Nvidia earnings takeaways: Data center revenue nearly doubles, report is strong but stock slides

The CNBC Investing Club raised its price target on Nvidia shares following the beat-and-raise quarter. In a reaction note for members Wednesday night, Jeff Marks, the Club's director of portfolio analysis, explained why the muted reaction in extended trading is not shaking their faith in the company's stock.

"This was an extraordinary quarter. Demand has gone parabolic," Huang said as he closed out the earnings call. "The reason is simple: Agentic AI has arrived."

Huang then summarized his own takeaways from the quarter's results.

1. Nvidia's involvement with the world's largest AI models is growing: "Nvidia is the only platform that runs every frontier AI model," Huang said, mentioning Anthropic, OpenAI, SpaceXAI, Meta and Google's Gemini.
2. The chipmaker is powering every hyperscaler, "supporting their core data processing and machine learning workloads, internal AI services, as well as supporting their demand for Nvidia users in the public cloud services," Huang said.
3. The company's broad range of products is helping it break into "new AI data center segments, new AI-native clouds and sovereign AI clouds, and on-premises enterprise and industrial infrastructure."
4. Physical AI is "the next wave," Huang said, adding that Nvidia's CUDA software platform helps it extend into autonomous vehicles and robotics, in particular.
5. The Vera CPU is a "major new growth driver," opening up a $200 billion revenue opportunity if the company can capture the market.

"The world is rebuilding computing for agentic AI and robotic physical AI," Huang said in closing. "Nvidia sits at the center of these transitions."

"We built it ahead of this moment so that when agentic AI arrived, Nvidia would be ready," he added. "It has arrived."

—Katie Tarasov

Nvidia acknowledged in a 10-Q filing that the data center landscape has changed, with its own customers potentially becoming semiconductor rivals as they develop custom computer chips tailored for their own uses.

"Some of our customers are developing their own ASICs and other products, including designs optimized for certain workloads that may not require all of the features and functionality our data center systems provide," Nvidia said in the filing, referring to a category of computer chips known as application-specific integrated circuits.

Although Nvidia did not name the customers, tech hyperscalers like Google, Amazon, Meta and Microsoft have all been developing their own custom ASICs. Meta, for instance, showcased in March four custom chips that the social media giant designed that Taiwan Semi will manufacture.

Google, on the other hand, has been developing its own AI-specific silicon dubbed tensor processing units, and on Monday confirmed that it would create a new AI infrastructure company centering around those computer chips. As part of that yet-to-be-named AI firm, asset-management giant Blackstone will invest $5 billion in equity capital.

"This new company has enormous potential as it helps to meet the unprecedented demand for compute," Jon Gray, President and COO of Blackstone, said at the time in a statement.

In addition to tech giants developing their own custom AI chips, Nvidia said in the filing that other unspecified customers "may offer cloud-based services that compete with our AI cloud service offerings, and we may not be able to establish market share sufficient to achieve the scale necessary to meet our business objectives."

"If we are unable to successfully compete in this environment, demand for our products, services, and technologies could decrease, which may negatively impact our business," the company said in the financial filing.

Besides Nvidia's frenemies potentially taking market share, the chip giant noted in the filing that the firms could also "impinge on our ability to procure sufficient foundry capacity and scarce input materials during a supply-constrained environment, which could harm our business."

—Jonathan Vanian

Nvidia's custom AI chip that was the result of its $20 billion acquisition of Groq's tech "will be a niche product for some time," Huang said on the earnings call Wednesday.

That's despite signs that Big Tech is hungry for alternatives to the company's costly, sold-out GPUs.

Nvidia unveiled a new Groq language processing unit at its annual GTC conference in March, as well as an entire data center rack filled with the custom chips, called LPX.

"LPX is designed for low latency and high token rate, but its throughput is low," Huang said. "The use case for LPX is not broad."

The new Groq chips in LPX are an example of what's known as application-specific integrated circuits, or ASICs. They're lower power chips programmed for more specific tasks.

Custom ASIC maker Cerebras Systems made a blockbuster debut on Thursday, in the clearest signal yet that the market is hungry for an alternative to Nvidia's GPUs.

ASICSs are also made in-house by the likes of Google, Amazon, Meta and Microsoft, while an increasing number of startups like SambaNova and D-Matrix are entering the space.

— Katie Tarasov

Huang explained the company's decision to overhaul its quarterly earnings that will break the reporting into two segments: Data center, which has two subgroups, and edge computing.

"On the segmentation and the description of the business, we wanted you to understand our business better," Huang told analysts during the earnings call.

Huang pointed to AI as the nexus of the company's overall business, but some companies have vastly different technology approaches and initiatives than others.

