AMD buys MEXT to boost data centre memory efficiency

AMD has acquired MEXT, adding memory optimisation technology to its data centre and AI portfolio.

The deal targets a pressure point in modern computing infrastructure as demand for memory rises across artificial intelligence, data analytics, virtualisation and high-performance computing workloads.

MEXT develops predictive memory software designed to make flash storage behave more like DRAM. The technology can expand usable memory capacity while maintaining system performance and efficiency, helping customers manage infrastructure constraints in cloud and enterprise environments.

Memory availability has become a growing issue for operators of large-scale compute systems. As models and datasets increase in size, organisations must often balance performance, hardware cost and deployment speed, particularly as DRAM remains one of the more expensive parts of server infrastructure.

Memory focus

By bringing MEXT into the business, AMD aims to strengthen its position in data centre computing with software that addresses one of the practical bottlenecks in running AI and other demanding workloads. The technology will be integrated across its data centre portfolio.

The move also reflects a broader trend in the semiconductor industry, as chipmakers look beyond processors to surrounding software, systems design and infrastructure efficiency. For vendors serving enterprise and cloud customers, the ability to improve memory use can affect both hardware utilisation and the cost of scaling services.

AMD framed the transaction as part of its wider effort to offer a more complete compute stack for customers building AI systems. In that context, memory management has become increasingly important because the performance of accelerated computing platforms is shaped not only by processing power, but also by how quickly and efficiently data can be accessed.

MEXT's tools are intended to reduce infrastructure costs and improve resource use. Those claims centre on using lower-cost flash memory in ways that lessen dependence on DRAM, potentially changing how operators configure systems for both general-purpose and AI applications.

Data centre strategy

The purchase adds a team with specialist experience in memory systems and AI infrastructure. AMD said that expertise would support its work on challenges tied to modern data centre build-outs, where operators face pressure to deploy larger systems without proportionate increases in cost or power use.

The deal comes amid intense competition among chip companies to supply the infrastructure behind AI services. While much attention has focused on graphics processors and AI accelerators, memory and interconnect performance have emerged as equally important factors in determining how effectively systems can run large models and data-heavy tasks.

For enterprise buyers, memory bottlenecks can slow deployment plans and force changes to system design. Technologies that extend the practical use of existing memory resources may offer an alternative to simply adding more hardware, especially where budgets, power limits or physical space constrain expansion.

AMD said the acquisition would help customers improve performance per dollar and increase efficiency at scale. It also argued that adding MEXT would support faster deployment of AI infrastructure by easing one of the main constraints operators face.

No financial terms were disclosed. AMD said combining its data centre platforms with MEXT's software marks another step in its effort to help customers deploy workloads more efficiently, cost-effectively and at greater scale.

"Demand for memory is growing across every category of enterprise compute. By combining AMD leadership in high-performance computing and data centre platforms with MEXT's memory optimization technology, we are taking another step to help customers deploy workloads more efficiently, cost-effectively and at greater scale. We look forward to welcoming the MEXT team to AMD," AMD said.