Why Qualcomm Returning to Samsung for 2nm Would Upend the Chip World

Five years ago, Samsung fried a flagship. Now it is making its most aggressive push to fix that mistake.

The relationship between Samsung Foundry and Qualcomm broke in late 2021, when the Snapdragon 8 Gen 1 — manufactured entirely on Samsung’s 4nm process — arrived in Galaxy S22 phones plagued by overheating and poor power efficiency. The scandal did not just damage Samsung’s reputation. It relocated an entire customer.

Qualcomm moved every subsequent Snapdragon flagship to TSMC. The Snapdragon 8+ Gen 1, the Gen 2, the Gen 3, and the current Gen Elite line all shipped from TSMC fabs. Samsung’s foundry business absorbed the loss as a permanent fixture, branded in industry parlance as having an instability problem that no amount of marketing could shake off.

Now Samsung thinks it has fixed the problem. And the window to prove it is dangerously narrow.

The yield gap is closing, but it has not closed

According to industry sources cited by Chosunbiz, Samsung’s 2nm yield — the percentage of functional chips on each silicon wafer — sat in the low 20 percent range through much of last year. By mid-2024, it had climbed to roughly 50 to 60 percent. That is meaningful progress, but it is not enough to win a flagship order on its own.

TSMC’s 2nm yield sits at 60 to 70 percent, according to TrendForce data from April. More importantly, TSMC’s standard for mass-production qualification is understood to be around 70 percent. Qualcomm is widely expected to hold that line. Samsung has not reached it yet.

The numbers matter because yield is not just a cost metric. It is a proxy for process stability. A fab that consistently produces 60 percent good dies from a 2nm run is telling a customer something specific about how predictable its manufacturing pipeline is. That prediction is what Apple, Nvidia, and AMD pay for.

Capacity is the real story

Yield is the headline. But the quieter variable may decide the outcome.

TSMC’s 2nm capacity through the fourth quarter is estimated at roughly 60,000 wafers per month. Apple has already reserved more than half of that for the iPhone 18 and M6 processor lines. Nvidia and AMD have effectively sold through their 3nm allocations through 2027. What remains for Qualcomm — or any other customer seeking large volumes at short notice — is thin.

Samsung, by contrast, is not facing the same demand squeeze. Its Pyeongtaek facility has excess 2nm capacity that it wants filled. From a pure logistics standpoint, Samsung offers something TSMC cannot: availability.

That asymmetry is what makes a dual-source arrangement plausible, even likely for at least part of Qualcomm’s order. Industry observers have floated the possibility that Qualcomm may split its Snapdragon 8 Elite 6th-generation production between the two foundries — sending the base model to Samsung and reserving TSMC for the higher-margin Pro or Extreme variants.

Or Qualcomm may do what many assume and go TSMC-only again, betting that yield discipline trumps capacity concerns.

What is at stake for Samsung goes far beyond one chip

A Qualcomm 2nm win would be Samsung Foundry’s most consequential customer recovery in a decade. The financial impact is real — Qualcomm is among the top five foundry customers globally by revenue — but the strategic impact is larger.

Samsung has spent years trying to shed the label that followed it from the Exynos 2100 thermal controversy into the 2024–2025 period. A qualified 2nm relationship with Qualcomm would function as a market-certified seal of reliability. Other design-win prospects — Apple’s next-generation Wear OS chips, Nvidia’s mobile GPU lines, possibly even Amazon’s next custom silicon — would be watching closely.

Conversely, if Qualcomm sticks with TSMC for both Snapdragon 8 Elite 6th-generation variants, Samsung’s 2nm narrative remains incomplete. It would have demonstrated incremental yield improvement without achieving the breakthrough that changes customer behavior.

The Apple shadow over everything

The geopolitical subtext here is often missed in Korean and American coverage alike. Apple is the single largest consumer of TSMC’s most advanced capacity. Every wafer Apple takes at 2nm is a wafer unavailable to everyone else. Samsung is one of the few fabs with the technical capability to absorb overflow — if it can match TSMC’s yield and reliability thresholds.

Should Qualcomm commit significant 2nm volume to Samsung, it signals that at least one major American fabless company has decided the risk of taping out on a non-TSMC node at 2nm is acceptable. That is a threshold moment. The question is not whether Samsung can make a working 2nm chip. It already has. The question is whether it can make thousands of them, consistently, at the cycle times and power targets that Qualcomm’s architecture demands.

What happens next

Qualcomm is set to announce its next flagship at the Snapdragon Summit in Hawaii on September 22. The event will reveal whether the company is willing to split production between Samsung and TSMC or return to a single-source model. Until that announcement, the dual-source scenario remains a probability, not a certainty.

For Samsung, the calculus is straightforward: win Qualcomm at 2nm and you reposition the foundry business for the next generation of customers. Lose it and the yield gap becomes a credibility gap that will take another two node cycles to close.

Five years after the Snapdragon 8 Gen 1 thermal disaster, Samsung finally has a reason to believe the conversation has changed. The summit in Hawaii will confirm whether it is right.