What This Is
Nvidia this week doubled the official MSRP of the RTX PRO 6000 Blackwell workstation GPU: the 96GB variant jumped from under $8,000 at last year's pre-order to $16,000 — same chip, just relabeled for "AI inference" (running trained models to produce actual outputs).
This isn't an isolated case. Former A16Z partner Gavin Baker recently posted on X that multiple private companies have explicitly stated they will renew GPU procurement contracts at double the price once existing ones expire. Compute is shifting from "commodity on demand" to "strategic resource."
Industry View
Supporters argue this is a normal reflection of supply-demand dynamics: Blackwell yield improvements are slow, capacity is being squeezed by hyperscale data center orders, so high-VRAM cards allocated to workstations are naturally scarce, and price increases are justified.
But the opposition is equally sharp. The reaction on LocalLLaMA (an English-language forum where independent AI developers gather) was nearly unanimous: some point out that Nvidia has long practiced differentiated pricing between consumer and workstation cards, with the same chip selling for several times the price across channels — the current hike just puts "price discrimination" on the table openly. Others worry that DGX Spark (Nvidia's newly launched desktop AI workstation) will follow suit.
Our take: this is an open test of pricing power. Nvidia is betting that the AI inference market can't live without it, and customers have no choice.
Impact on Regular People
Enterprise IT departments: Hardware budgets for AI training and inference this year need to be recalculated. The depreciation models originally built on "GPU prices drop 30% annually" no longer hold.
Individual careers: It's getting harder for small and mid-sized teams deploying AI locally. The hardware entry threshold has shifted from "a few tens of thousands to build one machine" to "starting at over a hundred thousand," and the space for independent developers continues to shrink.
Consumer market: Consumers won't feel it immediately, but the price gap between gaming cards and professional cards will keep widening — the cost pressure on DIY high-end builds will eventually pass through to end users.