Nvidia’s Blackwell GPUs now command $4.08 per hour, a 48% jump from $2.75 two months ago, while CoreWeave has lifted prices 20% and stretched minimum leases from one to three years. The spike reflects a tightening supply chain that has finally seeped into AI workloads, forcing cloud providers to re‑evaluate capacity budgets and prompting a scramble for alternative compute sources.

Anthropic has limited its newest model to roughly forty organizations, turning cutting‑edge AI into a gated privilege for firms that can afford the high compute cost. Sarah Friar, OpenAI CFO, warned that the industry faces “tough trades” because compute is scarce, a reality that could stall innovation for smaller players and threaten the pace of new model releases.

The scarcity forces a shift toward relationship‑based selling, higher bids, slower delivery, and forced diversification. Developers may pivot to smaller models or on‑premise solutions until data‑center capacity catches up—a process that could take years. The era of abundant AI is over, and the new reality will endure for years, reshaping how enterprises budget for compute.