HUGE: Huawei's 1.4nm Claim Has The Chip Industry Talking

追加: 2026-05-25

[00:00:00]You know, this could be the end of Moore's law, the end of the idea that the only way to build a faster chip is to make it transistors smaller. For over 50 years, the semiconductor industry lived by one rule. Make things smaller.

[00:00:15]Smaller transistors, nodes, even smaller nanometers. From Intel to TSMC to Nvidia, the entire industry has been racing down the same road. But now QA says the road itself is the wrong path and instead of chasing smaller transistors, it wants to chase something else entirely which is time. Hua calls it the thao scaling law. And if it works, China may have found a new way to compete in the global chip race without ever catching up in manufacturing. So well today on front page we are going to speak about one of the most ambitious semiconductor announcements to come out of China in years. At the ILE E international symposium on circuits and systems in Shanghai Uway unveiled what it calls the TAO scaling law a new framework that it believes can take the semiconductor industry beyond the limits of MOS law. The company says future Kirin processors launching later this year will be the first chips built on this approach and by 2031 Hua believes it can achieve transistor densities comparable to 1.4 nanometer chips. Quite an extraordinary claim I must say especially from a company that remains under some of the toughest semic semiconductor sanctions in the world.

[00:01:36]Well to understand this better let us try to simplify this. For quite some time, Moo's law guided the chip industry. The idea was every couple of years, transistor density doubles. More transistors means more computing power.

[00:01:50]The formula worked beautifully well.

[00:01:53]Computers phones got faster and AI became possible. But there was a challenge. Eventually, physics fought back. When transistors became so small that the electrons started behaving unpredictably, making them smaller became incredibly difficult and incredibly expensive. Today, moving from one generation of chips to another cost tens of billions of dollars. The gains from that exercise are shrinking and the costs are ballooning, which the industry knows. So, naturally, everyone is looking for alternatives. Huay's argument is surprisingly simple. coaxing everyone to stop obsessing over transistor size instead focus energies on the signal travel time. So instead of asking how small can we make this transistor, one needs to ask how fast can information move through this system. That's where toao comes in. Tao of course represented by the Greek symbol of the same word measures delay.

[00:02:55]the amount of time it takes signals to move through a chip. Huawei's belief is that reducing delay can unlock major performance gains even without relying on cuttingedge manufacturing nodes. So in other words, work smarter not smaller. The most interesting part of the announcement is something Huawei calls logic folding. So to elaborate on this, let us think about a modern city.

[00:03:20]If the roads are badly designed, traffic slows down even if the cars are fast.

[00:03:26]And that was inspired by Bengaluru.

[00:03:28]Anyway, Hua Huawei, sorry, believes today's chips have a similar problem.

[00:03:33]Signals travel across long physical distances inside processors that creates delays. Logic folding attempts to redesign the internal architecture so those distances become shorter. shorter parts, lower latency, less power wasting, higher performance. Of course, the company claims this can improve both transistor density and overall computing efficiency. Now, here's why this kind of matters geopolitically. Huawei has spent years operating under US restrictions.

[00:04:04]We know this. It cannot freely access the world's most advanced manufacturing equipment. It cannot easily buy cuttingedge chips from American suppliers. and it faces barriers across the semiconductor supply chain. So the obvious question becomes is thous scaling a breakthrough or is it a workaround because when you can't access the best manufacturing tools the next best option is to become dramatically better at design and that's exactly what Huawei appears to be doing. Huawei says the first kion chips built entirely around logic folding will launch later this year. Even more interestingly, the company claims that by 2031, it can achieve transistor densities equivalent to 1.4 nanometer processes. That statement immediately caught the industry's attention because 1.4 nanometers represents territory that even the world's most advanced chip makers are still working toward. Some critiques, of course, argue this is more marketing than reality. Others point out that transistor density alone doesn't guarantee real world performance. And of course, questions remain around power consumption, thermal management, manufacturing yields, and scalability.

[00:05:16]Those, of course, are all very fair questions. Well, because semiconductor history is full of impressive presentations that never really tend to become successful products. This story isn't really so much about smartphones anymore. It's about AI. Every major AI breakthrough eventually hits a roadblock, which is compute. Training models requires enormous amounts of processing power. Inference requires enormous amounts of processing power as well. The world wants more AI, but the world also wants cheaper AI. If Huawei can genuinely extract more performance from mature manufacturing processes, changes the economics of computing. No, suddenly the conversation shifts from who has the smallest node to who has the smartest architecture. And that I believe is a much more open competition.

[00:06:06]Here of course is the front page tick.

[00:06:08]So what's fascinating here is that we are actually seeing the same pattern across the entire technology industry.

[00:06:14]OpenAI has changed software by focusing on models. Nvidia has changed computing by focusing on architecture. Enthropic well coding by focusing on orchestration. And now Huawei is suggesting that the future of chips may depend less on shrinking transistors and more on redesigning how they work together. Now whether Tao scaling succeeds or fails remains to be seen but the direction is clear. The error when semiconductor progress was defined solely by nanometers is well starting to transform. Please do let us know what are your thoughts in the comments below.

[00:06:52]This ladies and gentlemen is front page by the AIM network. Like, share, subscribe and always remember thinky.