DIGITIMES reported on November 18, 2026 that Nvidia will move customers to new pricing in fiscal 2028. Until then, the company has been covering the increased supply costs on orders already booked. In other words, part of the price hike is staying on Nvidia's books for now rather than showing up on customers' invoices.
The logic is simple: management is counting on accelerator shortages and rising GPU rental rates in data centers to give it room to claw back margin. The second storyline is product expansion: the report names Vera Rubin, CPUs, and enterprise AI as growth drivers. The third, less pleasant one: the biggest customers are pushing their own chips harder than ever.
The full DIGITIMES piece is paywalled, so the open portion contains no details — no margin percentages, no actual price lists, no shipping dates. Read everything below with that caveat in mind.
Our read: this is a rare case of Nvidia talking about price before hardware. That sequence is usually a signal to procurement teams: brace yourselves, there will be no cheap contracts on the Vera Rubin transition. The "fiscal 2028" timing is no accident — it leaves a year and a half of negotiating room.
What this changes in practice: hyperscalers — Microsoft, Google, Amazon, Meta — now have to factor in not just FLOPS per dollar but the cost of migrating from Hopper/Blackwell to the next cycle. If GPU rental rates really are climbing, that's one more argument for their own chips. We think that's exactly why "in-house chips" get their own line in the source: it's not background noise, it's a direct threat to Nvidia's model.
Who wins: Nvidia — if the shortage holds until Vera Rubin shipments start; otherwise the pricing promise hangs in the air. Cloud GPU providers also come out ahead — rising rental rates work in their favor. Who loses: mid-size customers with no leverage — they have neither their own silicon nor the volumes to lock in the old terms.
The article is behind a paywall, and all we have is the opening paragraph. No cost-growth percentages, no breakdown by product line, no confirmation of whether this came from an investor call or off-the-record comments. One source, a paraphrase of a statement with no direct quote. Don't take "margin recovery" at face value: that's a company goal, not a result.
Should Nvidia lock in pricing for the Vera Rubin transition — or let it float, lose some customers, and protect margin?
Whose argument is stronger in the in-house chips debate: hyperscalers already have working accelerators, or their unit costs are still higher than Nvidia's?
If GPU rental rates keep rising through fiscal 2028, will the clouds actually benefit — or will they lose customers who simply postpone purchases?
Forecast: Nvidia won't publish an open price list for Vera Rubin before the end of fiscal 2028 — new pricing will flow through individual contracts. You can check this by whether a public price catalog appears.
Disclaimer: Data comes from a single source, and the full article is paywalled — the details have not been independently confirmed.
Source: DIGITIMES
Source: www.digitimes.com · post in Telegram