Samsung Raises Chip Prices Up to 15% While Losing a Third of Its Foundry Business

Samsung has raised prices on its advanced chipmaking services by 5% to 15% — and it did this right after losing more than a third of its foundry market share in a single year. New orders on Samsung’s 4nm (SF4) and 5nm (SF5) production lines are now costing customers in the US and China 10-15% more than they were a year ago, while Taiwanese customers see a smaller 5-10% bump. Even the older 8nm node, which most chipmakers treat as legacy capacity at this point, is up nearly 10%. I’ve been tracking the RAM and SSD price spiral since our coverage of the 2026 memory shortage, and this Samsung move is the clearest sign yet that the AI-driven chip crunch has stopped being a memory-only problem.

The Numbers That Don’t Add Up (At First)

Here’s what makes this genuinely strange. Samsung’s foundry business — the division that manufactures chips designed by other companies, competing directly with TSMC — saw its global revenue share fall to 7.3% in Q2 2026, down from 11.5% just a year earlier. That’s not a small dip. Samsung effectively lost more than a third of the business it had. Meanwhile TSMC did the opposite: its share climbed from 62.3% to 70.2% over the same stretch, according to reporting first surfaced by Korean outlet The Asia Business Daily and picked up widely since.

Normally, a company hemorrhaging market share cuts prices to win customers back. Samsung did the opposite. And once you look past the headline percentages, the logic actually makes sense — it’s just not the logic of a company competing on price. It’s the logic of a company that knows customers have nowhere else to go.

Why Samsung Can Raise Prices While Losing Customers

TSMC is running at full capacity. Every leading-edge wafer TSMC can produce is already spoken for, mostly by Nvidia, Apple, AMD, and the rest of the AI accelerator crowd. Intel Foundry is still years away from being a credible alternative for most of the customers who’d need it. That leaves Samsung as the only other option in the room for anyone who needs advanced-node capacity right now and can’t wait in TSMC’s queue.

That’s pricing power that belongs to the market, not to Samsung specifically — the foundry business overall has been running at a loss since 2022, and this is effectively Samsung using its remaining customers to start clawing that back. If you need 4nm or 5nm wafers today and TSMC can’t fit you in, you pay Samsung’s new number or you don’t get chips.

The Geography Is the Interesting Part

What really stands out to me is who’s paying the most. Chinese customers are absorbing the steepest increase, at 10-15%, the same bracket as US customers — but for a different reason. US export controls already block Chinese firms from buying TSMC’s most advanced nodes and from accessing Nvidia’s top-tier chips outright (something I covered when SK Hynix’s Nasdaq listing highlighted just how squeezed the whole memory and logic chip supply chain has become). Samsung is one of the few advanced-node doors still open to Chinese buyers, and Samsung knows it. Taiwanese customers, who have more realistic alternatives closer to home, got the gentler 5-10% increase.

That’s not a coincidence. It’s price discrimination based on how few options a customer actually has — and it’s a preview of how chip pricing is going to work for the next few years as capacity stays tight and export rules keep reshaping who can buy what from whom.

Quick Reference: Samsung’s July 2026 Foundry Price Hikes

What This Means Beyond Samsung’s Balance Sheet

If you’ve noticed graphics cards, RAM kits, and even some laptops creeping up in price this year, this is part of the same story. Foundry capacity is the bottleneck underneath almost everything with a chip in it, and when the two companies that control the overwhelming majority of advanced-node manufacturing both start charging more, that cost doesn’t stay contained to enterprise AI orders. It works its way down through every product that shares a fab queue with an Nvidia GPU or an AI accelerator, which by 2026 is most of consumer electronics in some form.

I don’t think this is the last price increase we’ll see this year, either. Samsung’s foundry unit still needs to turn a profit after four straight years of losses, and as long as TSMC stays sold out, Samsung has very little reason to hold prices where they were.

Frequently Asked Questions

Why did Samsung raise prices if it’s losing market share?

Because the price increase isn’t about winning new customers — it’s about extracting more value from existing ones who have limited alternatives. With TSMC fully booked and Intel Foundry not yet a mature option for most advanced-node customers, Samsung is one of very few places left to get 4nm or 5nm wafers made, regardless of its shrinking share of the overall market.

How much more will chips cost because of this?

Samsung’s own price increases are 5-15% depending on the node and customer region, but the downstream effect on finished products (phones, laptops, GPUs) is harder to pin down — it depends on how much of a given product’s bill of materials comes from Samsung-fabbed silicon versus TSMC or other suppliers, and how much manufacturers choose to absorb versus pass on.

Is this connected to the RAM and SSD price increases from earlier in 2026?

They share a root cause — AI-driven demand eating into global chip and memory capacity — but they’re different supply chains. RAM/SSD price spikes trace back largely to memory fabs (Samsung, SK Hynix, Micron) reallocating output to HBM for AI accelerators. This foundry price hike is about logic-chip manufacturing capacity, which is a separate (though related) bottleneck.

Are Chinese companies really paying more than everyone else?

Chinese customers are in the same 10-15% bracket as US customers, the highest tier Samsung is charging. The reporting suggests this reflects how few alternatives Chinese buyers have, since export controls already restrict their access to TSMC’s leading-edge nodes and to top-tier chips from companies like Nvidia.