A lot here is worth thinking through.
1. Everyone agrees 2027 is going to be huge. The real uncertainty is 2028. Look at what Oracle and the Chinese hyperscalers are saying. A decline in capex in 2028 is becoming the consensus view.
2. Anthropic still hasn’t convinced investors it can hit $150B in ARR by the end of 2027. Most people see around $120B as more realistic.
3. If I had to buy a semiconductor stock right now, TSMC would be my only pick. It runs its business as conservatively as possible, even if that means losing orders to competitors. Remember that recent Taiwanese media report saying TSMC reviewed customers’ financial statements and required cash up front if they wanted to place additional orders. TSMC really is the Coca-Cola of SOXX.
4. Many market research firms and sell-side analysts expect memory prices to start falling in Q3 2027. If it starts to look like prices can avoid a collapse in the second half of 2027, I think memory stocks could rebound as early as Q1.
5. With Jensen gone, CPO is the most at risk. He was the only one pushing CPO despite low yields, and now even Nvidia seems to be leaning toward NPO. Long NPO, short CPO.
6. 3–5 GW of capacity in Southeast Asia? That’s more than I expected.
Thanks for sharing these great insights from China,
@AgenticTeddy!