These days, you can hardly read a semiconductor news story without hearing about substrate bottlenecks. Okay, so we get it. But do you know which substrates are we actually talking about? In this article, I’m gonna explain the types of substrates and what they do, then looks at why AI chips use more substrate, why substrate supply cannot be expanded quickly, and why larger substrates yield fewer good units. It then briefly covers where the supply chain is most blocked, which companies keep the profits, and who secures substrate supply. I am confident that once you have read it, you will be able to tell for yourself which companies each piece of substrate news favors.

See? What did I say? Value for money and word of mouth are the best marketing. Kimi’s model isn’t bad, but its usage quota is the lowest among all coding plans I’ve subscribed to.

Great insight. If you're into $CRDO, follow @TransLog1c. Adding on: We joined a group meeting with $CRDO early investors and $CRDO R&D execs last quarter. Walked through the full AEC arc — from first design to today's near-monopoly scale. Credo's team is materially stronger than the market prices in. Their forward judgments and executions are hard for peers to match. Investing in $CRDO now for ALC + OmniConnect feels a lot like that PE investor buying the AEC story early — same pattern, now with a proven track record and a ridiculously cheap fwd multiple.

Order uncertainty and concerns about China localization have been suppressing WFE names. Apart from customers’ interest, China is now placing orders for US WFEs. Ryan’s industry insights are never wrong. $ASML $AMAT $LRCX $KLAC

So, China may not allow AI labs to release their technical reports or share code for their algorithms and inference stack. This will be pretty bad for the ecosystem. "China isn't simply aiming to restrict product exports, but rather to prevent overseas entities from replicating the relevant underlying technologies and limiting their access to the technology. For example: China won't restrict the export of final chip products, but will restrict the transfer of architectures, IPs, and software to foreign entities."

$NVDA Zephyr's reading of China's draft is right, and the annex backs it almost word for word. The AI chip entry covers "chip architecture, self-developed instruction sets, AI IP cores and the supporting software stack," the same list he gives. China keeps selling the products and puts a license on the know-how behind them. The model entry is where DeepSeek comes in. Model weights, checkpoints, source code, and training and inference techniques need a license if any one of three tests is met: 10^26 FLOPs of training compute, 2 trillion parameters, or a top-10 place in public global rankings, which the draft says include usage. DeepSeek's V4.1-Flash-1017, released the day after the draft, has 763 billion parameters and no published compute figure. The V4.1 Flash it replaces was No. 1 by tokens on OpenRouter last week. On open source, the draft leaves the door open for now. Technology already public through open source or papers is out of scope, and open-sourcing these items gets an "automatic license." DeepSeek put 1017's weights on Hugging Face under an MIT license on Saturday. Cheryl Wu's framing fits: the state now holds a switch, and today it's on. Taiwan is tightening from the other side. According to Bloomberg, its economy minister said AI servers and advanced chips will be added to export controls for China, with unlicensed shipments becoming a crime. There's no start date yet. DeepSeek's own Ascend tests in September show why that matters. Its matrix kernels reach 99.8% of a single Ascend 950DT's limit, but all-to-all bandwidth drops from 374 GB/s at 8 cards to 317 GB/s at 128. The single chip is close to done, and linking thousands of them is still the gap. China's draft lists cluster interconnect at 800 GB/s or more as restricted technology. What China has is price and reach. 1017 lists at $1.20 per million output tokens at peak, against $20 for Claude Opus 5.5. Chinese models held 7 of OpenRouter's top 10 last week and 84% of the tokens those ten processed, and CXMT took 9.5% of global DRAM revenue in 2Q26. Our read on the way out: ship open models that stay under the size lines, keep the frontier know-how at home, and build compute where chips can still be bought. The draft's automatic-license list already reads that way, since open-sourcing, Chinese-owned compute centers abroad and taping out chips at overseas fabs are all fast-tracked. The cost lands on Zephyr's ecosystem point. If the usage test survives into the final version, China's most-used open model would sit above a control line, and each open release would depend on the state keeping the switch on. Comments close Nov. 15. The clause we'd watch is the usage ranking in the third test.

“China is now expected to localize more than half, or close to half, of most equipment categories, excluding lithography and track. What does this mean? It means China can now expand mature node capacity far more easily than before, using equipment made domestically within China.” Interesting thought. The only remaining bottleneck(excluding litho) for China is the localization of equipment components, and it is currently facing a component shortage. I guess they can localize these components within the next 5–10 years—it’s just a matter of time🤔

Let’s recap what CC Wei said: “To verify whether AI demand is real, TSMC has even sent teams to the United States to inspect data center construction progress and power supply conditions on the ground.” They have conducted extensive fieldwork, much like @SemiAnalysis_ . How can anyone view them as reckless or uninformed?

“On the ABF substrate side alone, Liwei is hearing that customers have already put down deposits to secure capacity through 2032. Do you seriously think Taiwan’s suppliers would expand without customers putting money on the table first?” Taiwan’s suppliers understand both the present and future demand from CSPs, and CSPs understand the present and future demand for AI adoption. They see strong demand for the next few years, but Ryan don’t want to stop the market from FUD because you can always buy stocks at a much cheaper price😉