AMD Acquires MEXT to Tackle Data Center Memory Constraints

AMD acquired MEXT to address growing memory constraints in data centers through predictive memory technology that enables flash storage to behave like DRAM. The move aims to reduce data center total cost of ownership (TCO). This technology enables flash storage to appear as DRAM to applications, making data centers more efficient.

AMD announced the acquisition of MEXT, a startup specializing in AI-driven memory optimization, to address a growing constraint across modern data center infrastructure. Financial terms were not disclosed. AMD shares rose approximately 7% intraday following the announcement, reflecting investor confidence that the deal advances the company's data center positioning after a strong Q1 2026 in which data center revenue hit a record $5.8 billion, up 57% year-over-year.

MEXT's core technology is a predictive memory tiering system that enables flash storage to appear as DRAM to applications. As AI models, analytics workloads, and high-performance computing tasks scale in size and complexity, the gap between available DRAM capacity and workload demand has become a meaningful constraint. MEXT's approach targets that gap: rather than requiring customers to provision more expensive DRAM, the software predicts which data needs to reside in fast memory and manages the movement of data between flash and DRAM tiers transparently.

AMD says it plans to integrate MEXT's technology across its data center portfolio to help enterprise customers reduce infrastructure costs and improve resource utilization. For AMD, which already competes aggressively on both EPYC server CPUs and Instinct accelerators, a memory-efficiency layer could further differentiate its data center platform against INTC and Nvidia. The key watch item is integration depth: whether MEXT's technology ships as a software add-on or becomes native to AMD silicon will determine how durable a competitive moat it creates.