SK Hynix shares near record high on Microsoft AI chip report

SK Hynix shares approached a record high after local media reported the company is the exclusive supplier of advanced memory for Microsoft’s new AI chips.

According to Maeil Business Newspaper, each of Microsoft’s Maia 200 accelerators will use six units of SK Hynix’s HBM3E memory, citing industry sources.

The company did not confirm or disclose customer details, a spokesperson said.

The stock rose as much as 7.7% on the Korea Exchange, erasing early declines linked to tariff concerns and continuing a rally driven by AI industry optimism.

SK Hynix’s market value has neared US$400 billion, with its shares increasing 10x over three years, partly due to early supply deals with Nvidia.

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

Maia 200’s real impact on SK Hynix remains unproven

A claim that SK Hynix is the sole High Bandwidth Memory (HBM)3E supplier for Microsoft’s Maia 200 accelerator has been tied to a stock rally, yet production volume and rollout timing must be confirmed.
Until those details are known, the effect of the supply deal on SK Hynix earnings cannot be judged.
The uncertainty is reinforced by the Maia 100 accelerator, unveiled in 2023, which was reportedly not made in high volume or used for production workloads 1.
Microsoft’s broader custom silicon push has also been described as lagging other hyperscalers (the biggest cloud-computing providers), which leaves the near-term scale of Maia 200 in doubt 1.

Microsoft’s AI hardware shift creates an opening in data center cooling

Data center infrastructure vendors can gain business if Maia 200 ships widely, since advanced cooling will be needed.
Each Maia 200 chip runs within a 750W power envelope, so dense racks will run hot and may need liquid cooling beyond air cooling 2.
Microsoft is building out large data centers, including its “Fairwater” program with about 300MW AI buildings, which raises the bar for efficient cooling at scale 1.
Cooling vendors and retrofit teams can win work by matching their systems to the high-density layouts Microsoft is building for its AI fleet.