Artificial intelligence chip designers are turning to TSMC instead of Samsung.Many South Korean companies that previously worked with Samsung Foundry are turning to TSMC’s more advanced manufacturing technologies for their next-generation AI processors.
AI companies choose TSMC over Samsung
Several South Korean chip designers are choosing TSMC’s manufacturing architecture over Samsung Foundry for their next-generation AI processors. Companies such as DeepX, FuriosaAI and Mobilint are choosing to use TSMC’s process technologies for their chips.
While the companies’ earlier products will continue to be manufactured at Samsung Foundry, the next models will utilize TSMC’s production. DeepX will switch from Samsung’s 5nm node to TSMC’s 12nm process for its upcoming DX-V3 processor.
Google could use nuclear reactors for its continuous and energy-intensive operations, such as artificial intelligence.
FuriosaAI preferred Samsung’s 14nm architecture for its first chip named Warboy. It has switched to TSMC’s 5nm architecture for its more powerful Renegade AI chip. Mobilint will use TSMC’s 12nm technology for its low-power Regulus NPU chip instead of Samsung’s 14nm process.
The companies cite factors such as power, performance and cost optimization for the partnership changes. TSMC’s more advanced nodes provide better efficiency and pricing than Samsung technologies for next-generation designs.
Diversifying suppliers provides advantages such as backup options. But for small startups, it comes at the cost of additional resources. For the three Korean firms, however, TSMC offers a more competitive solution for scaling to their next products.
