AMD acquires MEXT to solve memory constraints with AI-powered tiering technology

AMD takes over MEXT to 'address growing memory constraints' in the data center -- memory tiering technology enables flash to appear as DRAM to applications

AMD on Monday announced that it had acquired MEXT, a startup that developed a memory tiering technology that makes NAND flash memory appear as DRAM to the operating system, which enables operators of data centers to save money on DRAM. AMD expects the acquisition to help customers improve system efficiency, lower operating costs, and deploy large-scale workloads more quickly. As AI models continue to expand and datasets grow larger, memory availability has become an increasingly important factor affecting overall system performance. In many cases, memory resources, not CPUs or GPUs, are becoming a performance bottleneck. Meanwhile, in many cases DRAM is used inefficiently. MEXT addresses memory efficiency challenges with an AI-based memory tiering technology that moves infrequently accessed data from expensive DRAM to NAND storage, which costs orders of magnitude less per unit of capacity, and in a way that's transparent to applications. MEXT's Predictive Memory Engine continuously analyzes memory access patterns and uses AI models to anticipate which data stored in flash will be needed next. Those memory pages are proactively transferred back into DRAM before applications request them and enable software to access data as though it were in main memory, thus preserving performance levels. By increasing the amount of usable memory available to applications, MEXT's technology aims to improve utilization of existing infrastructure and at the same time reduce needs for expensive DRAM. This approach can potentially lower total cost of ownership for cloud providers and enterprise customers and enable larger workloads to run on existing hardware. AMD believes that these capabilities can benefit both traditional data center applications and modern AI deployments, where access to large memory pools is often critical for efficiency and scalability. AMD plans to incorporate MEXT's technology into its data center product portfolio and expand its capabilities to address memory-hungry AI workloads. The company already offers integrated solutions that combine processors, accelerators, networking technologies, and software, so MEXT's Predictive Memory Engine will complement the already broad portfolio. As an added bonus to the technology itself, AMD gains a team with expertise in memory architectures, infrastructure software, and large-scale computing systems. Terms of the deal are unknown. Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.

AMD's Mext buy shows how AI could solve the RAM shortage it created

With no end in sight to the memory crunch, AMD thinks that AI, the main cause of the shortage, could be part of the solution. This week, the House of Zen acquired predictive memory startup Mext for an undisclosed sum, setting the stage for a world where bots decide which data to put into RAM and which to store in less-expensive flash. Founded in 2023, the Mext proactive memory platform uses machine learning algorithms and learned heuristics to proactively offload "cold" memory to flash storage, and, based on data access patterns, restore it before its needed again. Modern flash arrays are already approaching main memory in terms of aggregate bandwidth, but swapping to disk still imposes a stiff latency penalty. Mext claims it can expand the effective memory of a system by 2 to 4x using flash, which gig for gig is still vastly less expensive than DRAM. This flash memory is exposed to the operating system like regular memory simply by running the Mextd daemon. Memory tiering is nothing new and has seen various reincarnations over the years with some being software based and others, like Intel Optane persistent memory, using special 3D XPoint memory tech co-developed by Micron. Mext stands out for its use of machine learning to migrate data from hot memory to cold storage almost like a branch predictor -- something AMD has an awful lot of experience with. Mext isn't using one model to decide when to shuffle your data. Instead it uses a series of heuristics, long short term memory, and modern transformer architectures depending on which combination renders the best results. "This approach has the potential to reduce infrastructure costs, improve resource utilization, and help customers more effectively scale general-purpose and AI workloads," Dan McNamara SVP of AMD's compute and enterprise AI biz wrote in a blog post this week. Beyond enterprise applications, the technology could have implications for AI serving. Modern mixture of experts (MoE) models are, as their name suggests, comprised of multiple sub-models. For each token predicted, a different selection of experts may be used. In practice an LLM may use some experts more frequently and others rarely. We wouldn't be surprised to see AMD use Mext's prediction algorithms to offload infrequently utilized experts from HBM to slower system memory, enabling enterprises to take advantage of larger more capable models with fewer resources. That's just speculation of course, but we've reached out to AMD for comment; we'll let you know if we hear anything back. ®

