Nvidia's AI Boom: A Double-Edged Sword for Server Makers

Nvidia's AI boom brings riches to its partners -- at a cost

When Dell completes a $5 billion deal to supply servers to Elon Musk's xAI startup, it will mark an immense win for the 40-year-old computer seller. But the proposal, tentatively agreed to recently, comes with a detail all too common for server makers in the AI era: a narrow margin of profitability. The plan to supply xAI with AI-optimized servers is expected to yield gross margins in the mid-single-digits percentage, according to people familiar with the deal. Dell's dilemma is playing out for the largest U.S. server makers as they navigate the artificial intelligence boom. Though technology from Nvidia has created a sales surge for Dell, Hewlett Packard Enterprise and Super Micro Computer, the companies have seen shrinking margins and reduced influence over major customers. Nvidia, meanwhile, is often the one calling the shots. As the primary chip supplier for AI computing, it can make demands about the way its technology should be used by customers and partners. And that's been frustrating to companies including Dell, according to workers at the server makers. Nvidia has also expanded the products it sells beyond semiconductors, reducing opportunities for companies such as Dell or HPE to bundle higher-margin items. For example, in Dell's xAI deal, Nvidia will be providing the networking, a category of products that Dell also sells, according to people familiar with the plans. More than 20 workers at the U.S. server makers were interviewed for this story, and they requested anonymity to speak candidly about the relationship with a crucial partner. What some at the companies see as overbearing behavior by Nvidia, the chipmaker sees as necessary innovation to propagate the use of AI, something it argues will boost the entire economy. In a statement, Nvidia said the server manufacturers are "long-term Nvidia partners." It works with the companies to build AI systems and infrastructure that help customers "achieve a new level of productivity and capability," Nvidia said. Chipmakers have long wielded power over computer manufacturers, going back to the days of the Intel Inside campaign. But Nvidia's chips are in such tight supply that server companies have little control over the components they get. "The key issue for Dell or HPE is that Nvidia has the power to determine allocation," said Simon Leopold, an analyst at Raymond James. The pool of buyers for high-powered AI chips is small enough that Nvidia is largely able to manage the relationships itself, said Matt Bryson, an analyst at Wedbush Securities. Traditionally, component makers were more reliant on companies including Dell to communicate with customers. So when the most important clients, such as cloud services provider CoreWeave Inc. or Musk's xAI, want to buy high-powered servers, they often first approach Nvidia to work out which technology will be used in the project and how much, according to people familiar with the matter. That makes it tougher for server makers to sell ancillary products including storage and networking, the people said. Cross-selling equipment and services is how vendors typically pad earnings. Dell, for example, posted more than $16.2 billion in storage sales over the last year. Dell, HPE and Super Micro each declined to comment. Dell has said that while AI servers are less profitable than other products it offers, the deals still add to the company's bottom line and are an opportunity to sell services and other equipment. Nvidia also can swing the share price of its server partners. During a speech at the chipmaker's annual user conference last March, CEO Jensen Huang praised Dell's systems. Michael Dell, that company's founder and CEO, would be "happy to take your orders," he joked. That week, Dell's stock jumped almost 5%. Dell and Nvidia have a longstanding relationship, dating to the early days of personal computer gaming. Today, they dub their co-branded AI server offering the "Dell AI Factory." Over the last year, many Dell sellers have been flown to Nvidia's headquarters in Santa Clara, Calif., to get training and be certified as AI advisers. But Nvidia has also irked partners by pushing its own lines of AI servers and other gear. Early in the artificial intelligence boom, before Dell had an equivalent AI server product, Dell's salespeople offered Nvidia-made devices to customers. Last year, the Texas-based company cut incentives for sellers to push those systems, according to people familiar with the matter. AI servers are built to specifications from Nvidia. With a rapid cadence of product releases in the AI era, partners have been left scrambling with frequent and last-minute tweaks to Nvidia's guidelines, said workers at the server makers. That can make it tough to launch products based on the newest chips as quickly as Nvidia itself is able to, the people said. Analysts estimate that Dell shipped more than $10 billion of its AI servers in the fiscal year ended in January. Super Micro, meanwhile, expects a 62% jump in sales for the fiscal year ending in June, largely due to the new AI business. Among the three, HPE has posted smaller sales in this emerging category. But there are signs the company is gaining traction. It recently inked a deal with Musk's social network, X, for more than $1 billion. HPE also has a private cloud offering that bundles high-powered servers with products like storage, networking and software. This month, HPE CEO Antonio Neri plans to appear at Nvidia's annual conference, according to a spokesperson. He didn't attend last year. Nvidia's leverage is likely to continue -- at least until rival chipmakers such as Advanced Micro Devices or Intel launch products that compete with Nvidia's best offerings. Dell and HPE could gain a stronger position in another way. They have existing relationships with a broad swath of corporate customers, and the spread of AI servers to more of those businesses will potentially give the companies an edge, said Leopold, the Raymond James' analyst. But even there, the server makers face competition from Nvidia. The chipmaker announced an extension of its partnership Tuesday with Cisco Systems to incorporate Cisco's networking gear into Nvidia's AI servers in an effort to make it easier for corporations to deploy artificial intelligence systems. "We're only in the very, very early innings of enterprises figuring out how to deploy AI," Dell Chief Operating Officer Jeff Clarke said during a recent earnings call. "I like our chances there."

