Jensen Huang walked into Nvidia’s fiscal Q1 2026 earnings call with the numbers Wall Street expected — $81.6 billion in quarterly revenue — and then casually dropped three pieces of information that tell you exactly where the AI infrastructure war is headed next. First: Nvidia expects $91 billion in Q2 revenue, beating analyst estimates by roughly $3 billion. Second: the company unveiled Vera, a new central processor designed for data centers, targeting what Nvidia calls a $200 billion addressable market that it currently doesn’t compete in. Third: an $80 billion stock buyback — the largest in semiconductor history.
Each of those headlines alone would dominate a normal earnings cycle. Together, they paint a picture of a company that isn’t just riding the AI boom — it’s expanding the definition of what the AI boom includes.
The $91 Billion Quarter Nobody Predicted
Let’s start with the headline number. Nvidia’s data center revenue — which is now functionally its entire business — drove the $91 billion Q2 forecast. Wall Street consensus was roughly $88 billion. Nvidia beat that by a comfortable margin, which tells you something specific: the hyperscaler spending cycle isn’t slowing down, and the inference workload ramp that every analyst warned about being “the next growth driver” is arriving faster than the models predicted.
The growth rate here deserves context. This time last year, Nvidia posted $26 billion in quarterly revenue and people thought the company was printing money. A year later, $91 billion. That’s a 3.5x jump in four quarters. No semiconductor company in history has done this. Intel at its peak couldn’t dream of this. TSMC, which manufactures Nvidia’s chips, generates less revenue than what Nvidia is now guiding for a single quarter.
The engine behind the surge is straightforward: Blackwell Ultra (GB200) GPUs are shipping in volume, and every major cloud provider — Amazon, Microsoft, Google, Oracle, and now SpaceX — is buying as fast as Nvidia can produce. SpaceX’s S-1 filing this week revealed that Anthropic alone is paying $1.25 billion per month for compute at SpaceX’s Colossus facilities, and most of that hardware is Nvidia silicon. OpenAI is spending at a comparable rate across Microsoft Azure. Google is burning through its own TPU supply and still buying Nvidia GPUs on top.
The demand picture isn’t just large. It’s accelerating.
Vera: The Quiet Bomb That Changes Nvidia’s Entire Business
The real story of this earnings call isn’t the Q2 forecast. It’s Vera.
Vera is Nvidia’s new data center CPU — a central processor, not a GPU. This is Nvidia entering a $200 billion market that it has, until now, largely left to Intel, AMD, and Arm-based challengers like Ampere Computing and Amazon’s Graviton. Jensen Huang told investors that Nvidia expects Vera to generate $20 billion in revenue by the end of the fiscal year, which would make it one of the fastest product ramps in chip history.
Here’s why this matters more than another GPU upgrade cycle. Every AI data center needs CPUs alongside GPUs. Right now, those CPUs come from Intel (Xeon) or AMD (EPYC). Nvidia has been dependent on those companies for the “other half” of its data center servers. With Vera, Nvidia is telling Intel and AMD: we’re coming for the rest of the server, too.
The strategic logic is ruthless. If you’re a hyperscaler building an AI data center, you’re already buying Nvidia’s GPUs, Nvidia’s networking (InfiniBand/Ethernet switches via Spectrum-X), and Nvidia’s software stack (CUDA, NeMo, NIM). The only component Nvidia doesn’t supply is the CPU. Vera closes that gap. A fully Nvidia-powered server — GPU, CPU, networking, software — is now possible. That’s vertical integration at a scale that should genuinely terrify Intel.
Intel, which is already fighting for survival after losing its process technology lead and watching its data center market share erode, now faces Nvidia entering its last remaining stronghold. AMD’s EPYC has been steadily gaining share in data centers, but even AMD should be nervous — Nvidia’s distribution advantage with hyperscalers is enormous, and bundling Vera with GPU purchases creates a lock-in dynamic that’s hard to compete against.
