What: Rising AI compute costs are forcing companies to manage tokens, contracts and infrastructure exposure as strategic financial risks.

Why it is important: This matters because AI adoption is becoming a financial discipline, requiring companies to measure usage, manage volatility and link spending to business outcomes.

BCG argues that the economics of enterprise AI are shifting as providers move away from flat-rate subscriptions toward metered pricing and, potentially, dynamic pricing. This exposes companies to the real cost of AI compute, especially as usage scales across customer-facing tools, internal workflows and large model workloads. The AI compute market is expanding rapidly, with BCG projecting growth from $360 billion in 2025 to roughly $2.3 trillion in 2030. Yet the market remains opaque, heterogeneous and difficult to manage. Prices vary by hardware generation, geography, timing and contract structure, while many large transactions are negotiated privately. BCG expects greater transparency, benchmarks, futures contracts and secondary markets to emerge, helping companies manage price and supply risk. The firm estimates that a more liquid AI compute market could unlock up to $140 billion in annual “dark value” through trading, price optimisation and lower borrowing costs for data center players. For end users, AI labs and infrastructure providers, the message is clear: AI compute must be managed like a strategic commodity, with disciplined workload planning, hedging, contract scrutiny and provider selection.

IADS Notes: As AI adoption expands, the economics of compute are becoming a strategic issue for companies that rely on large-scale automation, analytics and customer-facing AI. In July 2026, analysis of below-cost AI pricing warned that current assumptions may not hold as providers seek to recover infrastructure costs, making cost discipline essential for companies building workflows around AI. BCG’s July 2026 work on the true cost of AI similarly argued that token consumption must be measured by workflow and outcome, not treated as a generic software expense. WWD’s July 2026 analysis of retail AI spending showed that investment is rising while returns remain uneven, especially where companies have not redesigned workflows, governance and decision-making. BCG’s June 2026 research on AI in retail and CPG reinforced that winners are those linking use cases to measurable financial outcomes, while Retail Touchpoints’ January 2026 coverage of smaller, domain-specific models showed how better model selection can improve efficiency and reduce waste.

Is AI computing power becoming a commodity?