Samsung's Foundry Has Two Markets Now. The Qualcomm Standoff Is Where They Collide.

The narrative around Samsung's foundry business follows a predictable pattern: another customer lost, another yield problem, another quarter of losses, and the same conclusion — Samsung can't compete with TSMC.

That pattern is what makes this week's reporting notable. Samsung and Qualcomm, two of the largest semiconductor companies in the world, cannot agree on pricing for a 2nm chip manufacturing deal. Qualcomm wants a discount. Samsung is not giving one.

The headline framing suggests Samsung is struggling to keep a major customer. The supply-side reality points to the opposite. Samsung has stopped discounting. It raised prices on 4nm, 5nm, and 8nm processes by up to 15% in July on new orders. It is refusing to cut 2nm pricing for a customer that is offering relatively small volume. And it can afford the posture because its foundry business no longer depends on whoever will take a seat.

This is not a pricing dispute. It is a sign that Samsung foundry has split into two businesses — one locked into massive AI manufacturing contracts where it has pricing power, the other serving legacy mobile customers who still expect the old foundry economics. The Qualcomm standoff is the visible edge of that split.

The structural shift

To understand what a pricing standoff means in this context, the timeline matters. Samsung foundry lost an estimated $30 billion cumulatively between 2022 and 2025, with continued financial pressure projected through 2027. During that period, the strategy was clear: secure volume at any price. The company aggressively discounted advanced-node capacity to attract customers and keep utilization above the break-even threshold. It worked to fill fabs but not to generate profit.

Samsung itself acknowledged the problem in June 2026, revising its profitability outlook. Management told employees that the foundry division would remain under financial pressure through 2027 and would not return to profitability until 2028 — a full year later than some external analysts had projected. The internal message was blunt: losses were management's responsibility, and compensation would not improve until business performance did.

Then the customer composition changed.

In July 2025, Samsung announced a $16.5 billion multiyear contract to manufacture AI chips for Tesla, running through 2033. In July 2026, Samsung and Broadcom signed a memorandum of understanding covering memory, foundry, and advanced packaging, with an estimated value exceeding $200 billion through 2030. The Broadcom deal specifically includes Samsung's 2nm and sub-2nm processes for AI accelerators and communications chips.

These are not speculative pipeline items. They are multi-year commitments from customers that buy in volume and on long horizons. Together with orders from companies like Groq, they have fundamentally altered Samsung's capacity picture. The first Taylor, Texas fab, which will produce 2nm chips, was reported as fully booked ahead of its late-2026 initial operations.

When capacity is booked and the marginal customer is no longer the one that saves the business, pricing discipline becomes possible. That is what Samsung is exercising now.

Why Qualcomm matters less than it appears

Qualcomm's position in this dispute reflects its business model. As a fabless chip designer, Qualcomm doesn't manufacture chips — it contracts with foundries. Its current flagship, the Snapdragon 8 Elite Gen 5, runs on TSMC's 3nm process. Qualcomm had planned to move a next-generation chip to Samsung's 2nm in 2026 to rival Apple's A20 Pro, which uses TSMC's 2nm.

Qualcomm is content with Samsung's technical performance. 2nm yields have climbed to over 70%, a significant improvement from below 50% earlier this year. The barrier is purely economic: Qualcomm wants a competitive manufacturing cost, and Samsung is not reducing its price.

The volume under discussion is reportedly insufficient to persuade Samsung to discount. For context, Qualcomm's total revenue over the past four quarters is approximately $44 billion, with $191 billion in market capitalization, 25.5% operating margins, and $10.4 billion in trailing free cash flow. Even if Qualcomm committed significant spend to Samsung's 2nm node, it would not match the order magnitude of Tesla or Broadcom. Samsung's calculus has changed: it can be selective.

If the deal slips to 2027 or falls through entirely, Qualcomm's options are clear. It can return to TSMC, which manufactured its Snapdragon chips before the 4nm yield problems in 2021 drove Qualcomm to Samsung. It can absorb a year-long delay. Neither outcome is catastrophic for a company with that financial profile.

The two-market foundry

The Qualcomm dispute is the clearest public signal yet that Samsung foundry is operating as two businesses with different economics.

The AI-foundry business has pricing power. It is served by locked-in, multi-year contracts with customers whose demand trajectories are steep and whose alternative capacity at TSMC is tight. This segment is where the Broadcom and Tesla deals live, and where the Taylor fab's 2nm capacity is already committed.

The legacy-foundry business is still negotiating on thin margins. Mobile chipmakers, traditional application-processor designers, and customers outside the AI infrastructure stack still expect foundry pricing that reflects competition — not capacity scarcity. This is where Qualcomm sits, and where Samsung's historical customer base remains.

The market share numbers tell a parallel story. Samsung foundry held roughly 11% of global pure-play foundry revenue in early 2024. By mid-2025, that share had fallen to 7.7%. Counterpoint Research projects it at 7% through Q2 2026, while TSMC holds 73%. SMIC, benefiting from China-for-China demand and domestic subsidies, has reached 5% and is narrowing the gap. Samsung is losing share in units and revenue — but the composition of the customers it keeps has shifted.

That is the supply-side reading: Samsung is accepting lower volume and lower share in the legacy segment while building a concentrated, higher-margin AI foundry franchise. The pricing standoff with Qualcomm is the friction point where the two strategies collide.

What this means for investors

Samsung is not a U.S.-listed stock, but the structural implications flow through the semiconductor supply chain that U.S. investors hold.

For Samsung watchers: the foundry turnaround is not a question of whether Samsung can manufacture competitive chips. Yields above 70% at 2nm demonstrate the technical capability. The question is whether the concentration of AI-foundry customers — Tesla, Broadcom, a handful of others — creates sufficient revenue gravity to offset the ongoing losses in the legacy segment through 2028. The answer depends on execution: can Samsung deliver on the Broadcom and Tesla commitments at the margins that justify the capex, and can the Taylor fab ramp without the yield issues that plagued earlier nodes?

For Qualcomm investors: this is a supply chain negotiation that is visible but not material. Qualcomm's business model is built on designing chips and licensing patents, not on controlling its foundry relationship. A 2nm delay to 2027 doesn't change the company's near-term revenue trajectory, its operating margin profile, or its free cash flow generation. The risk is downstream — if Xiaomi and Vivo shift toward MediaTek chips to maintain their product launch schedules, Qualcomm could lose smartphone design wins. But that is a commercial risk, not a foundry crisis.

The broader implication is structural. The foundry market is no longer a single market where every customer competes on the same pricing basis. It is bifurcating along the AI-infrastructure line, with capacity and pricing power concentrated where the capital investment is directed. Samsung's willingness to let Qualcomm walk away signals that the constraint has shifted from "how do we fill the fab" to "which customer earns the capacity."

The key issue is not whether Samsung and Qualcomm will eventually agree on a price. The more important question is whether Samsung's AI-foundry customers sustain their spending through the commitment periods — because that is what determines whether the pricing discipline is a sign of a turnaround or a temporary capacity bump in a business that still hasn't solved its core profitability equation.