Nvidia's AI boom brings riches to partners like Dell, at a cost

Dell's dilemma is playing out for the largest US server makers as they navigate the artificial intelligence boom. Though technology from Nvidia Corp. has brought a sales surge to Dell, Hewlett Packard Enterprise Co. and Super Micro Computer , the companies have suffered from shrinking margins and less influence over major customers.When Dell Technologies completes a $5 billion deal to supply servers to Elon Musk's xAI startup, it will mark an immense win for the 40-year-old computer seller. But the proposal, tentatively agreed to recently, comes with a detail all too common for server makers in the AI era: a narrow margin of profitability. The plan to supply xAI with AI-optimized servers is expected to yield gross margins only in the mid-single-digits percentage, according to people familiar with the deal. Dell's dilemma is playing out for the largest US server makers as they navigate the artificial intelligence boom. Though technology from Nvidia Corp. has brought a sales surge to Dell, Hewlett Packard Enterprise Co. and Super Micro Computer , the companies have suffered from shrinking margins and less influence over major customers. Nvidia, meanwhile, is often the one calling the shots. As the primary chip supplier for AI computing, it can make challenging demands about the way its technology should be used by customers and partners. And that's been frustrating to companies like Dell, according to workers at the server makers. Nvidia has also expanded the products it sells beyond semiconductors, reducing opportunities for companies like Dell or HPE to bundle higher-margin items. For example, in Dell's xAI deal, Nvidia will be providing the networking, a category of products that Dell also sells, according to people familiar with the plans. More than 20 workers at the US server makers were interviewed for this story, and they requested anonymity to speak candidly about the relationship with a crucial partner. What some at the companies see as overbearing behavior by Nvidia, the chipmaker sees as necessary innovation to propagate the use of AI, something it argues will boost the entire economy. In a statement, Nvidia said the server manufacturers are "long-term Nvidia partners." It works together with the companies to build AI systems and infrastructure that helps customers "achieve a new level of productivity and capability," Nvidia said. Chipmakers have long wielded power over computer manufacturers, going back to the days of the Intel Inside campaign. But Nvidia's chips are in such tight supply that server companies have little control over the components they get. "The key issue for Dell or HPE is that Nvidia has the power to determine allocation," said Simon Leopold, an analyst at Raymond James. The pool of buyers for high-powered AI chips is small enough that Nvidia is largely able to manage the relationships themselves, said Matt Bryson, an analyst at Wedbush Securities. Traditionally, component makers were more reliant on companies like Dell to communicate with customers. So when the most important clients, such as cloud services provider CoreWeave or Musk's xAI, want to buy high-powered servers, they often first approach Nvidia to work out which technology will be used in the project and how much, according to people familiar with the matter. That makes it tougher for server makers to sell ancillary products like storage and networking, the people said. Cross-selling equipment and services is how vendors typically pad earnings. Dell, for example, posted more than $16.2 billion in storage sales over the last year. Dell, HPE and Super Micro each declined to comment. Dell has said that while AI servers are less profitable than other products they offer, the massive deals still add to the company's bottom line and are an opportunity to sell services and other equipment. Nvidia also can swing the share price of its server partners. During a speech at the chipmaker's annual user conference in March 2024, Chief Executive Officer Jensen Huang praised Dell's systems. Michael Dell, that company's founder and CEO, would be "happy to take your orders," he joked. That week, Dell's stock jumped almost 5%. Dell and Nvidia have a longstanding relationship, dating to the early days of personal computer gaming. Today, they dub their co-branded AI server offering the "Dell AI Factory." Over the last year, many Dell sellers have been flown to Nvidia's headquarters in Santa Clara, California, to get training and certified as AI advisers. But Nvidia has also irked partners by pushing its own lines of AI servers and other gear. Early in the artificial intelligence boom, before Dell had an equivalent AI server product, Dell's salespeople offered Nvidia-made devices to customers. Last year, the Texas-based company cut incentives for sellers to push those systems, according to people familiar with the matter. AI servers are built to specifications from Nvidia. With a rapid cadence of product releases in the AI era, partners have been left scrambling with frequent and last-minute tweaks to Nvidia's guidelines, said workers at the server makers. That can make it tough to launch products based on the newest chips as quickly as Nvidia itself is able to, the people said. Analysts estimate that Dell shipped more than $10 billion of its AI servers in the fiscal year ended in January. Super Micro, meanwhile, expects a 62% jump in sales for the fiscal year ending in June, largely due to the new AI business. Among the three, HPE has posted smaller sales in this emerging category. But there are signs the company is gaining traction. It recently inked a deal with Musk's social network, X, for more than $1 billion. HPE also has a private cloud offering that bundles high-powered servers with products like storage, networking and software. Next month, HPE CEO Antonio Neri plans to appear at Nvidia's annual conference, according to a spokesperson. He didn't attend last year. Nvidia's massive leverage is likely to continue -- at least until rival chipmakers such as Advanced Micro Devices or Intel Corp. launch products that compete with Nvidia's best offerings. Dell and HPE could gain a stronger position in another way. They have existing relationships with a broad swath of corporate customers, and the spread of AI servers to more of those businesses will potentially give the companies an edge, said Leopold, the Raymond James' analyst. But even there, the server makers face competition from Nvidia. The chipmaker announced an extension of its partnership Tuesday with Cisco Systems to incorporate Cisco's networking gear into Nvidia's AI servers in an effort to make it easier for corporations to deploy artificial intelligence systems. "We're only in the very, very early innings of enterprises figuring out how to deploy AI," Dell Chief Operating Officer Jeff Clarke said during a recent earnings call. "I like our chances there."

Nvidia's AI Boom Brings Riches to Partners Like Dell, at a Cost

But the proposal, tentatively agreed to recently, comes with a detail all too common for server makers in the AI era: a narrow margin of profitability. The plan to supply xAI with AI-optimized servers is expected to yield gross margins only in the mid-single-digits percentage, according to people familiar with the deal.