TSMC(2330.TW)2021 RECORD CAPEX：TSMC PUTS ITS MONEY WHERE ITS MOUTH IS

LT (mid-point) growth guidance lifted by 5%: 4Q20 EPS of TWD5.51 was 5%/4% above our/Bloomberg consensus estimates as operational efficiency outweighed a less favorable FX backdrop. 1Q21 mid-point sales guidance (-0.7% QoQ in TWD) beat our/consensus estimates on particularly strong N5-7 loading. Its 2021 US$ sales growth guidance (mid- teens YoY post 2020’s 31% growth), in our view, merits TSMC to lift its 2021-25 projection to 10-15% CAGR in the HPC era (prior: 5-10%) amid an unchanged structural GPM (50%).
Strategic value to grow further at long-lasting N3-5: While the market may construe TSMC’s 2021 record-high CAPEX guidance (US$25-28bn) as related to Intel’s CPU outsourced orders at N3 in 2022, we see it indeed conveying its growing conviction on N5 volume prospects: The 2021 CAPEX is largely for N5 expansion (at least 2x its current 55-60k WPM capacity), and N3 build is limited to pilot line/initial production (+10-15k WPM, though we expect N3’s capacity to be similar to N5’s upon full ramp-up). Despite advanced nodes’ rising CAPEX density (Ex.1), we note the bold expansion reflects N3-5 as long- lasting big nodes with expected robust future demand beyond the initial wave from key clients. We see TSMC’s strategic value rising with N3-5: strong N5 tape-out activities from HPC/ smartphone applications and more N3 client engagement vs N5/N7 at similar stages.
Despite its debt-funded CAPEX (+US$10bn YoY; supported by US$8bn bond issued in ’20), its net cash status (2020: 24%) still supports its sustainable dividend policy.
Unrivalled transistor (Tr) density offering matters: We expect TSMC will still lead its peers on Tr density offering (a key chip PPA metric), even if Intel sticks to its tech roadmap.
When Intel is 7nm production-ready (2x Tr density gain vs its 10nm) in 2022E, TSMC would also be at N3 production, for which we estimate Tr density above Intel’s 7nm. Maiden EUV adoption at 7nm may also present yield challenges. We see TSMC’s unchanged Tr structure (FinFET) at N3 as a prudent way to extract the most out of Tr structure since N16. Despite Samsung’s earlier shift to GAA at 3nm, its estimated Tr density will likely still be below that of TSMC’s N3. It is the ability to deliver PPA merits that clients value most, in our view. We expect imminent Intel CPU foundry outsourcing (TAM: US$20bn p.a. vs TSMC’s 2020 sales of US$45.5bn) and TSMC should benefit most, even if Intel opts for multi-sourcing, if order allocation is based on vendors’ PPA merits/cost proposition.
BUY and Top Pick; new TP: TWD865.00 (1-yr disc. 40x ‘22E P/E; prior: 30x ’21-22E avg). We see more LT investors valuing TSMC on expected profit power in the new CPU foundry TAM, where it should gain traction from ‘22E, and willing to further re-rate it on: 1) TSMC’s doubling LT mid-point growth (in the past, P/E was capped at 20-25x); 2) better clarity on its foray in CPU outsourcing; 3) more inflow from TMT-focused funds switching out of China’s ecommerce sector. Risks: Worse demand, industry rivalry, pace of tech progress.
【免责声明】本文仅代表第三方观点，不代表和讯网立场。投资者据此操作，风险请自担。

   【免责声明】本文仅代表第三方观点，不代表和讯网立场。投资者据此操作，风险请自担。