NVIDIA’s Huang says world will have ‘billions of agents’ as AI boom lifts fiscal results

Jensen Huang, founder and chief executive officer of NVIDIA, said “agentic AI has arrived” as the world’s most valuable company posted another quarter of strong growth, driven by soaring global demand for AI infrastructure.

“The world has a billion human users. My sense is that the world is gonna have billions of agents,” the NVIDIA chief noted.

For the first quarter of fiscal 2027, ended April 26, 2026, the tech giant’s net income reached $58.3 billion, a 211% increase year-on-year. Revenue touched $81.6 billion, up 85% year-on-year, while data-centre revenue climbed to $75.2 billion, also a record.

The results show surging demand for AI infrastructure, particularly for the company’s Blackwell systems, which are increasingly being used to train and run advanced AI models.

The scale of the growth is striking even for NVIDIA, whose market value and influence have expanded rapidly since the launch of ChatGPT in late 2022 triggered a global race for AI computing power.

Parabolic demand

Huang noted that the latest phase of AI adoption is fundamentally different from earlier waves because systems are now capable of autonomous reasoning and task execution.

“Demand has gone parabolic. The reason is simple. Agentic AI has arrived. AI can now do productive and valuable work. Tokens are now profitable, so model makers are in a race to produce more,” Huang said during the earnings call.

Tokens are units of text or data processed by AI systems. More profitable AI services mean companies are willing to spend heavily on computing infrastructure, benefiting suppliers like NVIDIA.

The company’s strongest business remained data centres, where computing revenue rose 77% year-on-year to $60 billion while networking revenue nearly tripled to $15 billion. Networking has become increasingly important because modern AI systems require thousands of chips to work together efficiently. NVIDIA said InfiniBand revenue grew more than four-fold, driven by deployment of its latest networking technology.

Chief Financial Officer Colette Kress said NVIDIA’s customer base had broadened well beyond the large US cloud providers that initially drove AI demand. The company has now reorganised its reporting structure into two broad segments within data centres - Hyperscale and ACIE, which stands for AI Cloud, Industrial and Enterprise. The shift highlights how AI spending is spreading across industries, governments and regional cloud providers.

Kress noted that sovereign AI infrastructure is now deployed in nearly 40 countries representing $50 trillion in GDP, while the number of partner data centres exceeding 10 megawatts nearly doubled in a year to more than 80 sites.

That diversification is becoming strategically important. While hyperscale cloud companies such as Microsoft, Amazon and Google remain major customers, NVIDIA increasingly sees growth coming from enterprises, industrial users and governments building their own AI systems.

“The world is rebuilding computing for agentic AI and robotic physical AI. NVIDIA sits at the centre of these transitions,” Huang explained this broader transformation.

The company also stressed that AI infrastructure spending could still be in its early stages. Kress said analysts now expect AI infrastructure expenditure to reach between $3 trillion and $4 trillion annually by the end of the decade.

A major reason for investor optimism is NVIDIA’s effort to evolve beyond graphics processors into a full-stack AI infrastructure company. Alongside Blackwell GPUs, the company is pushing networking products, software, AI tools and now CPUs designed specifically for AI workloads.

Beyond GPUs

One of the quarter’s most significant announcements was Vera, NVIDIA’s first custom CPU designed for “agentic AI”. Huang said the launch opens a “brand new $200 billion TAM” for the company.

The company expects nearly $20 billion in CPU revenue this year and plans to begin production shipments of its next-generation VeraRubin systems in the second half of 2026. Kress said VeraRubin could deliver up to 35 times higher inference throughput than Blackwell.

“VeraRubin is going to be even more successful than Grace Blackwell at this point. Every single frontier model company will jump on VeraRubin from the get go,” Huang noted.

Despite the strong momentum, the results also highlighted several challenges.

One major uncertainty remains China. Although the US government has approved licences for H200 chips to be shipped to Chinese customers, NVIDIA said it had not yet generated revenue from those approvals and remained uncertain whether imports would ultimately be allowed. The company excluded any China data-centre compute revenue from its guidance.

Supply constraints are another concern. Huang acknowledged that demand for VeraRubin may outstrip supply for an extended period, while Kress warned that NVIDIA was “not immune to supply challenges”.

Operating expenses are also rising rapidly. NVIDIA now expects full-year operating expenditure growth in the upper-40% range, driven by higher research spending and greater use of AI productivity tools. Consumer demand in parts of the edge computing business also weakened modestly because of higher memory and system prices.

Still, the financial position remains exceptionally strong. Free cash flow rose to a record $49 billion from $35 billion in the previous quarter. NVIDIA also announced an additional $80 billion share repurchase programme and increased its quarterly dividend to $0.25 per share from $0.01.

For the next quarter, NVIDIA forecast revenue of $91 billion, plus or minus 2%.

The company increasingly describes its products as the foundation for “AI factories” that generate intelligence in the same way traditional factories produce goods.

“The buildout of AI factories - the largest infrastructure expansion in human history - is accelerating at extraordinary speed,” Huang said.

And NVIDIA continues to sit at the centre of one of the fastest technology spending cycles in modern history.