Samsung’s ambitions to become a go-to silicon partner for the AI industry have hit a double setback. The South Korean tech giant has paused development on a custom artificial intelligence chip for Meta after the social media company asked for a hold on the project, according to sources familiar with the matter. The collaboration, which had been running since last year under Samsung’s System LSI division, was expected to be worth tens of trillions of Korean won—tens of billions of dollars—if it reached mass production. Meta described the move as a “temporary hold,” not a cancellation, but declined to provide a timeline for resumption or an explanation for the pause.
A pattern of stalled projects
This isn’t an isolated incident. Samsung’s custom system-on-a-chip project with OpenAI also stalled recently due to strategic differences, sources say. Two major AI customer projects hitting the brakes in quick succession raises uncomfortable questions about Samsung’s logic chip and foundry strategy. The company has long struggled to compete with TSMC in advanced manufacturing, and these high-profile pauses could further erode confidence among potential clients. For Meta, the decision may reflect a broader reassessment of its custom silicon needs—the company has been investing heavily in its own in-house chip designs and could be rethinking its reliance on external partners.
A strategic pivot or a warning sign?
Separately, Samsung recently participated in Anthropic’s Series H funding round as a strategic infrastructure partner. Notably, Anthropic explicitly mentioned “logic chips” in its announcement, a detail that industry observers see as a potential future customer signal. If Anthropic is looking for a custom chip partner, Samsung could be positioning itself for that role. But the timing is awkward: with two high-profile projects on ice, Samsung’s ability to deliver on custom AI silicon is under scrutiny.
The custom SoC team
Samsung had established a dedicated Custom SoC team in November 2025 under Vice President Park Bong-il, with about 70 engineers. The team was tasked with designing specialized chips for AI workloads, a market that Nvidia currently dominates with its GPUs. But building custom silicon for AI is notoriously difficult—it requires close collaboration between chip designers and the customer’s software stack, and any misalignment can derail a project. The pauses with Meta and OpenAI suggest that Samsung may be struggling to meet the demanding requirements of the world’s leading AI labs.
For now, Samsung insists its custom chip ambitions remain intact. But in the fast-moving world of AI hardware, a “temporary hold” can quickly become permanent if the customer finds another partner. The company needs to demonstrate that it can execute on these complex projects, or risk being left behind in the race to power the next generation of AI.
