TSMC – All about AI

50% inventory, 50% AI.

• TSMC has signalled better times ahead for the semiconductor industry, but I suspect that things will only be really rosy if one is selling chips that enable the current insatiable demand for AI training.
• TSMC reported better than expected Q4 2023 results and forecast that 2024 would see a return to growth which buoyed the entire semiconductor sector in the hope that the consumer-led, inflation-driven downturn of 2023 is over.
• Q4 2023 revenues / EPS were TWD547bn / TWD9.21 just ahead of forecasts of TWD541bn / TWD8.65 but this was already fairly obvious to the market given the monthly revenue reporting that Taiwan requires.
• What has driven the TSMC GDR in New York up 10% is the outlook where TSMC expects a smaller than average seasonal decline in Q1 2024 and where 2024 revenues will grow by “up to 25%”.
• TSMC expects the overall industry to grow by 10% in 2024 and that its outperformance will be driven by its No. 1 position in leading-edge manufacturing and its position in AI.
• Anyone who is making AI chips seems to want to use TSMC at the moment and Nvidia has more short-term demand that it can deal with.
• This means that Nvidia’s 2024 outlook is more about how much capacity it can buy from TSMC than demand from its customers.
• It is also clear that the inventory correction that consumer electronics suffered last year is now pretty much in the rearview mirror which Qualcomm signalled at its last set of results in November 2023.
• This stabilisation looks to have spread beyond smartphones into other areas with the one exception being automotive where Mobileye has signalled a significant correction.
• Consequently, the outlook for the semiconductor sector is improving with only two clouds on the horizon.
  • First, China: where there is no sign of an improvement in either the economy or the state’s policy towards the technology sector.
  • Further tightening of restrictions on imports into China from the US Department of Commerce or outright bans on USA technology from the Chinese state remain possible and will cause sector-wide consternation should they occur.
  • Despite efforts by China to fill the gap in leading-edge semiconductors, I think that China is miles away from being able to manufacture these chips economically and there is only so much money the CCP can pump into this sector.
  • Hence, there is a rising (albeit small) likelihood of real retaliation against US companies that drive a significant part of their revenues from China.
  • Apple would be the worst affected, but Qualcomm would also take a hard hit if its products were banned in China.
  • The chances of both of these remain remote as a hit on Apple or Qualcomm would have a substantial impact on Chinese companies operating in China.
  • Second, AI bubble: which I am certain that we are in, but I have no way of knowing how long it is going to last.
  • A bursting of the bubble would lead to a decline in sentiment and investment meaning that demand for AI chips would almost certainly decline from its current fever pitch.
  • This would reduce TSMC’s outlook significantly and would almost certainly trigger a sell-off of the sector more generally and the AI-exposed companies most of all.
• The net result is that I do not want to chase the higher-valued end of the semiconductor sector which will correct hardest should the AI bubble pop.
• Instead, I remain very comfortable with my position in Qualcomm and would not be adverse to a position in either TSMC or MediaTek where the outlook is good and the valuation much more reasonable.