Nvidia did its usual thing again, the company managed to exceed analysts’ expectations on practically every important line. Revenue rose 73 percent to $68.1 billion on a year-over-year basis, setting a new record. This exceeded the analyst consensus of $66.2 billion by nearly $2 billion.
The engine of growth is still the data center business, which generated $62.3 billion after growing 75 percent and now accounts for more than 91 percent of total revenue. It’s worth checking out
an acceleration in growth can be observed both for total revenues and for the data center business, even though the past quarters were about normalization.
The growth of the data center business was primarily driven by the fact that large cloud providers (hyperscalers) continue to build their AI infrastructure at full speed, and this is directly reflected in Nvidia’s order book. The CFO also highlighted that hyperscalers remained their largest customer group and accounted for “slightly more than half” of the business revenue, meaning that spending in the sector is still the company’s most important growth engine.
The key to success is that the company doesn’t just make chips, but offers a complete rack-level infrastructure – hardware, software and network connection from a single source. This strategy is also strengthened by the announcement of the new Rubin platform, which will begin shipping in the second half of the year. Compared to Blackwell, this system promises a tenfold cost reduction in artificial intelligence model running (inference), responding to global energy and capacity constraints.
The main product in the current cycle is the Blackwell generation, including the Grace Blackwell rack-scale configuration. The importance of this lies not only in the fact that this platform carries out the training, but also in the fact that the focus is increasingly shifting towards inference, i.e. when the models are already running and “producing” in a production environment. Accordingly, Jensen Huang repeatedly returned to the conference call „compute equals revenue” to think: the essence of the message is that with the spread of agentic AI, computing capacity can be translated more and more directly into business output and income,
therefore, according to him, the demand is not simple hype, but a structural trend.
Another spectacular, but often underestimated, element of the growth of the data center business is the rise of network components. In the data center business, Nvidia reported $10.98 billion in sales from sales of the company’s networking components, which are used to connect hundreds of graphics processing units. Network components sales were up 263% year-over-year, reflecting strong adoption of the company’s NVLink networking technology and Spectrum-X Ethernet switches, which won new contracts with giants such as Meta.
While it’s all about the data center business, let’s take a quick look at the other segments.
- By Nvidia gaming division, which used to be the company’s largest segment, registered a 47% increase in revenue compared to the previous year, reaching $3.7 billion. According to analysts, Nvidia may skip the introduction of new gaming GPUs this year as memory chip shortages force chip makers to prioritize AI processors. For Nvidia, this means that most AI accelerators will be sold in rack-sized systems, such as the 72-GPU Grace Blackwell. The lack of memory is a concern for investors. The company expects supply constraints to hamper Nvidia’s gaming business in the first quarter of the year and beyond.
- The in the automotive segmentwhich includes chips for cars and robots, Nvidia reported quarterly sales of $604 million, up 6% from a year earlier, but short of analysts’ expectations of $654.8 million.
- Professional visualization business, Nvidia reported quarterly sales of $1.32 billion, up 159% year-over-year and beating expectations of $755.4 million.
Net profit jumped by 79 percent to $39.6 billion, and earnings per share increased by 82 percent to $1.6. Both exceeded analysts’ expectations.
It looks like there is no need to worry about margins, as the company’s net profit margin returned to a historic high in the current quarter. The gross margin was around 75 percent, which shows a slight decrease on an annual basis due to the rise of new products and the costs of the supply chain.
The biggest surprise was the forecast: the management expects revenue of 78 billion dollars for the next quarter,
which is much higher than the 72.6 billion estimated by analysts.
It is particularly noteworthy that the calculations they do not include Chinese data center revenues. The $78 billion revenue would correspond to a 77 percent increase on a year-over-year basis, which would mean that the re-acceleration of Nvidia’s revenues would not be just a quarterly story (since the closed quarter had a 73 percent increase).
The chipmaker’s upbeat forecast coincides with ramping up production of the Vera Rubin solution, which will succeed Grace Blackwell as the next rack-size artificial intelligence system. According to Nvidia, the system’s 72 next-generation Ruby graphics processing units (GPUs) are expected to deliver 10 times more power per watt than its predecessors. This increase in efficiency, i.e. how many tokens I get per dollar, is one of the most important parameters for investors.
CFO Colette Kress said after the report that the company shipped the first Vera Rubin samples to customers this week, and that Nvidia expects all model makers and cloud providers to eventually adopt the system. He said the company now expects this year’s growth to exceed last year’s forecast of $500 billion in revenue opportunity between Blackwell and Rubin.
The key message for investors is that the demand side remains stable as the investment cycle of tech giants – Alphabet, Microsoft, Amazon and Meta – has not broken. Their infrastructure expenditures in 2026 could reach 635-665 billion dollars, of which Nvidia has a disproportionate share. Meanwhile, the emphasis is increasingly shifting from training models to real business use (inference).
In this area, system efficiency is the decisive factor, in which Nvidia’s ecosystem seems unbeatable at the moment.
At the same time, the risks are concentrated, the narrowing of the customer circle is spectacular: according to reports, two customers account for 36 percent of the revenue. Although capital-strong partners mean stability in the short term, in the longer term it increases exposure and the bargaining power of clients. The recently announced cooperation between Meta and AMD is also a warning sign. This indicates that the strategic goal of the largest customers is to build a multi-supplier model and reduce dependence, which the market may over time price into Nvidia’s growth prospects. Several large companies develop their own in-house chips, which partly aims to become independent from Nvidia.
The stock market reaction after the report remained moderate, which is a typical trap of peak expectations. After fourteen quarters of continuous positive surprises, investors expect not only good performance, but an outstanding sensation. Nvidia’s share price rose 0.2 percent after the close.
Cover image credit: Jonathan Raa/NurPhoto via Getty Images
This article does not constitute investment advice or investment recommendation. Detailed legal information
