- Nvidia data center compute revenue dips for first time in ten years
- Networking sales surge as compute slows while China restrictions create added pressure
- Dependence on three customers hints at Nvidia’s vulnerability despite record overall revenues
You’d be forgiven for thinking Nvidia’s meteoric rise has made it virtually bulletproof, but data from the second quarter of fiscal 2026 shows the chip giant’s data center compute revenue declining for the first time in a decade.
The drop, while small, marks a potential turning point for the company which has led the artificial intelligence hardware boom.
Nvidia once again reported record revenues overall, but a graph shared by The Next Platform, hints at a worrying plateau beginning to form in the GPU behemoth’s core data center compute business.
No need to panic… yet
The slide (which you can see above) showed that the sales of GPUs and CPUs that underpin Nvidia’s AI infrastructure dipped 0.9 percent compared with the previous quarter.
That’s not a big drop, but it is notable since it’s the first sequential decline since data center compute became the Nvidia’s main source of growth.
Networking revenues tell a different story, with InfiniBand networking nearly doubling, and Ethernet and other interconnects also showing big gains.
The combined networking segment delivered $7.25 billion in revenue during the July quarter, suggesting customers are channeling more spending into interconnects that link AI clusters at scale.
That shift comes as export restrictions to China have already cut into expected compute sales, creating more pressure on Nvidia’s leadership.
Chief executive Jensen Huang insisted during the earnings call that fiscal 2026 and 2027 will still be record years, but the curve on the data center chart is still a concern.
Another worry is the company’s reliance on a small number of customers, which we reported on previously.
More than half of data center revenue now comes from just three unnamed clients, with one alone accounting for over 20% of sales.
The report highlighted $9.5 billion from Customer A, $6.6 billion from Customer B, and $5.7 billion from Customer C in the most recent quarter.
Growth in GPU sales has powered Nvidia’s transformation into one of the most valuable companies in the world, but history shows that expansion curves eventually have to level off.
Huang has argued that 50 percent growth in data center compute can continue for years, but market realities may force slower trajectories.
For now, demand for Nvidia’s Blackwell GPUs remains solid, with supply reportedly sold out into next year.
The company continues to prepare new platforms, but the revenue plateau hints that Nvidia’s dominance may be entering a more uncertain phase.
