Patel says TSMC still uses DUV for seven nanometer despite having fifty EUV tools because DUV proved superior.
“So TSMC, to this day, even though they've been use they have over 50 EUV tools and they make hundreds of thousands of wafers a month with EUV, still uses DUV for their seven nanometer.”
Patel explains TSMC found DUV superior to EUV for seven nanometer on yield, cost, and performance after trying both.
“TSMC tried both. Right? And so at six nanometer and five nanometer, they use EUV, but they think cost efficiency wise, yield, right, which is how good the product's coming out, how many are good, how many need to be thrown away, the yield is better, the cost is better, the performance is better with DUV rather than EUV.”
Patel projects maximum 200 gigawatts of AI chip capacity by 2030 based on EUV tool supply.
“700 EUV tools by the end of the decade. 700 EUV tools, three and a half tools per gigawatt, assuming it's all allocated to AI, which it's not, but three and a half tools per gigawatt gets you to 200 gigawatts worth of AI chips for the data centers to deploy.”
Patel calculates one gigawatt of AI capacity requires only 3.5 EUV tools costing $1.2B.
“Because we're talking oh, what's a gigawatt cost? It costs like $50,000,000,000 roughly. Right? Whereas, what does three and a half EUV tools cost? That's like 1.2.”
Patel shows that $50 billion in data center capacity depends on just $1.2 billion worth of EUV tools.
“I need about three and a half EUV tools to do the 2,000,000 EUV wafer passes for the gigawatt. So three and a half EUV tools satisfies the gigawatt.”