Dell completes a $5 billion deal to provide servers to Elon Musk’s Xai startup, and it’s a big win for the 40-year-old computer seller.
However, the proposals that I recently agreed to tentatively come with details that are too general for server manufacturers in the AI era. It’s a narrow margin of profitability. According to those familiar with the deal, plans to supply Xai with AI-Optimized servers are expected to bring gross profit margins at half the subdigits.
Dell’s dilemma is being deployed for the largest US server manufacturers as they navigate the artificial intelligence boom. Nvidia’s technology has generated a surge in sales for Dell, Hewlett Packard Enterprise and Super Micro Computer, but businesses are reducing margins and reducing their impact on key customers.
Nvidia, on the other hand, is often the person who calls shots. As a primary chip supplier for AI computing, you can request how customers and partners use the technology. According to server makers workers, it’s frustrating for businesses, including Dell.
Nvidia has also expanded its offerings that sell beyond semiconductors, reducing the chances of companies like Dell and HPE bundled higher margin items. For example, in Dell’s Xai trading, Nvidia offers networking, a category of products that Dell also sells, according to people familiar with the plan.
More than 20 workers from a US server manufacturer were interviewed for the story, and they requested anonymity to speak openly about their relationship with key partners.
Chipmakers, what some companies see as overwhelming Nvidia’s actions, see them as the innovation needed to spread AI use.
In a statement, Nvidia said the server manufacturer is a “long-term Nvidia partner.” Nvidia said they are working with businesses to build AI systems and infrastructure that help customers “achieve new levels of productivity and capabilities.”
Chipmakers have long been powerful against computer makers, dating back to the days of the Intel Inside campaign. However, Nvidia’s chips are in a very strict supply, and server companies have little control over the components they get.
“The key issue with Dell or HPE is that Nvidia has the power to determine allocation,” says Simon Leopold, analyst at Raymond James.
Matt Bryson, an analyst at Wedbush Securities, said the pool of buyers for powerful AI chips is small enough to allow Nvidia to primarily manage the relationship itself. Traditionally, component manufacturers have relied on communicating with customers to companies, including Dell.
So, if the most important clients, such as cloud service providers CoreWeave Inc. or Musk’s Xai, want to buy a high-power server, then people familiar with the issue should first approach NVIDIA and see which technology will be used in the project and how many technologies will be used?
This makes it difficult for server manufacturers to sell auxiliary products such as storage and networking, people said. Cross-seller equipment and services are the way vendors usually pad their revenues. For example, Dell recorded more than $16.2 billion in storage sales last year.
Dell, HPE, and Super Micro each declined to comment. Dell says that AI servers are less profitable than other products, but the transaction is still adding to the company’s revenue and is an opportunity to sell services and other equipment.
Nvidia can also distribute the stock price of its server partners. In a speech at the chipmaker’s annual user conference last March, CEO Jensen Fan praised Dell’s system. “I’m happy to take your order,” he joked, Michael Dell, the company’s founder and CEO. Dell shares rose almost 5% that week.
Dell and Nvidia have a long-standing relationship dating back to the early days of personal computer games. Today, they offer co-branded AI servers “Dell AI Factory.” Last year, many Dell sellers flew to Nvidia’s headquarters in Santa Clara, California, where they were trained and certified as AI advisors.
However, Nvidia is also plaguing its partners by pushing its own AI servers and other lines of gear. In the early days of the artificial intelligence boom, before Dell had comparable AI server products, Dell salespeople offered Nvidia devices to their customers. Last year, the Texas-based company reduced the incentives for sellers to push those systems, according to people familiar with the issue.
The AI server is built to Nvidia’s specifications. With the rapid release of product releases in the AI era, partners are scrambled with frequent and last minute tweaks to Nvidia guidelines, a server maker’s worker said. This could make it difficult for Nvidia itself to launch products based on the latest chips as quickly as possible, people said.
Analysts estimate that Dell shipped over $10 billion in AI servers for the fiscal year ended January. Meanwhile, Super Micro expects sales to jump to 62% sales for the fiscal year ending in June, primarily through the new AI business.
Of the three, HPE has recorded lower sales in this emerging category. However, there are signs that the company is gaining traction. Recently, it signed a more than $1 billion deal with Musk’s social network X. HPE also offers private cloud services that bundle high power servers with products such as storage, networking, and software.
According to a spokesman, HPE CEO Antonio Neri will appear at Nvidia’s annual meeting. He didn’t attend last year.
Nvidia’s leverage could last at least continue to rival chip makers such as advanced micro devices and Intel-launched products that compete with Nvidia’s best products.
Dell and HPE can gain a stronger position in different ways. Leopold, an analyst at Raymond James, said they have wider relationships with existing relationships with corporate clients, and the spread of AI servers to more companies could potentially give companies an edge.
But even there, server makers face competition with Nvidia. Chipmaker announced an extension of its partnership with Cisco Systems, and has incorporated Cisco networking gear into Nvidia’s AI servers to make it easier for businesses to deploy artificial intelligence systems.
“We’re only in very early innings for companies thinking about how to deploy AI,” Dell’s chief operating officer Jeff Clark said in a recent revenue call. “I love our chance.”
