Strip away the share price and Nvidia’s results are one of the most useful planning documents in the industry. Three themes tell operators how long they will wait, what they will pay, and who holds the power.
By Steven Foley | Datacenter Focus | 13 November 2026
Nvidia reports its latest quarterly results after US markets close on Tuesday. Most coverage will treat it as a verdict on the AI boom, and the stock is already jumpy: it fell 9.3% on 5 November – its largest one day drop in a year. For data centre operators the share price is mostly noise – Nvidia sets the pace of nearly everyone’s build, so its numbers work more like a delivery schedule than a scorecard.
You are no longer buying chips
Vera Rubin began production shipments in August and it is sold as a whole system rather than a single chip, spanning a CPU, a GPU, networking and a further accelerator. Management reportedly puts what Nvidia expects to earn per gigawatt at $40bn for Rubin, against $25bn for Blackwell – that is roughly 60% more spent with one supplier for the same slice of power.
Our view is that the operator’s role is shifting from choosing components to choosing an ecosystem – which makes the supplier relationship more important, and the bargaining position weaker.
Supply, not demand, sets your timeline
Management expects supply to stay a bottleneck at least through the end of fiscal 2028. Demand is not the constraint – AWS is deploying an additional 2 million Nvidia GPUs in a programme running into 2028. For operators, a finished building and a signed power connection are no longer enough – there is a third queue, for the equipment itself. A powered shell with no delivery slot is an idle asset and so is a delivery slot with no power. The operators who do well will be those who secured both at the same time.
The squeeze nobody budgeted for
Memory is getting more expensive and it’s reaching Nvidia’s own costs. One analysis tells readers to watch gross margin against 73.5%, with memory costs squeezing it. Another expects margin to reach a low point of 71% to 72% as memory prices keep rising – another summary points to price increases taking effect in fiscal Q1 2028.
When a supplier’s costs rise, customers should expect the increase to reach their purchase orders – a budget built on last year’s cost per rack is probably already out of date.
Three things to listen for on Tuesday
- How much of the business is Rubin? The expectation is about 20% of data centre revenue this quarter. If it lands well short, delivery delays may be heading your way.
- Does the outlook hold? Management has said it expects sales to the largest cloud companies to pick up again as Rubin supply grows – a wobble would suggest the biggest buyers are slowing down. A firm outlook means competition for slots stays fierce.
- What does margin say about costs? A weak figure is an early warning about pricing.
None of this settles the wider argument about AI financing – Reuters reports that some banks are now reluctant to fund new AI deals even with guarantees from marquee tech names so a strong Nvidia result might help sentiment, but it doesn’t fix a lender’s worry about contract terms.
Operator takeaway
Plan around allocation, not announcements and treat equipment slots as seriously as power. Build price variance into every budget and be wary of depending on a single supplier for the whole stack.
Operators should listen carefully to what Nvidia says about supply on Tuesday evening – that is the part that matters most to them.
