The 2nm race is over before it started. Not because TSMC's transistors are better. Because ASML's 2027 EUV pool is nearly sold out. Yield press releases are theater. Machines print wafers.
The 2nm race is over before it started. Not because TSMC's transistors are better. Because ASML's 2027 EUV pool is nearly sold out. Yield press releases are theater. Machines print wafers.
This breaks if ASML's Oct 14 print guides 2028 low-NA EUV output above 120 systems, or confirms 2027 capacity expanding past 90. More tools means no zero-sum. Bookmark it.
TSMC just printed its best month ever: NT$514.8B in August, +53.3% YoY and +10.1% from July. Four record months in a row. Jan-Aug revenue is already NT$3.39T (+39.3%), the "slightly above 40%" full-year guide looks conservative. $TSM
Marvell just guided to $80B in revenue by 2031. It also guided gross margin DOWN. 56-59% long term. Versus 59.5% today. Nobody cheered that second number.
Two doors. Marvell the plumber: $37.5B interconnect, real moat. Marvell the XPU house: $30B custom, riding on Google's orders. XPU means custom AI chip. Which one are you buying? Pick one.
"My motto is very simple: under-promise and over-deliver. That has been my trademark." - LBT
Any of you dorks who think $Intc will spill its partners' names before the partners drop Intel's name first deserve to sell your shares. The Intel Foundry business is strictly hush-hush