Why TSMC abandoned the GaN market and what It means for the industry

TSMC sent shockwaves through the power semiconductor industry when they announced the wind down of their gallium nitride foundry service. How did they come to this decision? Was this an acknowledgement that GaN was never going to live up to its hype? Was it the end of the GaN foundry model?

Until this point, conventional logic dictated that GaN was more susceptible to the foundry model than silicon or silicon carbide. There was a much greater emphasis on chip design and a wider variety of products. Spin-outs and start-ups competed to define the future of GaN power electronics, with each defining their own path. The option of utilizing a foundry service helped these companies get off the ground and start providing products to customers without being too Capex heavy, making them much more attractive to investors. The fact that the world's leading foundry service, TSMC, was involved underpinned customer confidence in the technology's ability to scale.

Fast forward to now, and the industry is trying to understand the ramifications of TSMC's decision to make a tactical retreat. Growth in GaN power devices has been steady but not explosive. The upside is still high, but the path has become a lot more challenging.

For TSMC, they were undergoing the challenging task of upgrading their process from 150 mm to 200 mm. This move had extra significance to their customers due to the recent intense price competition in the GaN space. Innoscience, the leading Chinese GaN IDM, has huge capacity at 200 mm, while Infineon, the leading power IDM worldwide, has announced that their 300 mm process is approaching readiness. Keeping up with these and other players would have required serious amounts of investment at a time when TSMC was being inundated with requests for cutting-edge AI chips. Focusing there seems to make sense.

But this decision does not leave fabless GaN companies short of options. While TSMC's brand may have been appealing, there are other fabs available who may be able to provide closer collaboration. Navitas quickly moved to announce a partnership with Powerchip Semiconductor Manufacturing Corporation (PSMC), and services are also available from foundries such as Episil and DB Hitek. The foundry model will survive. The ability to support innovative fabless companies as they scale up, or to provide fast time to market for IDMs who are dipping their toe into GaN is valuable.

The TSMC decision highlights the need for a robust supply chain. Their exit from the market is being very well managed and the disruption should be low, but we may not always be so fortunate. The same concerns also apply to those who purchase power semiconductors. The richness of innovation in the GaN market is impressive. But it also poses a problem for a customer who requires a low-risk supply chain. A company like Nvidia can build a consortium to meet its needs, but for most others the movement needs to come from the supply side. GaN vendors should work together to ensure that the tide rises and pushes them all in the direction of prosperity.