'Demand Has Gone Parabolic': Nvidia's Earnings Reveal AI Data Center Shifts

Nvidia’s record earnings show surging demand for AI computing, and they reflect the changing contours of the AI data center landscape. 

The artificial intelligence computing giant reported its highest-ever quarterly revenue Wednesday, with top-line earnings of $82B marking an 85% year-over-year increase and a 20% jump from the prior quarter.

This accelerating growth, which slightly exceeded Nvidia’s projections, is expected to continue this quarter, with the firm projecting Q2 revenue of $92B.

“This was an extraordinary quarter,” Nvidia CEO Jensen Huang said Wednesday on the firm’s quarterly call with analysts. “Demand has gone parabolic.”

The vast majority of Nvidia’s quarterly inflows came through its data center segment.

Data center revenues nearly doubled, up 92% year-over-year, to $75B. Nvidia leaders said the number of data centers carrying more than 10 megawatts of its hardware has doubled over the past 12 months, now surpassing 80 sites. 

Nvidia’s skyrocketing data center sales parallel escalating AI infrastructure spending from the world’s largest tech companies, with Amazon, Microsoft, Google, Meta and Oracle planning to make more than $700B in capital expenditures this year. That spending goes toward data centers and the chips inside them, though the lion’s share is increasingly for the latter. 

The chipmaker expects that hyperscale spending will continue growing. On Nvidia's earnings call, Chief Financial Officer Colette Kress cited forecasts for Big Tech capex to exceed $1T in 2027 and reach $3T to $4T by the end of the decade.

She said this continuing growth reflects surging adoption of AI tools and services by companies across nearly every sector, a rising tide that is lifting the entire AI ecosystem. 

“AI is no longer a nice-to-have — AI is now a necessity for enhancing productivity across all industries and roles,” Kress said. “This is propelling revenue acceleration across all layers of the AI cake, including energy, chips, infrastructure, models and applications.”

Beyond hyperscale demand, Nvidia’s quarterly numbers reflected the rapid growth of smaller AI-specific cloud providers, known as neoclouds, as major demand drivers for AI data center capacity.

While around half of Nvidia’s data center revenue came from traditional hyperscalers, the rest came from neoclouds, enterprises and industrial users, with revenue from neoclouds tripling year-over-year. 

CoreWeave is by far the most prominent player in the fast-growing sector. But the neocloud ecosystem has more than 190 other operators, including Lambda, Nebius, Crusoe, Core Scientific and Nscale. 

Huang told analysts he expects more AI cloud providers to emerge and the segment to keep expanding its share of the overall market for AI computing. 

“That category is going to continue to grow at incredible pace,” he said. 

Neoclouds have carved out a niche in the cloud ecosystem by offering access to high-performance chips needed for AI and Nvidia graphics processing units in particular — demand that the major hyperscalers haven’t been able to meet on their own. Neoclouds have proven competitive with larger cloud providers by being faster to market and offering better pricing for many customers.

The companies are also increasing their market share due to growing demand for inference computing, the systems through which users interact with AI models in real time.

Unlike computing to train AI models, inference workloads often need to be close to major population centers. Neoclouds frequently offer small cloud deployments near major population centers where traditional cloud providers may not have that kind of computing power available. 

Nvidia executives told analysts the past quarter had an “inflection in inference demand,” with inference accounting for a growing share of revenue that comes increasingly from neocloud and enterprise customers.

The company's leaders said it is poised to capture the majority of demand for processing power from the neocloud segment.

Unlike hyperscalers, which often design custom, proprietary AI systems tailored to their own needs, neoclouds typically want plug-and-play systems that can be deployed as quickly as possible. Nvidia executives say it is the only chipmaker offering such turnkey, end-to-end computing systems with chips, networking and software included.

“There's a whole category of data centers that semicustom chips just don't apply because these data centers want to buy systems, they want to operate systems — they don't want to design, they don't want to build it themselves,” Huang said. “There's a whole second category of AI data centers that we serve almost uniquely.”

Despite the strong quarterly numbers and general Wall Street optimism around the firm, shares of Nvidia fell slightly in aftermarket trading Wednesday and remained down Thursday. 

Some analysts cited growing competition from other chipmakers, including hyperscalers like Amazon and Google, and concerns about manufacturing capacity as reasons for the dip.

But Direxion Head of Capital Markets Jake Behan said this kind of slide is expected following earnings from a fast-growing firm like Nvidia if it doesn’t vastly outperform Wall Street projections. Nvidia's share price has closed lower on the day following earnings for the last three quarters. 

“When expectations are this high, even a strong print that doesn’t materially raise the long-term outlook can weigh on the stock,” Behan said.