AMD has a string of AI-related investments; this one focuses on hyperscaler AI compute
In its continued pursuit of a stronger arsenal of synthetic intelligence capabilities, chip firm AMD introduced one other main acquisition: A $4.9 billion cash-and-stock settlement to amass New Jersey-based ZT Methods, which gives AI and general-purpose compute for hyperscalers.
“The strategic transaction marks the following main step in AMD’s AI technique to ship management AI coaching and inferencing options primarily based on innovating throughout silicon, software program and methods,” AMD mentioned in a launch. “ZT Methods’ in depth expertise designing and optimizing cloud computing options can even assist cloud and enterprise clients considerably speed up the deployment of AMD-powered AI infrastructure at scale.”
AMD mentioned that after closing—which is anticipated within the first half of 2025—it should search for a purchaser to amass ZT Methods’ U.S.-based knowledge heart infrastructure manufacturing enterprise. Printed press stories speculated that the 1,000 cloud knowledge engineers that ZT Methods brings to AMD are on the core of the worth that AMD noticed within the enterprise.
“Our acquisition of ZT Methods is the following main step in our long-term AI technique to ship management coaching and inferencing options that may be quickly deployed at scale throughout cloud and enterprise clients,” mentioned AMD Chair and CEO Dr. Lisa Su. “ZT provides world-class methods design and rack-scale options experience that can considerably strengthen our knowledge heart AI methods and buyer enablement capabilities. This acquisition additionally builds on the investments we now have made to speed up our AI {hardware} and software program roadmaps. Combining our high-performance Intuition AI accelerator, EPYC CPU, and networking product portfolios with ZT Methods’ industry-leading knowledge heart methods experience will allow AMD to ship end-to-end knowledge heart AI infrastructure at scale with our ecosystem of OEM and ODM companions.”
A brand new report from Futurum Intelligence estimates that the GPUs accounted for 74% of chipsets utilized in AI purposes inside knowledge facilities; it expects the GPU market in knowledge facilities to see a compound annual progress price of 30% over the following 5 years, with the market leaping from $28 billion in 2023 to $102 billion by 2028. Futurum mentioned that Nvidia held 92% of the market share for GPUs in early 2024.
In July of this yr, AMD introduced that it’s shopping for the largest non-public synthetic intelligence lab in Europe, Silo AI, for $665 million in money, to bolster AI options for enterprise that can leverage AMD’s chips; that transaction closed earlier this month. The transfer was extensively seen as an effort by AMD to not solely higher allow its clients to construct AI fashions primarily based on AMD chip platforms, however to assist the corporate compete higher towards Nvidia.
“AI is our primary strategic precedence, and we proceed to spend money on each the expertise and software program capabilities to help our rising buyer deployments and roadmaps,” mentioned Vamsi Boppana, AMD senior vp, AIG, upon closing of the Silo AI transaction.
AMD has been on an AI shopping for spree, buying Mipology and Nod.ai and investing greater than $125 million in a dozen different AI corporations up to now 12 months. The corporate mentioned that, excluding the ZT Methods transaction, it has invested greater than $1 billion to develop the AMD AI ecosystem and strengthen its AI software program capabilities.
Trying again past 12 months, the corporate spent tens of billions within the final couple of years to purchase Xilinx and Pensando Methods.