Samsung crossed a $1 trillion valuation on Wednesday after shares of the South Korean chipmaker surged more than 10%, making it the second Asian company ever to hit the mark after TSMC. The move caps a stretch in which AI infrastructure demand has rewritten Samsung's earnings profile, with last week's results showing profits eight times higher than the same period a year ago. The driver is high-bandwidth memory, the chip that sits next to every AI accelerator shipping today.
The Wednesday rally had a second catalyst. Reports surfaced a day earlier that Apple is in talks with both Samsung and Intel to manufacture chips for Apple devices on US soil, a break from Apple's near-exclusive reliance on TSMC in Taiwan. If Samsung lands any portion of that work, it would be one of the larger shifts in the global semiconductor supply chain in a decade.
HBM is doing most of the work on Samsung's margin line. The chip is essential to running large AI models because it feeds GPUs the bandwidth they need to keep accelerator cores busy, and supply has been tight enough that pricing power has shifted firmly to the three suppliers — Samsung, SK Hynix, and Micron — that can make it at scale.
Key facts
- 01Samsung crossed a $1 trillion valuation on Wednesday after shares surged more than 10%.
- 02It is only the second Asian company to hit the threshold, after TSMC.
- 03Last week's earnings showed profits 8x higher than the same period a year ago, driven by HBM.
- 04Apple is in talks with Samsung and Intel to manufacture chips for Apple devices on US soil.
- 05Samsung workers are threatening an 18-day strike later this month over a bigger share of the AI windfall.
All three have been pulling capital out of consumer-facing memory businesses to expand HBM lines for AI data centers, where margins are materially higher than in PC and smartphone DRAM. That's the core trade behind the 8x profit jump: less commodity DRAM, more HBM stacks shipped into Nvidia and AMD accelerator supply chains.
“Samsung posted profits eight times higher than the same quarter a year ago, with high-bandwidth memory margins doing most of the work.”— Jaeden Schafer
"Every company building AI right now needs chips, and Samsung makes the memory chips that power those AI systems," TechCrunch's Kate Park wrote in summarizing the dynamic. That sentence is also the bear case in disguise — Samsung's fortunes are now tightly bound to the capex cycles of a handful of AI buyers.
Competition inside Korea is sharp. SK Hynix, the country's other memory giant, has been aggressive on HBM and at points in the current cycle has led Samsung on qualification timelines with key AI customers. Samsung's premium pricing on HBM is only sustainable as long as it keeps pace on yield and generation transitions; SK Hynix is the reason that's not a given.
The Apple manufacturing talks are a separate bet. Samsung's foundry business has historically trailed TSMC on advanced-node yields, and winning Apple silicon work — even partial — would require demonstrating reliability at a scale Samsung has not consistently hit. Intel is the other name in the conversation, which itself signals that Apple is looking for US-soil capacity more than it is looking for a TSMC equivalent.
Domestically, the AI windfall is becoming a labor story. Samsung workers are threatening an 18-day strike later this month, arguing that the people on the lines deserve a larger share of the profits the company is now booking. A strike of that length at HBM and foundry facilities would land directly on AI customers already coping with constrained supply.
There is also an internal cannibalization problem. Samsung's phone and TV divisions buy the same memory chips that the company now sells at AI-data-center prices, meaning Samsung is effectively charging itself the AI premium to build consumer hardware. That squeezes margins on the consumer side even as the chip side prints money.
The skeptic case on a $1T Samsung is that memory has always been cyclical, and HBM, despite its current scarcity, is a memory product. SK Hynix and Micron are both adding capacity, AI accelerator roadmaps will eventually move to next-generation memory standards, and any softening in hyperscaler capex would hit Samsung's earnings line before it hit Nvidia's.
Samsung's trillion-dollar moment is really a verdict on where the AI buildout's economics are accruing. The headline names — Nvidia, OpenAI, Anthropic — sit on top of a supply chain in which a handful of memory suppliers and one Taiwanese foundry hold the actual leverage, and the market is now pricing that in. The interesting question is whether Samsung can convert this cycle into a durable position in foundry as well, or whether the next downturn finds it back where it was: a memory company with a trillion-dollar memory of better days.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
