What about yields? Even if it is 600% faster but yields are 1000x worse, it doesn’t mean much commercially. But if this is an old chip that is easy to fab with equipment China has easy access to, then it is a big deal.
SMIC makes the Ascend 910c using its N+2 process—basically TSMC’s 7nm equivalent—and ASML’s TWINSCAN NXT:2000i DUV immersion lithography machines. But with yields stuck around 20% late last year, it's far from commercial-grade. GlobalFoundries and Intel have also wrestled with yield issues on this node, so it's no walk in the park. Expect the Chinese government to keep funding it for a few more years, but by then, the tech might be a bit dated—TSMC's been producing 7nm chips commercially since 2018.
The yield is much higher as indicating by Huawei using their 7nm process in mid range phones and tablets for export. Bottle neck now is likely in high bandwidth memory.
- The Kirin 9000s die used in Huawei's phones/tablets is 107 mm².
- SMIC's Ascend 910B has a die size of 665.61mm², which is more than 6x larger.
- The AI chiplet of Ascend 910 itself is already significantly larger than the whole Kirin 9000s die. There's no data on it from SMIC, but when TSMC produced it with equivalent N7+ process, it was 456.25mm² (>4x larger)
I don't know about the rumored 20% yield, but this difference in die-size alone would dramatically impact yield for the Ascend design. And that's just ONE factor.
SMIC could have achieved a 90% yield with Kirin 9000s and still be at 20% with Ascend 910...
That is for the old 910B which was fabbed on TSMC. The rumors I have seen is that 910C is a chiplet design.
Anyway. Since the company doesn't publish these numbers everything is secondary sources and dominated (at least in the English language media) by politics.
What you can do is look at objective indicators such as what gets sold (phones, computers) where (China, outside China). And prices and availability of Chinese cloud computing and AI products.
A big thing would be for the Huawei 910C to show up outside China.
> That is for the old 910B which was fabbed on TSMC. The rumors I have seen is that 910C is a chiplet design.
Yes.
As written, the (Virtuvian) AI processor CHIPLET of Ascend 910 with TSMC N7+ process has a size of 456.25mm², >4x larger than the whole Kirin SoC.
Rumor is that 910C was aimed to not redesign them significantly compared to the 910B iteration but double the amount of those AI chiplets in one package.
-
> What you can do is look at objective indicators such as what gets sold (phones, computers) where (China, outside China). And prices and availability of Chinese cloud computing and AI products.
The size of the chiplet in the comparable TSMC N7+ process is objectively known, the size of the Kirin 9000s SoC die is objectively known.
Those are good indicators that the commercial volume of the mass-produced die (Kirin) doesn't tell you much about the potential yield for the 4x larger AI-chiplet (Ascend 910 Virtuvian)
> And prices and availability of Chinese cloud computing and AI products.
We have no idea if the chips are just being heavily subsidized to deal with the low yields. If it isn’t free market, the prices they offer aren’t going to tell us much, and if it subsidized heavily it won’t be very sustainable.
To a point sure. Commercial viability doesn’t mean they need to match NVIDIA, they just need yields (and power consumption) that are just good enough. They are going to go for scale, so even power consumption is going to dictate how many nuclear plants they need to build.
Electricity prices being cheaper than in the US doesn’t necessarily mean electricity is actually cheaper. They subsidize a lot of it since utilities are state run and rates are set by politicians just as much as they are set based on costs. Farmers still freeze in the winter when they are forced to switch from in-home coal to electric heating (natural gas isn’t viable in rural China, and propane tanks are often banned for purposes other than cooking), electricity is cheap but not cheap enough compared to what they earn.
Your critique contains multiple factual inaccuracies:
1. Price Formation: China's electricity pricing isn't politically manipulated. Since 2021, 60.8% of electricity has been traded in competitive markets (NDRC 2022), with industrial users paying 20-40% more than residents.
2. Rural Heating: 12 million rural households received $3.5B in 2023 heating subsidies (MOF), reducing coal-based PM2.5 by 54% since 2015 (MEE). Freeze incidents decreased 78% post-2020 grid upgrades (State Grid Corporation).
3. Energy Access: Over 98% villages now have LPG access via 53,000 licensed stations (MEM), with 300m CNG cylinders in rural circulation - 4x more than in 2017.
4. Affordability: Rural electricity costs average 5.9% of income vs 8.7% in US farm households (OECD 2023). China's residential rates remain 30% below commercial tariffs to protect vulnerable groups.
While transitional challenges existed during 2017-19 coal-to-clean shift, WHO-certified data shows rural respiratory hospitalizations dropped 22% since 2020. The "cheap energy=state control" narrative oversimplifies complex market structures evolving since 2015 reforms.
It is national security issue for China, it doesn't need to be commercially competitive. Like Ford/GM has long lost their competitiveness against Japanese car makers, but the government will keep them around as long as they can.
> This suggests that Huawei's AI processor's capabilities are advancing rapidly, despite sanctions by the U.S. government and the lack of access to leading-edge process technologies of TSMC.
So chipping away at Taiwan's "silicon shield" (the defence strategy that an attack on Taiwan would deny China of the chips it needs itself).
Soon China can have a security situation with Taiwan (doesn't really matter if it actually invades, or just has a hotting-up conflict with rocket exchange) to choke off supplies of high-end chips to the West whilst domestic alternatives ramp up?
