AMD is acquiring server maker ZT Systems to strengthen its data center technology as it steps up its challenge to Nvidia in the competitive AI chip market.
The acquisition is important because AI’s growing demands need large clusters of connected chips for processing power.
By buying ZT Systems, AMD strengthens its ability to build these high-performance systems, boosting its competitiveness against rivals such as Nvidia.
“ZT Systems’ extensive experience designing and optimizing cloud computing solutions will also help cloud and enterprise customers significantly accelerate the deployment of AMD-powered AI infrastructure at scale,” AMD said in a statement.
ZT Systems has over 15 years of experience in designing and deploying AI compute and storage infrastructure for major global cloud companies, AMD added, noting that the company is a key provider of AI training and inference infrastructure.
Raising the stakes against Nvidia
The acquisition would allow AMD to have a direct interface to real-world customer deployments as well as feedback, according to Prabhu Ram, VP of the industry research group at Cybermedia Research.
AMD can leverage these insights to refine and customize chip designs to meet specific needs and requirements.
“By integrating ZT Systems’ expertise, AMD aims to accelerate its development of AI-optimized chips and potentially gain a competitive edge in influencing customer decisions,” Ram said. “While ZT Systems has historically worked with other chipmakers, the acquisition introduces strong incentives to prioritize AMD solutions, potentially expanding AMD’s market share in AI servers.”
Notably, the largest demand for Nvidia’s systems during the past two years of the generative AI revolution has come from hyperscalers, said Neil Shah, VP of research and partner at Counterpoint Research. ZT Systems effectively meets the specific design and integration requirements of these companies.
“ZT Systems has been a key partner for Nvidia in custom designing and integrating advanced Nvidia server racks and clusters for hyper scalers,” Shah said. “This expertise in custom server design, driving tighter integration across all components from compute, accelerators, and storage to networking, makes ZT Systems a unique solution provider and a significant factor in Nvidia’s success.”
Avoiding direct competition in manufacturing
Notably, AMD plans to sell the manufacturing division of ZT Systems after the acquisition. Sharath Srinivasamurthy, associate vice president at IDC, pointed out that this means the main reason for this acquisition seems to be the design capabilities that ZT will bring in.
“For manufacturing, AMD’s strategy seems to be to leverage the proven ecosystem that has worked well for them,” Srinivasamurthy said.
Shah views this as a smart move allowing AMD to avoid direct competition with its partners. Manufacturing is a specialized skill set that AMD can leave to its server partners in Taiwan and other regions.
“But whichever company acquires the ZT Systems manufacturing division will gain valuable access to key hyper scalers and AMD-ZT Systems designs, positioning them to win major hyper scaler contracts,” Shah added.
Building on recent acquisitions
The acquisition marks another move in AMD’s recent investment surge. In July, the company agreed to acquire Silo AI, bringing an AI model developer into its fold.
Analysts note that while Nvidia remains well ahead in ecosystem and hardware traction, AMD is closing the gap quickly, making it increasingly difficult for a third player like Intel to catch up.
“AMD is growing as a major challenger to Nvidia with the series of acquisitions to grow in the rest of the decade and beyond as we are still in nascent stages of the AI revolution,” Shah said. “Having the right building blocks from chip to software to solution design will be key for AMD to create a highly valuable, sticky, and differentiated ecosystem similar to Nvidia.”
From a broader market perspective, AMD’s recent acquisitions also underscore that AI success relies on the seamless integration of hardware and software, not just hardware alone.
“AMD’s strategy is evolving in the right direction as it tries to accelerate towards an integrated stack approach,” IDC’s Srinivasamurthy said. “This also helps enterprises spring up AI infrastructure in a faster and more efficient way so that they can focus more on creating AI applications for relevant use cases for their enterprises.”
Source:: Network World