AMD buys data center memory optimization startup Mext

Chipmaker Advanced Micro Devices Inc. said today it has snapped up the flash memory optimization startup Mext Corp. in an effort to help its customers solve their growing headaches around memory supply constraints in artificial intelligence data centers. The price of the acquisition was likely small as the terms of the deal were not disclosed, but AMD said the startup's technology can help its customers to enhance system efficiency, lower their operating costs and deploy large-scale AI workloads much more rapidly. Mext is the creator of an AI-native memory tiering technology that shifts infrequently accessed data to NAND-based flash storage rather than keeping it held within expensive dynamic random-access memory. Flash storage can cost orders of magnitude less than standard memory. The startup's technology is based on a predictive memory engine that continuously analyzes memory access patterns and uses AI algorithms to try and guess which data stored in flash systems might be needed next. It proactively transfers that data back into the DRAM when it believes it might be needed, so that applications can access it without any latency, thus ensuring the same level of performance for AI applications. The deal takes aim at a growing challenge facing operators of large-scale data centers. As AI models, high-performance computing, virtualization and data analytics workloads become larger and more complex, they require vast amounts of memory, which is facing acute supply problems. Simply put, there aren't enough memory chips to go around, and that has had the impact of driving up prices and limiting availability. The lack of memory resources is therefore becoming a major performance bottleneck in some data centers, yet in many cases, the way DRAM is employed by applications is extremely inefficient. AMD believes that Mext's technology can help to address this problem, solve data center efficiency challenges and accelerate data center performance. Mext basically increases the amount of usable memory available for applications, which improves the utilization of existing infrastructure while reducing the need to buy more expensive DRAM. The chipmaker believes the technology can help to reduce the total cost of ownership for cloud providers and enterprises alike, and enable larger workloads to run on their existing data center hardware. Moreover, it's not just AI data centers that can benefit - even traditional data center workloads can enjoy the same advantages. AMD said it will incorporate Mext's technology into its broader data center portfolio. It already sells integrated hardware and rack-scale systems that combine its processors, AI accelerators, networking technologies and software, and it seems likely that Mext's tech will be able to complement these systems. AMD's senior vice president and general manager of Compute & Enterprise AI Dan McNamara said the acquisition expands the company's ability to offer differentiated full-stack compute and AI solutions. "By integrating Mext's technology across the AMD data center portfolio, we expect to help enterprise customers unlock greater value from their infrastructure investments while accelerating AI deployment," he added. Additionally, the deal also brings Mext's talented team of engineers with extensive experience in memory architectures, infrastructure software and large-scale computing systems to AMD. Benchmark analyst Code Acree told MarketWatch that AMD is unlikely to see any substantial boost in its near-term revenue, but gains a nice alternative for its memory architecture system design. "The deal is just a nice technology tuck-in that adds to its memory optimization capabilities," he said. Nonetheless, Wall Street seemed to react positively to the acquisition, with AMD's stock briefly surging during the regular trading session to take its market capitalization above $900 million for the first time. However, it later surrendered some of those gains to end the day up just 6%.

AMD Bets on MEXT's Predictive Memory Tech to Slash Data Center TCO While Rivals Scramble for DRAM

AMD has announced the acquisition of MEXT, a pioneer in AI-driven memory optimization technologies to help lower TCO. AMD & MEXT To Accelerate Memory Deployment Through Optimizations & Lower TCO For AI Data Centers As memory requirements continue to jump with Agentic AI adoption, major firms are looking towards key technological innovations that can not only optimize the usage of memory but also lower TCO in huge AI deployments. Today, AMD announced that it has acquired MEXT, and this marks a major step for them to accelerate memory deployments while addressing one major challenge: memory access. With MEXT, AMD aims to solve this by repurposing Flash to behave more like DRAM, expanding the total usable memory capacity with similar performance and efficiency characteristics. With this approach, AMD claims a reduction in total costs, improved resource utilization, and enables customers to scale general-purpose and AI workloads effectively. Press Release: Modern data center infrastructure is evolving rapidly, and customers are increasingly facing a common challenge: memory access. As AI models, data analytics, virtualization, and high-performance computing workloads grow in size and complexity, memory has become a critical constraint across cloud and enterprise environments. For customers, addressing these bottlenecks is essential to improving performance per dollar, increasing efficiency, and accelerating deployments at scale. AMD is addressing this challenge by acquiring MEXT, a pioneer in AI-driven memory optimization technology. MEXT has developed innovative AI-powered predictive memory technology designed to make flash behave more like DRAM, helping expand usable memory capacity while maintaining performance and efficiency. This approach has the potential to reduce infrastructure costs, improve resource utilization, and help customers more effectively scale general-purpose and AI workloads. The acquisition adds to our ability to deliver differentiated, full-stack compute and AI solutions. By integrating MEXT's technology across the AMD data center portfolio, we expect to help enterprise customers unlock greater value from their infrastructure investments while accelerating AI deployment. Just as important, MEXT brings a talented team with deep expertise in memory systems and AI infrastructure. Their innovation and technical leadership will further strengthen efforts at AMD to solve some of the most important challenges facing modern data center buildouts. Demand for memory is growing across every category of enterprise compute. By combining the AMD leadership in high-performance computing and data center 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. Follow Wccftech on Google to get more of our news coverage in your feeds.