The $80 Billion Buyback Is the Confidence Signal
Companies announce stock buybacks for two reasons: they believe their stock is undervalued, or they have so much cash that they genuinely don’t know what else to do with it. In Nvidia’s case, it might be both.
An $80 billion buyback is staggering. For context, Apple — the company that essentially invented the modern mega-buyback — authorized $110 billion earlier this year. Nvidia’s $80 billion puts it in the same conversation, except Nvidia is growing revenue at 3.5x year-over-year while Apple’s revenue is growing at single digits. On a growth-adjusted basis, Nvidia’s buyback is the most aggressive capital return program in technology history.
The message to investors is clear: Nvidia isn’t worried about a demand cliff. It’s not worried about the “AI bubble” narrative. It’s not worried about custom silicon from Google (TPUs), Amazon (Trainium), or Microsoft (Maia) eating into its market share. If anything, the buyback says the opposite — Nvidia sees the current stock price, with all those concerns priced in, and thinks it’s still a bargain.
Whether that confidence is justified depends on one question: can the hyperscalers keep spending at this rate?
Follow the Money: Who’s Actually Buying $91 Billion Worth of Chips?
The spending math tells you everything. In 2026, the five largest cloud providers — Amazon, Microsoft, Google, Meta, and Oracle — are collectively spending over $630 billion on capital expenditure, with the majority going to AI infrastructure. Nvidia captures the most valuable slice of that spending: the GPU and accelerator layer.
But there’s a second wave now. SpaceX’s data center division is spending $30 billion annually on AI infrastructure. Saudi Arabia, the UAE, and sovereign wealth funds are building national AI compute facilities. India’s government-backed AI mission is allocating billions for GPU clusters. Japan, South Korea, and Singapore are all racing to build sovereign AI infrastructure.
Nvidia isn’t just selling to five tech companies anymore. It’s selling to governments, sovereign funds, defense departments, and a new class of AI-first infrastructure companies that didn’t exist two years ago. That diversification is why the $91 billion forecast is credible — and why Jensen Huang can announce an $80 billion buyback without the board blinking.
The China Problem Isn’t Going Away
There’s one shadow hanging over all of this, and Huang addressed it directly. Nvidia has “largely conceded” the Chinese market to Huawei, he acknowledged this week. That’s roughly $16 billion in annual revenue that Nvidia can no longer access due to US export controls on advanced AI chips.
Huawei’s Ascend 910C processor is now the default AI accelerator in Chinese data centers, and Chinese cloud providers are building their own ecosystems around it. Nvidia can sell lower-end chips to China, but the high-margin, high-volume data center GPUs that drive Nvidia’s profit margins are off the table.
The market seems to have already priced this in. Nvidia’s stock didn’t flinch at the China concession. But the long-term risk is real: if Chinese AI models — which now account for 60% of developer usage on platforms like OpenRouter — continue closing the capability gap with Western models, the demand for Chinese-made AI chips will grow proportionally. Nvidia’s loss is Huawei’s permanent gain.
What Comes Next
Nvidia is no longer just a GPU company. It’s becoming a full-stack data center company — GPUs, CPUs, networking, software, and increasingly, the reference architecture that defines how AI infrastructure gets built. The Vera CPU launch is the clearest signal yet that Jensen Huang wants Nvidia to own the entire compute layer, not just the accelerator.
The $91 billion Q2 forecast, the $80 billion buyback, and the Vera CPU launch all point to the same conclusion: Nvidia believes the AI infrastructure buildout is still in its early innings. And when the company supplying the picks and shovels to the gold rush thinks the rush is only getting started, you either bet with them or bet against every hyperscaler, sovereign fund, and government on Earth simultaneously.
For Intel, this is an extinction-level event. For AMD, it’s a warning shot. For investors, it’s the most important earnings call of the year — not because of what Nvidia reported, but because of what it just told you it’s building next.
