Explosive growth, annual revenue of 1.3 trillion yuan. NVIDIA: China's competitors are making progress, H200 in China has not yet generated revenue! Jensen Huang: AI has reached a turning point; computing power equals revenue.

Reporter |Yang Hui

Editor******|Cheng Peng Chen Junjie Du Hengfeng Proofread******|Chen Kemin****************

On February 25th, local time, NVIDIA announced its financial results for Q4 and the full fiscal year 2026. Notably, the data center performance saw exceptional growth. In Q4, NVIDIA’s data center revenue reached $62.3 billion, a record high, with full-year revenue of $193.7 billion (approximately RMB 13,251), up 68% year-over-year.

During the subsequent earnings call, NVIDIA CEO Jensen Huang and CFO Colette Kress repeatedly emphasized that AI (artificial intelligence) has reached a turning point, with explosive growth in computing demand. In this new era, computing power equals revenue.

Huang also revealed during the call that the infrastructure based on Blackwell (NVIDIA’s most advanced AI chip) deployed by major cloud service providers, large-scale cloud companies, AI model developers, and enterprise clients has reached 9 billion watts and is operating at full capacity. Additionally, earlier this week, NVIDIA delivered the first Vera Rubin samples to customers, with mass production plans on track for the second half of 2026.

Regarding the industry’s widespread concern about NVIDIA’s previous-generation H200 chip, senior executives stated during the earnings call that exports of the H200 product to Chinese customers have not yet generated revenue. NVIDIA also mentioned that a domestic Chinese competitor recently completed its IPO, strengthening its capabilities and potentially causing long-term impacts on the global AI industry landscape.

Record Data Center Revenue Raises Questions:

“Is demand too concentrated?”

Data shows that NVIDIA’s revenue for Q4 FY2026 hit a record $68.1 billion, up 20% quarter-over-quarter and 73% year-over-year. Net profit was $42.96 billion, an increase of over 90%, with diluted earnings per share of $1.76.

Meanwhile, NVIDIA disclosed its full-year FY2026 results: revenue of $215.9 billion, up 65%; gross margin of 71.1%; net profit of $120.067 billion, with diluted EPS of $4.90.

Breaking down by business, most of NVIDIA’s revenue comes from its data center segment, which has been boosted by massive investments in AI infrastructure across the tech industry. In Q4, data center revenue reached $62.3 billion, a new high, up 22% quarter-over-quarter and 75% year-over-year; annual revenue was $193.7 billion, up 68%.

However, concerns about “demand being overly concentrated” have persisted since last year.

During the earnings call, Kress also noted that the top five global cloud providers and large-scale cloud companies contributed over 50% of NVIDIA’s data center revenue in FY2026, fueling worries on Wall Street about demand being too centralized.

In response, Kress emphasized that NVIDIA still believes that the shift from traditional data center workloads to GPU-accelerated computing, along with AI empowering existing large-scale cloud workloads, will generate about half of the company’s long-term growth opportunities. Just like electricity and the internet, every country will develop and operate its own AI infrastructure.

Huang added that NVIDIA’s customer base is diverse, including AI model developers and platform providers, with increasing global diversification. This trend will positively impact the company’s ecosystem development.

When asked by analysts, “What if customer capital expenditures stop growing?” Huang responded confidently that he is optimistic about customers’ cash flow growth. “The reason is simple: AI has reached a development inflection point, and its practicality has been widely validated across global enterprises, leading to explosive growth in computing demand. In this AI era, computing power equals revenue—if computing power grows, so does revenue. Currently, there is about $300-400 billion in available market capital, which is substantial. This is now shifting toward AI. To generate tokens in AI, you need computing power, which directly translates into growth and revenue.” Huang also revealed that NVIDIA is close to signing a cooperation agreement with OpenAI and is excited about a long-term partnership.

Other segments saw strong growth: Q4 gaming revenue was $3.7 billion, up 47% year-over-year but down 13% quarter-over-quarter; professional visualization revenue was $1.3 billion, up 159%; automotive and robotics revenue was $604 million, up 6%. For FY2026, gaming revenue reached $16 billion, up 41%; professional visualization $3.2 billion, up 70%; automotive and robotics $2.3 billion, up 39%.

NVIDIA’s H200 for China Has Not Yet Generated Revenue

Another focus of this earnings report is the ramp-up of NVIDIA’s most advanced AI chips, Blackwell and Vera Rubin.

Huang revealed that the infrastructure based on Blackwell deployed by major cloud providers, large-scale cloud companies, AI model developers, and enterprise clients has reached 9 billion watts and is operating at full capacity. Regarding Rubin, NVIDIA stated that earlier this week, the first Vera Rubin samples were delivered to customers, with mass production planned for the second half of 2026. Although it’s still uncertain how much of the early ramp-up will be achieved this year, market demand and interest are very strong.

As for the widely discussed H200 chip, senior executives said during the earnings call that exports of the H200 to Chinese customers have not yet generated revenue, and it remains uncertain whether this product will continue to enter China in the future.

NVIDIA also mentioned that a domestic Chinese competitor recently completed its IPO, strengthening its capabilities and potentially causing long-term shifts in the global AI landscape. To maintain its leadership in AI computing, the U.S. needs to attract developers worldwide and become the preferred platform for all commercial entities, including Chinese companies.

Additionally, shortages of storage chips and power supplies are key concerns among analysts. Kress stated that all data centers face power constraints, and given this situation and the goal to maximize AI factory revenue, customers are making critical architectural decisions based on watts-per-performance. Huang noted that while the company hopes for more supply, supply constraints are expected to be tight over the coming quarters.

Furthermore, during the earnings call, an analyst raised the hot topic of “space data centers.” Huang responded that currently, space data centers are not very cost-effective, but this will improve over time. AI applications in space are expected to have promising prospects in the future.

(Disclaimer: The content and data in this article are for reference only and do not constitute investment advice. Please verify before use. Risks are borne by the user.)

******|Daily Economic News nbdnews Original Article|

Unauthorized reproduction, excerpting, copying, or mirroring is prohibited