Providing new chips and software is a big part of enabling AI acceleration in the data center, but how that technology gets integrated into data centers is just as important. AMD has boosted its hardware case with chips like the Instinct MI300X GPU, and strengthened its software approach through multiple acquisitions, and this week it took steps to upgrade its AI data center design and deployment capabilities with an agreement to acquire longtime systems integration partner ZT Systems for $4.9 billion.
Secaucus, New Jersey-based ZT reportedly has around 2,500 employees and makes about $10 billion in annual revenue working with some of the world’s largest hyperscalers on data center manufacturing, design, deployment, and integration activities. AMD’s plans to keep the part of the firm that focuses on design, deployment and systems integration–accounting for about 1,000 employees–and sell the 1,500-employee manufacturing operation to another company.
That intention to sell, which AMD’s plans to follow through on after the acquisition closes in the first half of next year, is driven in large part by the fact that AMD does not want to directly compete with its own existing network of server manufacturing partners, according to AMD President and CEO Lisa Su.
As far what AMD plans to keep, the large staff of systems integration experts and capabilities it gains from the deal will help it to more quickly optimize and deploy its AI acceleration components for data center customers, speeding up a cycle of design, optimization, production, and deployment otherwise takes “several quarters” to play out, Su said during a webcast arranged to discuss the deal
“The ZT team complements our silicon and software capabilities with critical expertise needed to deliver full AI solutions, from silicon to software to rack and cluster level solutions,” she said. “With ZT, our largest hyperscale customers will be able to more rapidly deploy optimized AMD AI infrastructure, while our OEM and ODM partners will have access to rack level designs that will speed their time to market.”
She later added, “To make AMD the AI infrastructure leader and meet the compute demand of next-generation AI models, we must offer more than leadership components. We must have expertise to optimize solutions at the systems, rack, and even data center level.” ZT’s ability to optimize AMD chip deployments also is likely to become more refined through the closer affiliation with AMD.
But that’s not all. AMD may realize a richer and more long-term return on this investment by gaining a feedback channel through ZT that will give the chip giant a “more inclusive view of hardware information” that will be required to drive cutting-edge software innovations, Su said, a comment that comes just a month after AMD's announcement the acquisition of Silo AI, its latest in a trio of software-related acquisitions.
The guidance could benefit both hardware and software developmente at AMD in the future. Jack Gold, president and principal analyst at J. Gold Associates, observed to Fierce Electronics, “(AMD) gets an experienced cloud-scale systems integrator that can feed expertise back to AMD to enable it to design more targeted and customer-centric chips. Having an actual deployment arm means they can see the places where their products could improve. So this is an attempt to get expertise based on real customer installations and quickly use that for new products.”
That is the kind of edge AMD increasingly will need as it chases Nvidia in the AI chip market, and looks to further separate itself from any other competitors. In addition, the acquisition also may also give AMD a persuasive voice in the ears of customers as they are consider whose AI chips to buy, an aspect of the deal lent further intrigue by the fact that ZT also acts as a systems integrator and manufacturer for products with Nvidia and Intel chips.
Gold agreed that ZT “has the ability to ‘steer’ deployments to AMD chips. I’m sure as a result of this, AMD will get preferential treatment in deciding any new or upgraded AI solutions that ZT Systems puts in place… I don’t expect ZT Systems to not support other vendors’ solutions, as they need to go with customer preferences, but it now has major incentives to put AMD first, which creates an expanded channel for AMD products, especially in the AI server market.”
That edge also could extend to even to non-AI deployments in the data center market. Gold noted, “Since ZT also supplies non-AI solutions, this is a direct attack by AMD on Intel Corporation by providing an additional ability to promote its EPYC CPUs against Xeon for some very large installations. This is a parallel strategy to the AMD vs. Nvidia capabilities in the AI space, and both will provide AMD with increased market leverage.”
Patrick Moorhead, founder, CEO, chief analyst at Moor Insights & Strategy, offered a different opinion regarding ZT’s ongoing support of AMD competitors, stating on X, “Once the deal is closed, I would expect all design activities for NVIDIA and Intel to cease. AMD says that manufacturing for the competitive systems will continue, which makes sense assuming that the manufacturing group is indeed spun out and sold.”
AMD is coming off a second quarter earnings report in which it showcased 115% year-over-year revenue growth for its data center business. The company expects to reach $4.5 billion in sales of AI chips for the full year, has identified a $400 billion addressable market opportunity in AI chips, and clearly sees an opportunity to up its game at a significant moment in the market’s evolution.
It remains to be seen how quickly AMD can sell off the ZT manufacturing business, and furthermore, how effectively can translate the visibility ZT will give it into the data center market into new product improvements, the latter being more of a long-term bet.
Moorhead observed, “Precision execution is exactly what is needed to turn this investment into gold for the company by increasing revenue and taking market share. Compared to AMD’s Xilinx acquisition of a few years ago, this one looks simple. This deal further reinforces the lead that AMD and Nvidia have built on the rest of the pack for AI chips. I believe this acquisition will be positive for the company and enable it to take much more of that projected $400 billion datacenter AI market in 2027.”