SAN JOSE, California, March 19, 2026, 02:04 PDT
Nvidia’s networking arm now tops $31 billion, carving out a strong second leg for the AI chip giant. The unit just hauled in $11 billion last quarter—over 3.5 times what it managed a year ago. As the company shakes off its headline $68 billion quarter, investors are turning to see how much fuel is left in the tank.
That’s become key as the conversation about Nvidia moves on. Selling GPUs is a given now; the focus has turned to whether Nvidia can lengthen the AI spending boom by moving whole systems — what it dubs AI factories, essentially data centers designed to train and deploy models, bundling compute, networking, and storage.
At Nvidia’s March 16 GTC event, Huang doubled down: the company now pegs a $1 trillion revenue opportunity for its Blackwell and Rubin chips through 2027, jumping from the $500 billion estimate for 2026. He referred to this as the “inference inflection”—meaning the pivot from training AI models to deploying them live in real-world use. Reuters
These are hefty figures. Nvidia clocked $68.1 billion in record quarterly revenue, with data center revenue hitting an all-time high of $62.3 billion for the quarter ended Jan. 25. CFO Colette Kress pointed to networking as “a standout,” noting customers ramped up purchases of scale-up and scale-out gear for massive clusters. NVIDIA Newsroom
Kevin Cook, a strategist at Zacks, pointed out to TechCrunch that Nvidia’s networking sales each quarter have now topped Cisco’s entire networking business—a sign of just how rapidly the segment has shifted from peripheral to central.
Inference work is a tougher battleground. Nvidia finds itself facing off with Intel’s CPUs, Google’s in-house chips, plus Chinese tech giants like Baidu. The company is racing to launch a Groq-based product that will meet China’s requirements and, according to Reuters, is also restarting some H200 production for the Chinese market.
This is the pocket of risk. Cleo Capital’s Sarah Kunst pointed out that after a stretch of hypergrowth, investors might have to get used to a “normal growth stage.” She cited stiffer year-over-year comps and headwinds from China, both factors that could drag on Nvidia’s quarterly fireworks—even if the company stays at the heart of AI capital flows. Still, fiscal 2026 revenue landed at $215.9 billion. 24/7 Wall St.
Nvidia slipped roughly 0.8% in Thursday’s session from its last close, according to the latest trading data.
On March 17, Huang clarified that the $1 trillion figure leaves out Nvidia’s CPUs, networking chips, Groq-derived offerings, and Rubin Ultra. Rubin-powered systems are set to arrive from Amazon Web Services, Google Cloud, Microsoft Azure, and Oracle in the latter half of 2026. That timeline puts a near-term spotlight on whether Nvidia’s expanded platform can push the story past chips alone.
