Analysts expect revenue of $78.8 billion, almost 80% higher than just a year earlier. Earnings per share are projected at $1.77, nearly double last year. There’s ample reason to believe Nvidia will meet those lofty goals: according to data from The Motley Fool, the chipmaker has beaten Wall Street’s estimates in 21 of the last 23 quarters, totaling to five years of outperformance.

So the reason Nvidia’s earnings are so closely watched isn’t because anyone expects it to suddenly fail. A beat is almost a given. It’s because Nvidia is still the big winner of the AI boom, the company at the center of every hyperscaler’s capital spending plan and every investor’s portfolio anxiety. Its last earnings report in February was a 7% beat on earnings per share. The stock fell 6% that day, and was down 11% a month later, according to 24/7 Wall St.

CEO Jensen Huang has bemoaned that no-win dynamic— a stock that can do everything right and still get punished — and it’s why why investors won’t be watching the headline revenue number. They’ll be watching a less famous line, called gross margin.

What is gross margin?

Gross margin is the percentage of every sales dollar a company gets to keep after paying to make its product. So if Nvidia sells a chip for $100 and it costs $25 to make, the gross margin is 75%. The remaining $75 goes toward everything else— profits, salaries, taxes—but the 75% itself shows how much pricing power a company actually has.

For context of how large that gross margin is, a grocery store runs on gross margins around 25%. Walmart hovers near 24%. Apple, often considered one of the most profitable hardware companies in the world, sits near 46%. Microsoft, which at this point sells mostly software, sits at around 70%.

Nvidia, which sells physical chips, runs at 75%, a number you almost never see in the physical economy. It exists because Nvidia’s customers—the hyperscalers like Microsoft, Meta, Amazon and the frontier model providers like Google, OpenAI—currently have no real chip alternative. Nvidia is the first mover in the most important part of the AI supply chain; but that doesn’t mean competitor aren’t building. 

Why it matters tonight

Nvidia told investors in February to expect a non-GAAP gross margin of about 75%, give or take half a percentage point, for the current quarter. There are a few ways it could miss.

The first is pricing pressure. Hyperscalers like Microsoft and Meta have been Nvidia’s most reliable customers, but they have also been the loudest about wanting alternatives. 

Google now sells access to its in-house TPU chips and recently signed a multi-gigawatt deal with Anthropic. Amazon released its Trainium3 chip in late 2025 and claims customers can save 30% to 40% versus Nvidia, according to AWS executive Dave Brown. Microsoft unveiled its Maia 200 chip in January and is already deploying it inside Azure data centers. Meta announced four generations of its own AI processors in March, and in April,they agreed to buy millions of Amazon’s custom AI CPUs—chips that compete directly with Nvidia’s own Vera CPU. The day the deal was announced, Amazon shares hit a near record.

That doesn’t mean the hyperscalers can afford to abandon Nvidia—they can’t. But its leverage that can squeeze on the price of Nvidia’s chips.

The second is the cost of building Nvidia’s current chips. Blackwell, the 2024 chip architecture that drove roughly 70% of Nvidia’s data center compute revenue last quarter, is more complex and more expensive to manufacture than its predecessor. If those costs are rising faster than expected, the margin shrinks.

The third is product mix. If a bigger share of revenue this quarter came from Nvidia’s lower-margin products, which it used to be known for—gaming cards, older chips—the average comes down.

Analysts’ consensus sits at 74.5%, slightly below Nvidia’s own guidance. A preview from CoinDCX put the stakes well: “A gross margin print below 74.5% would be the single most bearish data point in this Nvidia earnings report regardless of headline revenue.”