Samsung Wins AI Chip Foundry Orders from U.S. Cloud Giants for Second Consecutive Quarter, Mass Production Set for 2028
nashnova research
Samsung Electronics has signed AI chip foundry contracts with major US cloud providers two quarters in a row, with commercial shipments expected in 2028 — driven directly by TSMC's tightening grip on advanced-node capacity.
Two quarters of wins — what exactly was signed?
Korean outlet Ddaily, citing industry sources, reported that Samsung signed foundry contracts with two major US cloud providers in Q2 and Q3 2026. The orders cover application-specific integrated circuits (ASICs) — chips custom-built for AI workloads, not general-purpose processors.
Samsung confirmed on its Oct 29 earnings call that it won 2nm projects from cloud and AI-HPC clients in two consecutive quarters, but disclosed neither the client names nor the contract values.
This means → these are not exploratory test runs. Back-to-back signings signal that cloud companies' confidence in Samsung's foundry is building quarter by quarter.
Why are cloud giants turning to Samsung?
The most direct reason: TSMC doesn't have enough capacity. TSMC's 2nm node accounted for 8% of its wafer revenue in Q3, up from just 3% the prior quarter. Demand for advanced nodes keeps climbing, and the queue keeps growing.
In plain terms = TSMC's waiting list is too long. Cloud companies designing their own AI chips can't afford to wait, so they're booking a second shop.
Samsung's pitch is the bundle: advanced logic fabrication + HBM high-bandwidth memory + chip packaging, all under one supply chain. This reflects a real pain point for cloud clients — coordinating multiple vendors slows everything down.
What's the Faraday Technology deal about?
Taiwan-based chip design house Faraday Technology (智原科技) announced on Oct 27 that it won a 2nm GAA cloud-AI ASIC design project from a US client, with Samsung handling fabrication. The contract is worth tens of millions of dollars and is expected to start contributing revenue in H1 2027.
This means → Samsung's 2nm foundry already has a live design partner on the ground, not just a signed letter of intent.
One caveat: Faraday's contract has not been publicly linked to the cloud-provider deals reported by Ddaily — the two may be related, or they may be separate projects.
Can Samsung's capacity keep up with 2028 mass production?
Samsung has acknowledged that demand for 8nm and more advanced nodes is expected to exceed existing capacity by 2027. It is expanding 8nm, 4nm, and 2nm lines simultaneously.
The linchpin: Samsung's new fab in Taylor, Texas is scheduled to begin operations in 2027. The first cleanroom is active, equipment installation is underway, and qualification wafers are already being run.
In plain terms = the factory is going up, the machines are going in, and test production has started — but the gap between "test run" and "stable mass production" is still real.
Has the profitability problem been solved?
Not yet. Samsung does not break out foundry operating results separately. Foundry chief Han Jin-man said in June that full-year profitability in 2027 will be difficult; a turnaround is more likely in 2028.
This means → winning orders and reaching mass production is one thing; making money is another. Korea Investment & Securities analyst Ko Young-min expects these contracts to start contributing foundry revenue only in 2028.
This reflects the core tension in Samsung's foundry business: clients are arriving and contracts are being signed, but upfront expansion costs are enormous. Whether Samsung can hit mass production and profitability at the same time in 2028 is the make-or-break test for this recovery story.
市场有风险，内容仅供研究参考，不构成投资建议。