- AMD is acquiring AI infrastructure company ZT for $4.9 billion
- Analysts said the deal will help it shore up gaps in its software and systems capabilities
- The end result could well be a more ‘sustainable challenger to Nvidia’
Watch out, Nvidia. Analysts think chipmaker AMD’s $4.9 billion deal to buy artificial intelligence (AI) infrastructure company ZT could help it snag a much greater share of the expected $400 billion AI market in the coming years and become a true challenger to the GPU leader.
ZT has a long history of designing and deploying AI systems in data centers. As Moor Insights and Strategy founder Patrick Moorhead pointed out on LinkedIn, that history includes rumored work building Nvidia systems for the likes of Amazon and Azure.
AMD CEO Lisa Su said the combination of AMD’s hardware and networking technology with ZT’s data center expertise will allow the company to “deliver end-to-end data center AI infrastructure at scale.”
Related:
AMD to buy ZT Systems for $4.9B to expand data center AI punch
Moorhead noted the deal will help AMD shore up two major deficiencies it has in the data center AI infrastructure space faster than it could do organically on its own.
“AMD currently has two main competitive disadvantages in AI infrastructure: its software and its system scale and maturity. AMD already addresses this well for non-AI EPYC servers and PCs, but not so much for AI racks,” he wrote.
Indeed, neXt Curve Founder Leonard Lee told Fierce that part of the reason AMD has been lagging Nvidia in the AI accelerator realm is because both it and fellow competitor Intel "have shied away from going too far downstream." That is, instead of building it all themselves, both AMD and Intel have chosen to rely on a more "open" approach that relies on ecosystem partnerships to bring it all together.
Why is this a problem? Well, as Moorhead explained "Chip’ vendors are expected to deliver the full rack and software stacks to achieve year-on-year improvements to performance, efficiency, quality and time-to-market. And the ZT acquisition is targeted at accelerating AMD's ‘above the chip’ and ‘below software’ capabilities for AI servers.”
Analysts at New Street Research agreed, pointing out in a note to investors that rack-scale and system-scale integration are two areas where AMD has trailed Nvidia. This has led to AMD GPUs generally being used for “easier, low-scale use cases” such as simulation and inferencing. By stirring ZT’s capabilities into the mix, AMD will be able to expand the use cases it can be used for over time, New Street said.
Bringing infrastructure systems engineering capabilities in-house will allow AMD to better compete with Nvidia's optimized and generally pre-packaged generative AI data center stacks, like the GB200 NVL72, Lee said.
"I do think that this play will help AMD better compete with Nvidia than merely selling GPUs," he told Fierce. "Nvidia is selling GenAI data centers in a box. AMD (and Intel) will have to do the same to play the GenAI supercomputing game."
Analyst Jack Gold thinks the deal will also help AMD compete against Intel. How? Well, he stated on LinkedIn that since ZT also supplies non-AI solutions, the deal amounts to a "direct attack by AMD on Intel" since it provides an additional opportunity for AMD to promote its CPU products over Intel's in some "very large installations."
All in all, New Street Research predicted that by 2027, AMD's market share will "comfortably exceed 10%" which would imply data center GPU sales of around $20 billion. That's "a sustainable scale, as a challenger to Nvidia," New Street's team wrote.
From Nvidia's perspective, Moorhead indicated the deal could be a knock against it given its history working with ZT and the likelihood that ZT’s design activities for others will cease once the deal closes in the first half of 2025.
He concluded that if executed correctly, the ZT deal will help AMD “take much more of that projected $400 billion datacenter AI market in 2027.”