Nvidia posts a record quarter, a big forecast — and dares the market to flinch

Nvidia $NVDA’s latest quarterly earnings report arrived the way those do now: as a global group project, graded in public, by an audience that has already decided what an A looks like. So the company walked into the room and tried to out-A the A.

For months, the market’s relationship with Nvidia has looked like a never-ending treadmill sprint: The quarter has to be enormous, the guide has to be bigger, and the commentary has to sound inevitable while still somehow offering something new. On Wednesday, Nvidia delivered a record quarter and then, more importantly, handed Wall Street a forward number that had some jaws dropping.

For Q4 FY26, Nvidia posted record revenue of $68.1 billion, up 20% from the prior quarter and 73% from a year earlier, with GAAP diluted EPS of $1.76. The engine room stayed the engine room: Data Center revenue hit $62.3 billion, up 22% sequentially and 75% year over year.

Nvidia shares initially rose about 3% in after-hours trading, a move that suggested relief more than shock — the kind of response you get when the company clears the bar and then the bar starts getting raised again. It beat the quarter, it beat the forward quarter harder, and it managed to do both while keeping margins pinned near the number investors have treated as a proxy for pricing power.

The market, predictably, cared most about the part that hasn’t happened yet. Nvidia told investors it expects $78 billion in revenue for the current quarter (plus or minus 2%) — a number that landed above what Wall Street had been modeling: revenue expected at $66.16 billion and first-quarter revenue expected at $72.46 billion. That’s a couple of billion above the quarter’s baseline and more than five billion above the Street’s plain-vanilla next-quarter expectation — the kind of gap that turns “beat and raise” into “reset the math.”

Nvidia’s AI boom has matured into something that looks a lot like industrial logistics. The company is shipping more than just chips; it’s shipping systems. And the supporting cast (networking, interconnects, full-stack integration) is starting to show up in the numbers with a little pep in its step. In the quarter, networking revenue was nearly $11 billion, up 34% sequentially and 263% year over year, with Nvidia pointing to the ramp of its NVLink compute fabric for GB200 and GB300 systems alongside Ethernet and InfiniBand growth.

Margins — the other religion — held up. Nvidia reported GAAP gross margin of 75% for the quarter (non-GAAP 75.2%), up sequentially as Blackwell ramped and inventory provisions eased. But the full-year picture is messier: FY 2026 GAAP gross margin was 71.1%, down from 75% in fiscal 2025, a reminder that scaling a hardware transition at this size tends to come with some financial scuff marks.

Sure, the market had already decided the quarter was going to be strong. The pre-earnings anxiety wasn’t really about whether Nvidia would beat. It was about whether it could still surprise anyone in a world where surprise has been priced, modeled, hedged, and litigated in advance.

The company’s post-earnings framing leaned hard into inevitability. CEO Jensen Huang, never a man to understate a computing cycle, said in the press release that “computing demand is growing exponentially — the agentic AI inflection point has arrived.” He added, “Grace Blackwell with NVLink is the king of inference today — delivering an order-of-magnitude lower cost per token — and Vera Rubin will extend that leadership even further.” Nvidia isn’t interested in soothing the market’s nerves. Nvidia is interested in setting the frame. And you can roll your eyes at the rhetoric and still respect the receipt: $193.7 billion in full-year Data Center revenue, up 68% year over year, will buy a lot of narrative confidence.

The guide does most of the talking

Nvidia’s outlook is what, even ahead of time, was turning this earnings report into a market event. The company guided first-quarter revenue to that $78 billion (± 2%) — and then stapled a geopolitical caveat directly to the number: The company isn’t “assuming any Data Center compute revenue from China” in that outlook.

That one sentence does two jobs at once: It boxes the export overhang into a disclosed assumption, and it quietly preserves China as an upside lever that doesn’t need to be promised, modeled, or defended tonight. Nvidia has used versions of this tactic before, but the explicitness here matters because the market has spent the past year treating China like a ghost in the machine: always present, never fully quantified, ready to spook guidance when investors are already jumpy.

