TSMC--FY25--Top holding reports--FAB leader Riding the AI wave

 REPORT ON TSMC FY25

Summary

The numbers TSMC are producing are very good, even excellent. I have analysed them below. The overriding story here is that the AI investment chain needs to succeed. That means use cases and token usage need to keep growing to fill the DCs and, in turn, TSMS’s new FABs. Data centres are the largest growth area, AI use cases require more complex nodes that are being developed, and TSMC is planning huge capex into FABs to supply the chips that will be used to satisfy this demand.

The semiconductor value chain is incredibly integrated. If AI progress pauses or slips and is shown to have issues being implanted and generating ROI, the whole chain will be impacted. There are portfolio consequences for this, and this correlation must be considered. My base case is that AI will succeed, but there will be bumps in the road, maybe some large ones. Therefore, the overall exposure needs to be assessed for risk control.

Notes from FY25 results.

Main Points

1.      TSM delivered revenues up 38% yoy, NPAT +53% and ROE 31.5%. All better than my estimates, see below.

2.      TSM are satisfied with interaction with the end customers, ie the hyperscalers, that demand is long-lasting, with 2-3 year planning timeframes mentioned. Important to this view is that the efficiency use cases and token usage continue to expand at a rapid rate. TSM notes that it has seen AI improvements at its own FABs and believes the benefits are tangible.

3.      GM increased to 62.3%, very high due to improvements in capacity utilisation. Q126 guide 63-65%

4.      TSM FY26 revenue growth forecast at +30%, and 5Y revenue growth cagr 2024-29 of 25%.

5.      Capex was FY25 UD$41b and is forecast at $52-56B for FY26. TSM believe that silicon is the bottleneck as volumes and complexity increase. Mentioned the increasing cost of FAB tools, good for the Fab 5. Productivity gains are the focus for 2026 and 2027; the large capex increases will impact 2028 and 2029.

6.      The overseas FAB build continues, and the incremental news is positive. Although dilutive, they appear to be progressing better than expected. Some mention of US yields approaching the Taiwanese levels.

7.      TSM mentioned that cross-node capacity is being utilised, which is allowing greater efficiency. That is seen as a significant competitive advantage. That is capacity from older nodes is being used for the newer nodes savings costs. Other advantages mentioned are better power utilisation and speed of production, innovation, and product leadership are maintained.

8.      The large increase in memory costs was not viewed as a barrier to growth in TSM's higher value skew smartphone market clients.

9.      Advanced nodes continue to take a greater share of the revenues, N3 is expected to cross the corporate average margin, while N2 will be dilutive at the start. The more complex nodes appear to be more profitable. N2 should do better than N3- positive.

10.   The results were well above my estimates at the NPAT line, driven by huge +GM. Last 4 years versus my estimates.—pretty good record.

BUSINESS RISKS

TSMC remains the leader in chip manufacturing, with Samsung and Intel both behind. The market share graph is below. Intel has made a couple of efforts to catch up, but the base case is that these competitors will continue to lag. As the largest FAB operator, TSM has an advantageous position with equipment suppliers, although skilled labour acquisition, especially overseas, will be a struggle. At this stage appears manageable. Customers are NVDA, AAPL, AMD, as well as a slew of chip designers. The end customers through NVDA and AMD are the hyperscalers such as GOOG, AMZN, META and MSFT. The growth plans of these end customers remain vital to TSMC’s growth path. The market size is ultimately dependent on the ongoing use cases driven, at this stage, through the hyperscalers, but likely to spread to sovereign and wider enterprise demand.

Geopolitical risk exists, and the base case is that any Chinese invasion will likely lead to a huge escalation that will see all incumbents worse off. Taiwan remains too important at this stage. A scenario where China considers invasion due to not being able to access Taiwanese technology remains the most probable but low chance risk to the base case.

The tables of revenue splits below clearly show the growth drivers and evolution of the business. The business is being driven by North American demand for high-powered computing, with an ever increasing demand for the latest and more complex nodes. Clearly, data centre demand and lately accelerated computing are becoming a more important story. These numbers are good for mix and profitability. Plans exist for 2nm and smaller, more complex nodes after that. 

OPERATING HISTORY

The GM, eps, ROE and return on capital have all been positive. There is a volatility involved, which is, imo, due to the capital intensity of the business. Pricing power does quite clearly hold. GM has increased despite the dilution of the offshore FABs and newer nodes, due to operating capacity and cost discipline. Pricing is important as well, from a cost recovery perspective. TSM is a strong business. 

CAPEX

TSMC guided to a range of USD$52-56B in capex for 2026. Unlike many others in the semiconductor chain, TSM is very capital-intensive; it has to commit massive amounts of capital to the new FABs. Therefore, we can see that TSM is heavily increasing capital spend and obviously needs increased and sustained demand to fill these new FABs. Any significant shortfall in spend will result in operational deleveraging, a terrible outcome. Highlights the major risk, IMO. TSM notes that the current capex spend will generate revenues in 2028/9.

Management

TSM management remains best in class, with operational and strategic leadership. The management is conservative, and guidance is usually exceeded.

Balance Sheet

TSM remains in a strong net cash position and has been over the LT. Cash flow generation is strong, easily funding capex as shown in the chart below.

OTHER ISSUES

The market's main concerns are geopolitical, including a Chinese invasion and secondly, the sustainability of the AI-generated demand, since TSM is committing the capex to build the chips. The market agrees that TSM is the leader in its industry, and the growth outlook is positive.

Reconciling CFO to NPAT+ depreciation shows some volatility, but fine over the 5 year period.  The business cash can swing with the semi-cycle; TSM absorbs cash through the cycle, which is why the balance sheet needs to stay strong.

Cash coverage
3year	5 year
95%	100%

 

VALUATION

The base valuation is as follows, at $342, assuming 23% 5Y eps cagr and a 21X PE exit generates 14%pa returns, very attractive in my world. Of course, this all hinges on the AI trade playing out well. If demand disappoints and TSM suffers operational deleveraging, then the earnings and PE both could suffer. From a portfolio perspective, it comes to exposure to the AI trade in total and through which stocks specifically. A reasonable exposure is warranted, imo, but I play that through the stronger names in the semi value chain, of which TSM is a main player. TSM remains a significant holding.

Please note disclaimer