against GPUs and AI clusters works, how large it has become, and why it depends on price discovery.
ReadWhy GPU price discovery is fragmented
Data-center capex is headed for the trillions, yet the hardware at its center has no shared reference price. Why, and what changes it.
ReadSelected
Selected reading
American AI Companies Can't Get Enough Chips
How AI chip production is becoming a binding constraint on compute buildout.
Read at CNASGPU Procurement Trends in U.S. Data Centers
How operators are planning GPU supply around power, cooling, and long-term capacity.
Read at BIS ResearchCan US Infrastructure Keep Up With the AI Economy?
Power, grid capacity, and supply chain gaps shaping AI data center growth.
Read at DeloitteAI Chip Lifespans: A Note on the Secondary Market
Why generational refresh cycles and downstream demand matter for secondary GPU clearing.
Read at Princeton CITP