Affaan Mustafa and Alejandro Soumah Fernandez describe how long-term GPU capacityAI compute capacity is the available ability of hardware and supporting systems to perform AI training or inference work over time. is negotiated through private bilateral contracts and broker networks rather than a widely adopted public exchange. Mustafa distinguishes the cost of model intelligence from the demand for hardware capacity.
They outline how upfront payments and long commitments can strain a growing AI company, and compare owning infrastructure with renting capacity. The speakers explain how specs, delivery terms and access affect the economics of each deal.
The talk surveys brokers, price references, potential hedges and policy considerations. The speakers argue that standardized terms and intermediaries able to manage riskRisk assessment identifies potential harmful outcomes and examines their likelihood, impact and uncertainty in a specified context. It provides evidence for deciding how a system should be managed. could help connect suppliers with buyers whose needs change.
Watch on YouTube
