NVIDIA (NVDA, Financials) CEO Jensen Huang said the computing industry is at the early stages of an AI-driven transformation, predicting that reasoning AI and inference workloads could require exponentially more computing power in the coming years.
Speaking on the fourth-quarter results call for the firm, Huang underlined that as businesses and cloud providers increase infrastructure spending, artificial intelligence adoption is quickening.
“The demand for Blackwell is extraordinary,” Huang said. “We are at the beginning of reasoning AI and inference-time scaling, and it will consume significantly more compute power than anything we've seen before.”
Driven by great demand for AI-related processing, NVIDIA announced a fourth-quarter income of $39.3 billion, a 78% rise from the previous year. Though income was above projections by $1.17 billion, the company's profits per share came in at $0.89, above estimates by $0.04. Thanks in great part to the company's Blackwell AI chips, data center income for the quarter came to $35.6 billion.
With OpenAI, Meta (META, Financials), and xAI among the clients adopting Blackwell for AI inference and model fine-tuning, Huang said NVIDIA's AI hardware is gaining increasing acceptance across sectors. He noted that thanks to reinforcement learning and customizing, AI models are currently being improved after the first training, hence driving even more computer demand. As companies include AI-powered decision-making, the firm also mentioned expansion in corporate AI applications like fraud detection, customer support, and healthcare.
Sovereign artificial intelligence is another main engine of development according to NVIDIA. Governments, according to Huang, are creating their own artificial intelligence systems to meet local demands. As evidence of the growing worldwide emphasis on artificial intelligence capabilities, he highlighted massive AI investment programs in Europe including a €200 billion AI development plan.
With first shipments now reaching key cloud providers like Microsoft Azure, Google Cloud, Amazon Web Services, and Oracle Cloud Infrastructure, Huang said Blackwell is growing quicker than any prior NVIDIA product. Following a replacement, Vera Rubin, in early development, said the Blackwell Ultra platform, the next edition of NVIDIA's AI processors, is on target for deployment in the second part of 2025.
In the fourth quarter, NVIDIA's gross margin dropped to 73.5% despite significant income growth. Kress said the early expenses of ramping Blackwell production were projected to cause a decline. She said as production rises, margins will stay in the low 70% range; however, as manufacturing efficiency rises later in fiscal 2026, margins could improve.
Driven by ongoing increases in AI-related infrastructure expenditure, NVIDIA projects income for the first quarter of fiscal 2026 to be $43 billion, plus or minus 2%. Data centers are moving capital expenditures toward artificial intelligence computing, according to Huang, and he sees this trend continuing. With business AI, sovereign artificial intelligence, and autonomous systems offering long-term development prospects, he said NVIDIA is still in the early phases of AI adoption.
