Samsung Electronics and Arm are reportedly developing a 2nm AI system-on-chip for smartphones and other smart devices, giving Samsung another chance to prove its advanced manufacturing process on custom AI silicon.
Arm is expected to provide AI accelerator architecture and core design technology, while Samsung’s System LSI division handles the SoC design and Samsung Foundry manufactures the chip using its 2nm process. BigGo Finance reported that Arm approved the project’s initial non-recurring engineering costs in late August, allowing development to move forward.
The project arrives as more AI workloads shift from cloud infrastructure onto local devices, where lower latency, reduced cloud dependence, and potential privacy benefits are becoming increasingly valuable.
The partnership comes as AI processing increasingly moves from cloud data centers onto consumer devices. Running some AI workloads locally can reduce latency, limit reliance on cloud infrastructure and offer privacy benefits.
OpenAI link remains unconfirmed
The project has another potentially significant angle: OpenAI is reportedly being considered as a customer. Neither Samsung, Arm nor OpenAI has publicly confirmed that OpenAI will buy or use the chip, so that connection remains speculation.
If OpenAI does become the customer, the project would give Samsung a high-profile opportunity to demonstrate that it can design and manufacture customized AI silicon for an AI company rather than simply produce chips for traditional consumer electronics.
Why Samsung needs this win
On-device AI is becoming more important as companies look for ways to run AI features without sending every task to a remote data center. Processing more workloads locally can reduce latency and cloud dependence while offering potential privacy and power-efficiency benefits.
The bigger prize may be what this project could say about Samsung’s foundry business.
The company would be combining two capabilities under one roof: System LSI would handle chip design and Samsung Foundry would handle manufacturing. If the chip reaches mass production, Samsung could use the project as evidence that its 2nm technology can support demanding AI workloads, potentially strengthening its pitch to future custom-chip customers.
That is particularly important as Samsung works to improve its advanced-node manufacturing. BigGo Finance reported that Samsung’s foundry utilization had recovered to about 70% to 80%, while 2nm yields remained an important factor in the division’s return to profitability.
The bigger semiconductor race
The timing is significant because custom AI silicon is becoming an increasingly competitive market. Broadcom is already working with major AI companies, while Google is diversifying its custom-chip supply chain and Anthropic is building its own custom-silicon capabilities.
Samsung and Arm’s project approaches the market from a different direction: smaller, local AI systems rather than massive data-center processors. That could prove strategically useful. The project may not rival the enormous economics of data-center AI chips, but consumer devices offer a potentially much broader volume opportunity.
The catch is that none of this becomes meaningful until the chip reaches production. Customer confirmation, technical performance, 2nm yields, and mass-production stability will determine whether the partnership becomes a major commercial win or simply another promising semiconductor development project.
Other News: Nscale is reportedly seeking $3.5 billion in pre-IPO funding, including a potential $2 billion investment from Nvidia, as the AI cloud provider races to expand its data center and GPU capacity.