After a decade-long comeback, chip giant Advanced Micro Devices is preparing to ship its first rack-scale artificial intelligence system, called Helios, to a growing list of customers that now includes Microsoft.
It’s the first competitor to Nvidia’s popular plans by Grace Blackwell and Vera Rubin, and aims to give the world’s most important chipmaker its first real competition in years.
Microsoft announced on Monday that it will use the Helios system in its data centers, join Meta, OpenAIOracle and others in a race to grab as much computing power as possible.
AMD will begin shipping to customers, including Microsoft, later this year.
Details about the financial terms or the price of the computer were not disclosed.
“We’re expanding the Azure infrastructure portfolio with AMD Helios to give customers the performance, scale and choice they need to build and run the next generation of AI applications,” Microsoft CEO Satya Nadella said in a press release.
The new Helios system will power Microsoft’s frontier model, its AI customers and support Azure AI services. Microsoft will also add two new computing environments running on AMD “Venice” central processing units, or CPUs, one for AI agents and data pipelines, and one for semiconductor design.
It’s a continuation of a long-standing partnership, with AMD chips powering Microsoft’s Surface PCs and Xbox gaming consoles for years. In 2023, Microsoft was also the first to adopt AMD’s MI300X graphics processing unit, or GPU, which competed with Nvidia’s AI chips. Microsoft also uses its Maia chips in its data centers.
Like its peers, Microsoft needs as much computing as possible, especially as it expands its own model and allocates more computing power to research and development. In June, it announced seven in-house models. Microsoft’s AI efforts so far have seen mixed results, from its 365 Copilot AI assistant to GitHub’s Copilot code agent. It is the worst performing “Magnificent Seven” stock so far this year.
Microsoft is part of a growing number of large companies turning to AMD for AI acceleration. AMD says eight of the top 10 AI companies are running a lot of work on Instinct GPUs, including OpenAI, Cohere and Elon Musk’s SpaceXAI, which is part of SpaceX.
AMD’s work installs the first Helios rack-scale AI system in a data center lab in Rockdale, Texas, on June 24, 2026. Helios comes in four customizable configurations, and this is the one Meta will use later this year.
Andrew Evers | CNBC
In February, Meta announced that it will use up to 6 gigawatts of AMD GPUs over time, starting with 1 gigawatt installed in Helios racks later this year. OpenAI and Oracle also made major commitments to deploy Helios this year, with India’s largest IT company, Tata Consultancy Services, committing to use it as well.
CNBC got the world’s first detailed look inside the Helios system, from the Texas data center lab where it is being developed and tested.
‘Lowest cost per token’
Named after the ancient Greek god who pulled the sun from the sky with the help of four horses, Helios combines four things that AMD makes in-house: GPUs, CPUs, networking and software.
“We’re very focused on providing the best total cost of ownership, the lowest cost per token, all in-house,” data center head Forrest Norrod told CNBC about AMD’s first-generation system. “And our customers tell us that we achieve that.”
In May, AMD CEO Lisa Su told CNBC’s Jim Cramer that Helios has “significant advantages” over Nvidia’s rack-scale systems, “when you talk about inference and when you talk about memory bandwidth and memory capacity.”
While AMD would not comment on cost, Futurum Group estimates that Helios will cost between $5 million and $5.5 million. That compares to Futurum’s estimates of $3.5 million to $4 million for Nvidia’s second-generation rack-scale system, Vera Rubin.
Weighing in at 7,000 pounds, the Helios is also wider and heavier than Nvidia’s Vera Rubin.
Nvidia controls more than 95% of the data center GPU market, according to Futurum Group. AMD only holds 4.5% of the market, but Helios could change that.
“I think there is a serious case where AMD is doing well and can reach 20% and 25%. And by the way, this is hundreds of billions of dollars in revenue,” said Daniel Newman, analyst and CEO of Futurum Group.
In the first quarter of 2026, data centers made up the largest portion of AMD’s revenue, up 57% year over year. AMD told CNBC it plans to book tens of billions in data center AI revenue starting in 2027, most from Helios.
Data center CPU market share, Intel remains the clear leader, but AMD keeps gaining momentum. This CPU leadership separates AMD from Nvidia, which launched its first server CPU in 2021 and changed strategies to focus on chips this year.
‘A very different AMD’
Norrod called Helios “our baby,” as he showed CNBC the core chips of the system. Each of its 18 compute trays has four Instinct GPUs powered by a single EPYC central processing unit.
It was these EPYC data center CPUs that helped AMD regain lost leadership in the data center market.
In 2003, AMD had a key data center CPU that helped it quickly gain nearly a quarter of the market, but that slice withered following a series of delays and missteps that led to massive layoffs and reduced revenue when Su took over.
“Under Lisa’s leadership over the past 12 years, it has been a very different AMD,” Norrod said.
Things changed after the company launched the first EPYC server CPU on stage in 2017.
“One thing we have done is that we have laid out our road map for three generations, which is unusual,” he said. “And we delivered what we said.”
Part of AMD’s roadmap also included plans to launch Helios with the current MI400 series of GPUs.
Each Helios tray also has up to 12 network chips made with technology AMD acquired when it bought Pensando in 2022.
It was one of several acquisitions that helped power Helios’ development over the past few years.
AMD’s biggest acquisition to date was programmable chip company Xilinx for around $50 billion by 2022. AMD also acquired ZT maker systems for about 5 billion by 2025, as well as a series of software companies that helped it develop ROCm, its open source alternative to Nvidia’s widely accepted CUDA software.
Counterpoint Research analyst Neil Shah said AMD’s Helios chips are “on par” with Nvidia GPUs and CPUs, but “the secret sauce is in the software and the optimization.”
“With CUDA, I think Nvidia has a huge ecosystem, and it’s way ahead compared to AMD,” he said.
With Helios, AMD has the opportunity to take drastic measures, depending on how well early shipments go.
“The question is going to be: Is AMD winning because it’s technologically superior? Or is AMD winning because there’s a barrier that’s so hard that if they can build it, nobody will buy it?” Newman said.
WATCH: First look at Vera Rubin, Nvidia’s AI program that is 10 times better than Grace Blackwell
– CNBC’s Jordan Novet contributed to this report.