While hyperscalers like Meta and Alphabet might be buying tons of computing gear and related GPUs as part of their gigantic data center build-out plans, governments and smaller businesses have different computing requirements, as do the companies researching cutting-edge robotics.

"It's the simplest way of understanding our business," Huang said. "Each one of them have different stacks in a lot of ways. They have different operating systems. They operate in a different way, and we go to market very differently in each one of them."

— Jonathan Vanian

Nvidia expects its next rack-scale system for AI, Vera Rubin, will be "even more successful than Grace Blackwell," Huang said on the earnings call.

The primary reason, Huang added, is because "we're growing share in inference very, very quickly" as the number of frontier model companies grows. He cited Anthropic as a key new customer this year, providing secure compute for Microsoft Azure, Amazon Web Services and CoreWeave.

The Vera Rubin system is comprised of 1.3 million components, including 72 Rubin GPUs and 36 Vera CPUs. Nvidia says it delivers 10 times more performance per watt than its predecessor, Grace Blackwell.

Huang said he anticipates that Nvidia "will be constrained throughout the entire life of Vera Rubin."

— Katie Tarasov

Nvidia is aiming to become the "world's leading CPU supplier," Chief Financial Officer Colette Kress said on the earnings call. It's a lane currently dominated by competitors Intel and Advanced Micro Devices.

The company's new Vera central processing unit "opens a brand new $200 billion tab for Nvidia," Kress said, adding that "every major hyperscale and system maker is partnering with us to get it deployed."

She anticipated $20 billion in total CPU revenue this year.

Until now, Nvidia has led the AI boom with its graphics processing units that excel at the parallel math necessary for training large models. But as agentic AI shifts compute needs, the once-quiet CPU market has seen a major resurgence.

Nvidia unveiled an entire CPU-only rack at its annual GTC conference in March.

— Katie Tarasov

Hyperscalers made up more than half of all data center revenue, reaching $38 billion and increasing 12% quarter-over-quarter, CFO Colette Kress said on a call Wednesday.

The other $37 billion portion of data center revenue comes from a segment Nvidia is calling ACIE. It's tied to AI clouds, industrial and enterprise markets. Kress said AI cloud revenue more than tripled year over year.

Nvidia is now enabling "rapid stand-up of AI compute capacity" in more than 80 data centers that exceed 10 megawatts, Kress said.

— Katie Tarasov

Nvidia revealed Wednesday in a 10-Q filing that its global supply chain has yet to experience any significant impact from instability in the Middle East due to the Iran war, which began in February.

"Our global supply chain for our networking products, including our Israel operations of approximately 5,900 employees supporting research and development, operations, and sales and marketing, has not been significantly impacted by the conflict in the Middle East," the company said in the filing.

Still, Nvidia noted that "If the conflict escalates or extends, it could affect future product development, supply chain, and revenue, and create business uncertainty."

— Jonathan Vanian

CFO Colette Kress told analysts on the earnings call that the build-out of AI factories is accelerating and the value of the chipmaker's infrastructure is rising.

Kress said the price of renting an H100 has risen 20% year to date, and A100 cloud pricing is up nearly 15%. She said customers are generating profitable revenue beyond the depreciable life of their GPUs.

— Chris Eudaily

Nvidia said in a related corporate filing that the company invested $18.6 billion in private companies and infrastructure funds during its fiscal first quarter.

Regarding that figure, Nvidia noted in the filing that "Some of these investments include AI model makers that may indirectly purchase or use our products in the cloud."

—Jonathan Vanian

Gene Munster of Deepwater Asset Management told CNBC's "Closing Bell Overtime" that Nvidia's revenue acceleration this quarter was "remarkable," although he notes it's still "working through some of the noise" when it comes to sales in China.

Tony Wang of T. Rowe Price added that agentic AI will be adding to how much compute we need, adding to the excitement.

—Katie Tarasov

The conference call with analysts is scheduled for 5 p.m. ET. The stream is accessible here.

—Chris Eudaily

Revenue in Nvidia's data center unit, which houses the company's market-leading GPUs and AI systems, doubled from a year earlier to $75.2 billion, accounting for 92% of total sales. That beat the average analyst estimate of $73.13 billion, according to StreetAccount.

—Ari Levy

Nvidia said it has updated its quarterly earnings report into the following two market buckets: Data center and edge computing.

The company's data center unit counts hyperscale-related revenue, and sales that are part of the ACIE division tied to AI clouds, industrial and enterprise markets. Nvidia's data center unit logged $75.2 billion in its fiscal first-quarter, representing a 92% jump from the previous year.

The chip giant's new edge computing segment includes data processing devices for agentic and physical AI, which it said includes gear like personal computers, video game consoles, robotics and automotive. Nvidia's edge computing group posted $6.4 billion in fiscal first quarter sales, representing a 29% year-over-year increase.

