TheChek News Flow

‘Computing has changed,’ CEO Jensen Huang says, citing agentic AI as driver of 94% profit surge

Nvidia NVDA 1.41%increase; green up pointing triangle reported a 94% increase in profit and record sales for the fourth quarter, helping ease concerns over a possible artificial-intelligence bubble that rippled through markets in recent months.

The chip maker reported net income of $43 billion, up from $22.1 billion in the year-earlier quarter, on sales of $68.1 billion, up 73% from $39.3 billion a year earlier, easily beating consensus estimates.

Analysts polled by FactSet had predicted net income of $37.5 billion and revenue of $66.1 billion for the quarter.

Data center hardware—the chips and networking equipment that Nvidia sells to AI and cloud-computing companies—accounted for 91.4% of the quarter’s sales, or $62.3 billion, and the segment’s revenue grew slightly faster than the company’s overall sales.

“The simple way to think about it is, computing has changed,” Nvidia Chief Executive Jensen Huang said on Wednesday’s earnings call with investors. “In this new world of AI, compute equals revenues…I am certain at this point that we’ve reached the inflection point” where agentic AI is upending how business is done worldwide and selling AI tools is starting to generate real profits.

With each passing quarter, the pressure grows on Nvidia—which at a market value of nearly $5 trillion is the world’s largest publicly traded company—to beat Wall Street’s expectations.

“It’s no longer enough for Nvidia to produce good quarterly results,” said Daniel Newman, chief executive of tech research and advisory firm Futurum Group. “They have to produce perfect quarterly results.”

Nvidia’s gross margins, which have been rising steadily for the past year, hit 75% in the January quarter, up from 73% a year earlier, in line with analysts’ predictions.

In recent months, investors have sent tech stocks on a wild ride as worries about the AI trade have risen, then subsided, only to resurface. Nvidia’s share price fell as low as $170.94 in mid-December but has recovered to over $196.

The biggest buyers of Nvidia’s chips include ChatGPT-maker OpenAI, Oracle, Microsoft, Meta Platforms, Google parent Alphabet and Amazon.com. In recent months, investors have grown concerned over OpenAI’s fundraising abilities and rising competition from other chip designers, including Google and makers of custom chips.

In January, The Wall Street Journal reported that Nvidia’s deal, announced in September, to invest up to $100 billion in OpenAI, was on ice. Instead, Nvidia has said it will participate in OpenAI’s latest funding round, making a smaller investment that people briefed on the situation said is in the range of $30 billion.

“They’re exposed to a subset of companies with fragile balance sheets,” Newman said.

A factor that will likely shape Nvidia’s fortunes will be the transition from AI-model training to inference, the process by which AI tools respond to queries. Training and inference require different types of computing, and as a result, different hardware.

Nvidia has for years dominated the training market with its graphics processors, known as GPUs—powerful chips capable of performing billions of simple tasks simultaneously. As more tech companies deploy AI tools in the real world, demand is expected to shift from training to inference, which relies more heavily on central processing units, or CPUs, a simpler type of data-center chip that more companies are capable of designing.

Last week, Nvidia announced a partnership with Meta that included its first major deployment of CPUs that aren’t connected in servers to GPUs, a sign that customers such as Meta need more inference-computing infrastructure to run their AI tools and other applications.

“It’s important to understand that inference equals revenues for our customers now,” Huang said. “As AI agents come into wider use, being able to quickly generate the computing tokens needed to operate them, customers have realized that their capital spending on Nvidia’s products leads to faster growth.”

In an interview following the earnings release, Chief Financial Officer Colette Kress said that Nvidia wasn’t worried about competition from custom-chip designers because the company’s latest generation of servers, known as Grace Blackwell, provides the most efficient inference computing power on the market.

“It doesn’t surprise us that others want to take a part of the great market that we’re a big part of,” Kress said. “It sounds like a good idea, but right now we’re the king of inference. That’s what we are: the king of inference.”

One factor that isn’t yet driving meaningful revenue growth for Nvidia is the Trump administration’s relaxation of restrictions on selling the company’s H200 chips in China.

Although small quantities of H200 sales have been approved, “we have yet to generate any revenue and we do not know whether any imports will be allowed into China,” Kress said on the earnings call. The company warned that homegrown Chinese chip designers are catching up and could disrupt the global AI landscape if Nvidia was prevented from bringing Chinese developers onto its computing platform.

Nvidia said Wednesday that it expected $78 billion in revenue in the current quarter, significantly higher than the $72.9 billion predicted by analysts, and gross margins of 75%, slightly higher than Wall Street’s prediction. The better-than-expected revenue and rosy guidance nudged Nvidia’s stock up less than 1% in after hours trading, to $198.39.

Nvidia faces potential headwinds if it becomes more difficult for important customers such as OpenAI to obtain financing, or if competitors that specialize in custom AI chips gain market share, said Brian Mulberry, chief market strategist for Zacks Investment Management, which owns about $300 million worth of Nvidia stock across multiple funds.

But even if demand shifts, he expects minimal impact on Nvidia’s margins in the short term because of how the company has developed best-in-class products for multiple computing categories.

“At the end of the day, they’re still the most in-demand piece of hardware in the AI market, regardless of what side of it you’re on,” Mulberry said.