AMD CEO Lisa Su Says Sourcing AI Chips From TSMC’s U.S. Plants Is 20% More Expensive, Highlighting the Complications of Building Supply Chains in America

While TSMC US operations are seeing massive attraction from American clients, it is still a very expensive venture, according to AMD’s CEO Lisa Su.

Getting Chips From TSMC US Is a Costly Venture For Big Tech, But They Have No Other Choice For Now

TSMC has expanded in America at a massive pace that no one could predict. The Trump administration catalyzed the efforts of the Taiwan chip giant, which was against sourcing chips from Taiwan, which is why TSMC announced significant investments in the US. Despite seeing massive demand, TSMC’s production in the US is 20% more expensive compared to what is available in Taiwan, and this was revealed by AMD’s CEO Lisa Su herself while talking at an AI event in Washington (via Bloomberg).

The American chip supply chain before TSMC was at a “negligible” position, with no other firm readily producing cutting-edge nodes in the nation. Apart from Intel, America didn’t have any other chip company capable of catering to the market demand, but fast forward a few years, TSMC started this with their Arizona facility, and now, they intend to scale up production even further, but the business in America is pretty expensive, which is why clients like AMD and NVIDIA are to pay 5% to 20% higher in order to source chips in the US.

There are plenty of reasons why producing in the US is expensive, but one of the more notable ones includes expensive labour, higher costs associated with importing equipment and setting up facilities, and the naivety of the chip supply chain. Despite being a costly venture, companies are still ready to get chips from TSMC US, mainly since the production lines in Taiwan are fully populated right now, and the second option for Big Tech is to get expensive nodes from Arizona, and it seems like this is happening, since the Taiwan giant has reported massive success with its US operations.

AMD is one of TSMC’s bigger customers in the US, being one of the first to place 4nm orders with the Arizona facility, and even has plans to scale up to 2nm, particularly for EPYC Venice data center CPUs. AMD’s CEO claims that the demand for AI chips isn’t slowing down anytime soon, and its partners are placing orders at an unprecedented pace, which is why she expects the total accelerator market to reach up to a whopping $500 billion in valuation in the next five years.