Prior to the AI boom, gaming was Nvidia's most important business, with more than half revenue coming from gaming in fiscal 2020 and only 27% coming from data center.

Now those numbers have flipped. Data center made up more than 90% of revenue last fiscal year, with gaming accounting for less than 8%.

—Jonathan Vanian

Nvidia's gross margin came in at 75%, meeting analysts' expectations, according to StreetAccount. That's in-line with the fiscal fourth quarter and up from an adjusted number of 71.3% a year ago. Nvidia's margin strength shows that it's able to maintain pricing power despite increased competition.

The company said its gross margin will hold at the same level in the current quarter.

—Ari Levy

Nvidia gave strong forward guidance, predicting $91 billion of revenue in its current fiscal second quarter. That's well above the average quarterly revenue estimate of $86.84 billion, according to LSEG.

Nvidia said it is not assuming any data center compute revenue from China in the outlook.

—Katie Tarasov

Nvidia will return more of its cash to shareholders. The chipmaker's board has authorized $80 billion for share repurchases. The company is also bumping up its quarterly cash dividend to 25 cents per share from 1 cent previously.

Nvidia generated $48.6 billion in free cash flow in the period, up from $34.9 billion in the prior quarter and $26.1 billion a year ago.

The dividend itself isn't new, having become available to investors in 2013. In August, Nvidia authorized $60 billion for repurchases.

— Jordan Novet

Nvidia said that net income rose to $42.96 billion, or $1.76 per share, from $18.8 billion, or 76 cents per share, a year earlier.

—Jonathan Vanian

"The buildout of AI factories — the largest infrastructure expansion in human history — is accelerating at extraordinary speed," Nvidia CEO Jensen Huang said in a statement. "Agentic AI has arrived, doing productive work, generating real value and scaling rapidly across companies and industries."

—Jonathan Vanian

Nvidia's revenue jumped 85% to $81.62 billion from $44.06 billion a year earlier, the company said.

—Chris Eudaily

Nvidia's earnings are expected to show booming sales of its current Grace Blackwell rack-scale system. But all eyes are on its next AI system, Vera Rubin, which CNBC got an exclusive first look at in February.

Later this year, top Nvidia competitor Advanced Micro Devices is expected to start shipments of its first rack-scale system, Helios, to compete directly with Vera Rubin.

—Katie Tarasov

Nvidia shares have tumbled after the chip giant's three most recent financial results. Despite huge demand for its AI chips, analyst expectations may have hit unattainable highs.

Nvidia beat expectations in 18 of the last 20 quarters, yet its stock fell 5% after reporting fiscal fourth quarter results in February. It was down 3% and 0.8% following the previous two reports.

The last time Nvidia saw a double-digit stock move in reaction to earnings was more than two years ago, in early 2024.

Nvidia's last miss on earnings per share and revenue came in 2022.

—Katie Tarasov

Investors are positioning ahead of Nvidia's earnings drop, and there are some unusual moves in the options market.

Options are implying a 5% to 7% move, with a lot of the action leaning bullish.

—Chris Eudaily

Memory prices are up across the board as the global shortage continues, and analysts are watching to see if this could prove a material headwind for Nvidia.

Insatiable demand for AI has largely exhausted supply of what's known as high-bandwidth memory, or HBM, from leading memory makers Samsung, SK Hynix and Micron.

HBM is made by stacking large amounts of Dynamic Random Access Memory, or DRAM, that enables fast, temporary data storage so chips can run parallel tasks. This is especially crucial for the high-power graphics processing units that make up the majority of Nvidia's business.

So far, consumer electronics have borne the brunt of the DRAM shortage, as chipmakers like Nvidia reserve their limited supply for AI chips instead of those used in less powerful devices. Gartner predicts PC prices will rise by 17% this year.

According to data from Counterpoint Research, the DRAM market has recorded 30% quarter-over-quarter growth for two consecutive periods, driving memory stocks to be some of the market's top performers this year.

—Katie Tarasov

Nvidia's revenue explosion in recent years has been mirrored by an expanding profit margin, underscoring the company's pricing power, supply chain efficiency and an increase of software in the mix.

From a gross margin in the mid-60s five years ago, Nvidia has lifted that number into the mid-70s today, and it's one that investors watch closely, particularly with the whole tech industry facing soaring memory costs from a worldwide shortage.

Analysts expect Nvidia to record a gross margin, or the profit left after accounting for the cost of goods sold, of 75% in the latest quarter. That would be in-line with the fiscal fourth quarter and up from an adjusted number of 71.3% a year ago.

Cantor analysts said in a report last week that they're expecting to see gross margin come in at 75.1%, "based on better volumes of rack-scale solutions and continued Blackwell cost improvements."