If guidance was the dare, margins were the reassurance. Nvidia said fourth-quarter gross margin was 75% GAAP and 75.2% non-GAAP. For the first quarter, it guided 74.9% GAAP and 75% non-GAAP, plus or minus 50 basis points. 

That matters because margin has become the market’s lie detector. The Nvidia story isn’t just “sell more chips.” It’s “sell more chips at a level of profitability that implies you still have pricing power even as your customers get bigger, smarter, and more motivated to negotiate.” When investors show up looking for cracks — memory costs, mix shifts, ramp friction — a steady mid-70s margin narrative is the cleanest way to deny them any sort of an easy hook.

And then there’s the fun-house-mirror detail: “other income.” Nvidia reported $5.6 billion in net other income for the quarter, driven by unrealized gains in equity holdings, including gains tied to its previously announced Intel $INTC investment. When your operating income is $44.3 billion in a single quarter, even the side quests show up with 11 figures and a nice little smirk.

Then Nvidia slipped in the kind of detail that won’t lead the next-day headlines but will absolutely show up in the morning analyst notes: Beginning in the first quarter of fiscal 2027, it will include stock-based compensation expense in non-GAAP financial measures. In other words, Nvidia is tightening its own definition of “adjusted.” In the same outlook section, it also quantified how much SBC it expects to carry: $1.9 billion within non-GAAP operating expenses, and a 0.1% impact on non-GAAP gross margin.

That’s how a company can signal confidence without throwing confetti or getting glitter everywhere. Nvidia is saying it can keep producing blockbuster numbers even with a less forgiving scorecard — while also acknowledging, in the most CFO-friendly way possible, that talent is expensive and it plans to keep paying for it.

What Nvidia’s earnings say about the AI boom

The question hovering over this earnings season has been whether the AI buildout is starting to produce the kind of returns that make its own spending feel rational — to boards, to investors, and to the same hyperscalers whose capex plans have started to unsettle the market (and then some).

Nvidia’s report won’t settle the “boom/bust” argument, because one company’s quarter can’t guarantee an entire industry’s economics. (Although, if there’s one company’s quarter that could…) But Wednesday’s earnings do offer up a data point: The AI hardware demand Nvidia serves isn’t fading away. Nope. It’s actually accelerating loudly enough that Nvidia is willing to guide to $78 billion next quarter while excluding China data center compute revenue from the math.

If you’re looking for the most telling line item, it’s still that $62.3 billion quarterly Data Center revenue; the AI boom is a revenue base. Nvidia said hyperscalers were “slightly over 50%” of Data Center revenue, with growth led by the rest of its Data Center customers as revenue diversified. The big buyers stayed big, but the long tail apparently got longer. The market can argue about who wins the next layer — custom silicon, ASICs, alternative accelerators, the pricing of inference — but the center of gravity remains Nvidia-sized.

And the company is still playing defense against its own success. Inventory rose to $21.4 billion, and Nvidia disclosed $95.2 billion in total supply-related commitments — a number that reads like a company trying to make sure it never has to tell a customer, “Sorry, ask again next quarter.” Nvidia said it has “strategically secured inventory and capacity to meet demand beyond the next several quarters.”

And Nvidia, never shy about the theatrical, leaned into the broader story with shareholder returns: $41.1 billion returned in fiscal 2026 via repurchases and dividends, with $58.5 billion remaining under its buyback authorization. In a market that has started to treat AI spending as a test of discipline, a number like that reads that the factories are humming — and the cash is very real.

Wall Street asked Nvidia to make the future feel faster than the past. Nvidia answered with a forecast big enough to force the market to react — and careful enough to show it knows exactly what the market is afraid of. The company handed the market a bigger “next quarter” number and backed it with signs that the plumbing — networking, systems, capacity commitments — is scaling alongside the silicon.

The market’s job, as always, is to ask the only question it still knows: Fine. Can you do it again — and even faster?