China's territorial claims over Taiwan starts from 1949 since the civil war ends. [1]
The semiconductor doesn't play significant factor here. It only becomes hot topics in past few years because of TSMC's lead on its process, but really doesn't have any direct relationships.
In other words, current situation is a result of U.S. "Asia-Pacific Rebalance" strategy and military deployment over the first island chain [2] that China wanted the island regardless of TSMC.
That's definitely reading the situation totally wrong.
This so called arms race is only in western minds. It would be stupid to not think about this possibility and its consequences of course but this is the most illogical option by China.
The best option would be to advance their own semiconductor tech beyond TSMC and make it irrelevant, which so far is the course of action they are taking (alas only because they are forced by US).
Expanding on why capturing Taiwan is stupid: TSMC already has fabs and people with knowhow outside Taiwan. The fab assets are not as strategicly valuable as their knowhow. Plus, even if you acquire the knowhow, there's the supply chain: ASML, others, ASML's supply chain.
I think US might force ASML and TSMC to move to US in the coming years even stronger but more than that this arms race isn't getting hot.
There's no way the EU will allow US to force ASML to move to the US, especially given the burning of the bridges of decades of diplomacy by the incumbent president.
If the trade war continues and escalates, which form our current perspective it looks like it will, the EU will not lose one of its greatest assets in that space.
The idea of Trump that the US is powerful enough to strong arm everyone to his will is mistaken, an alienated US without its allies will suffer, as will the rest of the world..
Totally agree. It is really in both sides interest to make TSMC irrelevant in Taiwan and avoid WW3.
So we will have a cold war, trade war and saber rattling until that time.
Then when the chips don't matter someday Taiwan will become old news to western minds like Iraq is today. When China takes Taiwan it will barely be a western news story because right now will feel so long in the past.
The CCP doesn't want Taiwan for the chips, they want it because having a successful democracy of ethnic Chinese off the shore means there are viable alternatives to communist rule, by and for Chinese people.
In the same way that Russians want to "take over" Ukraine or Argentinians want to "take over" The Falklands or Americans want to "take over" Greenland?
Territorial expansion is always popular with the masses before the shooting starts.
It is more nationalistic than political; in fact, many unpopular politicians start wars in order to get a sense of national unity and tamp down unrest at home.
I’m not sure Greenland should be lumped in with the others considering the very different (lack of) history, and polling does not support its popularity.
> A strong majority of Americans, 66%, are opposed to U.S. territorial expansion accomplished “by invading other countries and territories,” according to a recent poll from More in Common.
> In contrast, just 13% support the U.S. launching an invasion to acquire territory, and 16% were not sure
why you mixed [Argentinians want to "take over" The Falklands] between [Russians want to "take over" Ukraine] and [Americans want to "take over" Greenland]
> He stressed that China will not blindly follow trends or engage in unrestrained international competition. With a robust governance and regulatory system in place, China is confident in its ability to manage and utilize AI technology effectively
Taiwan has been touting it's "Silicon Shield" defence for years and trying to tie China's self-interest with leaving Taiwan alone for now.
But the Silicon Shield is tumbling.
On the one side the US (the carrots of the CHIPS act under Biden, the stick of sanctions on Taiwan under Trump) is trying get sovereign control of chip production.
And on the other side China is ramping up getting sovereign control of it's own top-end production.
If both stay on course then China will have one less big reason not to invade, and the US will have one big reason less to defend.
Of course if China wins the 'race' then it could be to it's advantage to have a rocket exchange or invasion with Taiwan to choke of the West's supply of chips before the West has brought chip production home?
I don't really think it will choke the supply, but only reduce in capacity in short term, for example TSMC Arizona[1] took 4 years to be in produciton. Intel's process is still acceptable and better than China.
Also semiconductor is very complicated manufacturing process not only invovles Taiwan but vendors all over the world, like the light source of EUV is provided by Cymer based in U.S. I don't think China taking over Taiwan would mean they can transfer whatever is there into something productive in any short time as because of supply chain problem. Note that many SMIC's (basically China's TSMC) leadership are actually from TSMC already[2].
This assumes Taiwan's involvement in Huawei processors are 0. Lots of things don't make the news. There are way more companies involved than just TSMC.
At the risk of whataboutism, this is looking a little differnet now on the world stage with the US-on-Denmark threats. A second source for AI hw from China could turn out to be a good card to have even for western countries.
The US also has comparable hardware, and it had it for a while: Intel Gaudi. Similar perf characteristics, works very well for inference, costs substantially less than H100. The entire field seems to be drowning in money, so no one gives a shit.
The key piece of information is missing though - at what price and energy cost. For the first sample it sounds exciting, but... they'll have to compete with the cost of smuggling Nvidia into the country.
Agree on price but what you are suggesting on energy is likely not true.
More energy is likely more power (if you decide that the performance matters). More power means you cant network enough of these chips together at the scales needed for training. And that means they havent really solved the problem (although likely they are progressing nicely towards the solution).
Basically they made training much more efficient and developed a better and cheaper product. They passed the stage of "AI in China is basically impossible" to "It's just a matter of time and money until they take the lead".
It wouldn't get killed, but it may end up like Baikal, Elbrus and others - kept around just in case, but not really used instead of bought/smuggled chips.
It always matters to some degree but you can spend much much more to get an edge for national security. 10x maybe 100x compared to the free market but not infinite.
