The AI memory boom hit a symbolic new high this week, as two pure-play chipmakers joined the most exclusive cohort in global equities. According to Bloomberg Technology, South Korea’s SK Hynix and US rival Micron have both crossed the $1 trillion market-capitalisation threshold, a level previously reserved for the largest hyperscalers and Nvidia. Bloomberg’s Caroline Hyde and Ed Ludlow framed the milestone as confirmation that high-bandwidth memory has become as strategically important to the AI build-out as the accelerators it feeds.
The driver is relentless demand for HBM — the stacked, high-bandwidth memory bolted onto AI GPUs and custom silicon inside every major data centre. Bloomberg reports that order books at both companies are stretching well into 2027, with hyperscaler customers locking in multi-year allocations to avoid being caught short during the next training cycle. That visibility is unusual in a sector historically defined by brutal boom-and-bust pricing, and is a key reason public markets have been willing to re-rate memory names so aggressively.
The milestone also reshapes the geopolitics of the AI supply chain. With Samsung, SK Hynix and Micron now controlling effectively all of the world’s advanced HBM capacity, governments in Seoul, Washington and Tokyo have fresh leverage — and fresh exposure — in any escalation around export controls or cross-strait tensions. Bloomberg noted in the same segment that Taiwanese prosecutors have opened a probe into alleged smuggling of restricted Nvidia chips into China via Japan, underlining how tightly memory, accelerators and enforcement are now intertwined.
For Silicon readers, the practical signal is on cost and availability. With HBM sold out and memory makers prioritising the largest buyers, UK and Scottish AI projects should expect tighter GPU server pricing and longer lead times deep into 2027.
