Samsung’s 2nm foundry push: on schedule, but AI orders are not sealed yet

For investors, chip customers and supply-chain watchers, that distinction matters. Samsung appears to be moving closer to a credible 2nm ramp. But a foundry turnaround is not proven by process milestones alone. It will be proven when advanced-node designs become signed orders, stable yields and revenue.

The current position at a glance

What Samsung says is ready for 2026

In its fourth-quarter 2025 earnings conference call, held in January 2026, Samsung said its second-generation 2nm process was being developed with a target of mass production in the second half of 2026 and that work was progressing toward yield and performance goals . The company also said it was carrying out PPA evaluation and test-chip collaboration with major customers, with pre-production technology validation proceeding as planned .

In practical foundry terms, that places Samsung in a pre-ramp validation phase. PPA work is important because customers are checking whether a process can deliver the right mix of performance, power efficiency and chip area for a real product . Test chips also help customers and the foundry validate design rules, process behavior and manufacturability before high-volume production. Still, these steps are not the same as stable mass production, and they are not the same as a binding multi-year supply agreement.

The 2026 milestone is generally discussed around Samsung’s second-generation 2nm node, often referred to as SF2P. ZDNet Korea reported that Samsung planned to begin producing a new mobile product based on the second-generation 2nm, or SF2P, process in the second half of 2026 . That fits the broader roadmap Samsung presented at its 2023 Foundry Forum, where it said 2nm would expand from mobile in 2025 to HPC in 2026 and automotive semiconductors in 2027 .

AI and HPC are the prize, but not yet a closed deal

The commercial opportunity is clear. AI accelerators, custom cloud chips and high-end computing processors are exactly the kind of products that can justify the cost of leading-edge foundry nodes. Samsung said during its first-quarter 2026 earnings call that its foundry business was actively discussing 2nm collaboration with multiple large AI and HPC customers and expected visible progress with some customers in the near term . EDaily reported the same call in similar terms, noting active 2nm discussions with large AI and HPC customers .

That is a meaningful signal, but it stops short of a confirmed breakthrough. The public wording so far centers on discussions, collaboration and expected visible results . It does not publicly identify the customers, product lines, wafer volumes or timing of production revenue. In other words, Samsung’s AI and HPC funnel may be improving, but public information does not yet support the conclusion that major 2nm AI or HPC volume contracts have been locked in.

Why the order-growth numbers need caution

Several Korean reports point to an expanding 2nm opportunity set, but they do not all use the same figure. EToday and Seoul Economic TV reported that Samsung expected 2nm order projects for AI and HPC applications to grow by more than 130% year on year . Global Economic and ZDNet Korea, by contrast, reported a figure of more than 30% year on year .

The safer interpretation is not to treat either number as the final measure of commercial success. The consistent message is that Samsung expects more 2nm opportunities around AI and HPC. The open question is whether those opportunities become design wins, wafer commitments and revenue. A larger project pipeline is encouraging; a confirmed production ramp is the stronger proof.

Samsung’s pitch: more than a 2nm process

Samsung is not selling 2nm in isolation. Its broader argument is that it can bundle advanced foundry manufacturing with memory and packaging. Chosun Biz reported Samsung’s first-quarter 2026 call as saying that ongoing memory supply constraints were increasing customer demand for turnkey arrangements, and that 2nm collaboration was becoming more visible among AI and HPC customers .

That matters because AI hardware customers often need more than a logic chip. They need high-bandwidth memory, advanced packaging, and the ability to integrate multiple components efficiently. Samsung has also said that its 4nm-based HBM4 base die is being recognized for strong performance and is helping drive 4nm demand, while the company is reviewing ways to expand supply . HBM4 work does not automatically translate into 2nm contracts. But it supports Samsung’s broader one-stop pitch across foundry, memory and advanced packaging .

Silicon photonics is another part of the same strategy. Samsung said demand is growing for high-bandwidth, low-power data transmission in data centers and that it is developing technologies that combine devices, advanced processes and 3D packaging . The company also said it was discussing commercialization with multiple large global silicon photonics customers and that it expected to begin project mass production with a major optical communications module company from the second half of 2026 .

The customer base is widening beyond AI servers

AI and HPC remain the main growth story, but Samsung is also trying to broaden the customer base for its advanced nodes. The company said it was discussing adoption of 2nm and 4nm processes with automotive and robotics customers in the Americas and Greater China . Ajunews and EDaily also reported that Samsung was in talks with automotive and robotics customers in those regions over 2nm and 4nm process adoption .

That diversification is useful. Automotive, robotics and optical communications could give Samsung more paths to fill advanced-node capacity. Even so, the commercial turning point for 2nm will likely still be judged by large AI and HPC products, because those chips can bring bigger volumes, higher strategic value and stronger validation for a leading-edge foundry process.

What to watch next

1. More specific customer evidence
The key upgrade would be a move from general language about discussions to concrete details: customer names, product categories, tape-out milestones, wafer volumes or production timing. Current public reports emphasize active talks and expected visible progress .

2. Yield and performance at scale
Samsung says the second-generation 2nm process is being developed toward yield and performance targets, with PPA evaluation and test-chip collaboration under way . The next test is whether that validation translates into repeatable high-volume manufacturing.

3. Pipeline conversion into revenue
The reported growth in 2nm order projects is encouraging, but the cited growth rates differ across reports . The more important question is whether those projects become signed production commitments and then sales.

4. Advanced-line ramp-up
ZDNet Korea reported that Samsung’s advanced-node line utilization had reached a maximum level, while Newsis reported that Samsung was preparing 2nm-based semiconductor production at its new factory in Taylor in the United States . If Samsung is to meet the second-half 2026 target, the stability of advanced manufacturing lines will be as important as customer announcements.

Bottom line

Samsung’s 2nm foundry program is showing real signs of progress. The company is still targeting second-half 2026 mass production for its second-generation 2nm process, is working with customers on PPA and test chips, and is seeing more visible AI and HPC engagement .

But the public evidence still calls for caution. Samsung’s position can be summed up as: development and validation are advancing, customer prospects are improving, and commercial certainty remains limited. The market’s view will change more decisively only if major AI and HPC customers move from talks to identifiable contracts, stable yields and meaningful production revenue.