—Ari Levy

In the years before the generative AI craze, Nvidia was primarily known for its gaming chips. In fiscal 2020, over half its revenue came from gaming, while just 27% came from the data center.

Fast forward to the present, and Nvidia is almost exclusively a data center company. That part of the business made up 90% of revenue last fiscal year, and there seems to be no stopping it.

Nvidia's Grace Blackwell rack-scale systems have long been sold out, and Wall Street is now watching for the ramping up shipments of its next system, Vera Rubin.

Meanwhile, gaming is now less than 8% of Nvidia's business. And as CNBC reported last month, Nvidia's relentless focus on data center clients has left gamers feeling betrayed, a reality that's become even more apparent of late due to a global memory shortage that's led Nvidia to prioritize Blackwell and Rubin over GeForce gaming GPUs.

—Katie Tarasov

One big area of uncertainty is China, specifically pertaining to Nvidia's older Hopper GPU, known as H200. Huang was a last-minute addition to President Donald Trump's China summit last week, but the visit did little to clear up whether H200 sales will be permitted in the country.

Huang said at Nvidia's GTC conference in March that Nvidia had received H200 orders from China.

"We're in the process of restarting our manufacturing," Huang told reporters at the event in San Jose, California. 

Reuters reported last week that a handful of Chinese companies have been approved by the U.S. Commerce Department to purchase H200s, including Alibaba, Tencent, ByteDance and JD.com.

But one U.S. trade representative said chip export controls were not discussed in the China talks last week, suggesting a major breakthrough on H200 sales may not be close.

China once accounted for at least one-fifth of Nvidia's data center revenue, but the company has been shut out of the country since being told by the Trump administration in April that it would require a license to export chips there and to a handful of other countries.

—Katie Tarasov

Cerebras Systems' monster Nasdaq debut last week was a clear signal that tech giants are hungry for alternatives to Nvidia's costly (and sold out) GPUs. The company's market cap swelled to almost $100 billion on its first day of trading.

Cerebras makes a different type of chip, known as a custom ASIC — application-specific integrated circuit — that's been gaining ground as agentic AI shifts compute needs toward inference. While GPUs excel at the parallel math necessary for training large models, inference can happen on less powerful chips programmed for more specific tasks.

It's an increasingly crowded space, with in-house ASICs now made by the likes of Google, Amazon, Meta and Microsoft. Cerebras operates its dinner-plate-sized chips inside its own data centers, pitting it against cloud providers Google, Microsoft, Oracle and CoreWeave.

Nvida also makes custom ASICs in-house after spending $20 billion to acquire Groq's technology in December, and then announcing custom Groq Language Processing Units at GTC in March.

—Katie Tarasov

Nvidia's stock is up roughly 20% so far this year, underperforming many of its semiconductor peers but still enough of a gain to support the biggest market cap in the world.

Nvidia became the first $5 trillion company in October, and inched closer to reaching the $6 trillion record last week, though after a bit of a pullback the number now sits at $5.5 trillion.

The company's record-breaking run comes as chip companies not named Nvidia hit historic highs. Intel had its best month ever in April, as agentic AI spins up a major renaissance for the central processing unit. Memory makers like Micron, meanwhile, have seen shares surge amid a shortage for the key type of chip needed to support AI.

Alphabet briefly surpassed Nvidia to become the world's most valuable company in after-hours trading earlier this month, but for now that appears to be a momentary blip. Google's parent is currently worth about $4.6 trillion.

—Katie Tarasov

Nvidia has been a leading beneficiary of the AI boon due to its graphics processing units, or GPUs, that are used to train and run powerful foundation models.

Data center revenue for Nvidia's fiscal first quarter is expected show an 87% increase from a year earlier to $73.1 billion, representing even faster expansion than the 75% year-over-year jump in the prior quarter and 73% growth rate in the same period a year ago.

The persistent growth reflects the exploding capital expenditures from hyperscalers, which are snapping up GPUs for their data center computing infrastructure that underpins their AI initiatives.

On the same day last month, Alphabet, Amazon, Meta and Microsoft all reported quarterly results, giving investors an updated glimpse into their capex forecasts for the year. Financial firms like Evercore and Bank of America are projecting the group will spend over $1 trillion on AI-related capex in 2027, which ultimately benefits Nvidia.

John Belton, portfolio manager at Gabelli Funds, said in an email on Tuesday that he's "looking for whether the company is broadening its customer base as that remains a major risk," adding that five names account for roughly half of Nvidia's business.

"I'm questioning things such as how durable the growth within that segment of the business is as well as if they're expanding the customer base and broadening the product set," Belton said.

—Jonathan Vanian