Microsoft has secured an exclusive high-bandwidth memory supply deal with SK Hynix for its next-generation Maia 200 AI accelerator, triggering a sharp rally in the chipmaker’s stock and tightening global AI hardware supply.
According to Bloomberg and South Korea’s Maeil Business Newspaper, SK Hynix will serve as the sole supplier of advanced HBM3E memory modules for Microsoft’s new AI chip. The news pushed SK Hynix shares up nearly 9% on the Korea Exchange, lifting the company’s market valuation close to $400 billion.
SK Hynix Becomes Sole Memory Supplier for Maia 200
Each Maia 200 accelerator integrates six units of SK Hynix’s HBM3E memory, delivering massive bandwidth for AI training and inference workloads. Maeil Business Newspaper reported that the configuration targets hyperscale data center performance rather than consumer hardware.
SK Hynix confirmed it cannot publicly disclose customer details but acknowledged strong demand for advanced memory products. The company has benefited heavily from the AI boom and previously secured early supply agreements with Nvidia, which helped drive its long-term stock surge.
According to TheElec, SK Hynix will supply its 12-layer 12H HBM3E chips, with each module delivering 36GB capacity. The full six-module setup provides a total of 216GB memory per accelerator and supports bandwidth up to 7TB per second.
Maia 200 Targets High-Performance AI Inference
Microsoft introduced Maia 200 as a dedicated AI inference accelerator built on TSMC’s 3-nanometer process. The chip contains more than 140 billion transistors and integrates 272MB of on-chip SRAM to reduce data bottlenecks and improve processing speed.
Microsoft stated that Maia 200 delivers over 10 petaFLOPS of FP4 performance and more than 5 petaFLOPS of FP8 performance within a 750W thermal envelope. The company positions the chip as capable of running today’s largest AI models with future scalability in mind.
Investors React as AI Demand Accelerates
The stock rally reflects rising expectations for high-bandwidth memory demand. Jung In Yun, CEO of Fibonacci Asset Management Global, told Bloomberg that dip buying and stronger HBM earnings expectations continue to support SK Hynix’s momentum.
Citigroup also raised its price target for SK Hynix by 56% to 1.4 million won and maintained a buy rating, citing improving memory pricing trends and long-term AI infrastructure demand.
Analyst Peter Lee noted that memory contracts increasingly require customers to secure supply one year in advance as the industry shifts toward semi-customized production models. He expects global DRAM and NAND pricing growth to outperform earlier forecasts in 2026.
Memory Market Tightens as Data Centers Take Priority
SK Hynix now controls more than 90% of the global HBM3E market, giving it significant pricing power in AI infrastructure. The company has shut down its consumer RAM production to focus entirely on enterprise and data center demand, increasing pressure on the broader memory market.
Exclusive supply deals between hyperscalers and memory manufacturers continue to push enterprise profits higher while raising component costs for consumers. Analysts also expect continued shortages in high-end GPUs and AI accelerators as cloud providers expand infrastructure aggressively.
Microsoft recently reported a 16.7% revenue increase, and the Maia 200 initiative strengthens its push into custom AI hardware to reduce dependency on third-party accelerators and improve cost efficiency at scale.
If AI infrastructure demand continues accelerating through 2026, high-bandwidth memory pricing and availability will likely remain one of the most influential factors shaping cloud expansion, GPU availability, and consumer hardware costs.
