America will make at least quarter of advanced chips in 2032, compared to China’s 2%
- Reference: 1715199366
- News link: https://www.theregister.co.uk/2024/05/08/us_china_sia/
- Source link:
Those figures are in a study out this month that primarily focused on the [1]impact of all that CHIPS Act funding and China's efforts to achieve processor independence. Both countries are seeking to bring advanced chip fabrication (defined by the report as smaller than 10nm nodes) to their own turf, and the Semiconductor Industry Association, or SIA, predicts the United States will by far be the biggest winner.
In its report
[2]PDF
titled "Emerging Resilience in the Semiconductor Supply Chain," the SIA predicts the state of the market in 2032 based on international trends and various sources.[3]
According to the trade association, it seems everything's coming up roses for US chip making, in part thanks to billions in CHIPS Act subsidies bankrolling factories and other efforts. As a reminder, this [4]funding is supposed to help America reduce its reliance on foreign semiconductor suppliers and grow its own domestic output.
[5]
[6]
In 2022, virtually all advanced chips were made in either Taiwan or South Korea, which produced 69 percent and 31 percent of sub-10nm dies respectively. In 2032, the report says, we should expect the US's share to increase to 28 percent, while China's will sit at just a couple of percent, while Europe and Japan projected share is six and five percent respectively. Taiwan is expected to bag 47 percent of sub-10nm capacity, down from 69 percent in 2022.
China is predicted to increase its share of the market for 10nm to 22nm chips from six to 19 percent, plus a slight bump in legacy silicon (28nm or higher) production from 33 percent to 37 percent. However, substantial decreases in DRAM and NAND fabrication in the Middle Kingdom will mean China's overall share of the semiconductor industry will decline from 24 percent to 21, while the US is projected to rise from 10 to 14 percent. See the chart below for the full breakdown.
[7]
Here's your usual reminder that most applications do not require cutting-edge chips, and your fridge, car, power supplies, and so on, don't need sub-10nm processors right now – though by 2032, we imagine things may well be different.
[8]
Projected per-country chip wafer output capacity ... Bear in mind these figures come from the SIA, using various sources. Source: Page 14 of above report
Additionally, the SIA predicts South Korea will see a dramatic drop in advanced chip production from 31 percent to just 9 percent. However, thanks to expansions in DRAM and NAND fabbing, South Korea's overall share of the silicon market will slightly increase from 17 to 19 percent.
Part of China's decline overall is doubtlessly due to how much capacity countries and regions are expected to add. While China saw a massive 365 percent increase in manufacturing capacity from 2012 to 2022, from 2022 to 2032 that growth figure will apparently be just 86 percent. That's below the predicted world average of 108 percent, and significantly behind Europe at 124 percent, South Korea at 129 percent, and the US at 203 percent.
Producing cutting-edge chips from scratch is hard
While going from zero to two percent in advanced semiconductors isn't nothing, it is perhaps much less than hoped for, especially since Chinese tech titans like Huawei and SMIC are already [9]making progress on producing 7nm chips. The report however seems to agree with Commerce Secretary Gina Raimondo, who insisted last year that [10]China couldn't make these kinds of processors at scale.
China's theorized inability to break into the advanced chip scene in a big way might be down to how much money Beijing's predicted to inject into its semiconductor industry, which the SIA says will be $156 billion from today to 2032. That's about equal with Europe's projected $154 billion but nowhere near South Korea's $300 billion budget, the US's $646 billion pile, and Taiwan's hoard of $716 billion.
[11]Where there's a will, there's Huawei to develop one's own chipmaking kit
[12]So what if China has 7nm chips now, there's no Huawei it can make them 'at scale'
[13]Teardown confirms Huawei's Pura 70 contains SMIC 7nm process node
[14]US commerce department yanks back Huawei export licenses
Even if China was spending more money on its chipmaking industry, it has problems in acquiring production tools such as the extreme ultraviolet (EUV) machines that ASML makes but is not permitted to sell to China. It's a significant issue for the Middle Kingdom as EUV tech is certainly one way to get below 10nm in volume, since older deep ultraviolet (DUV) tools achieve poorer yields at that node. Huawei is attempting to [15]make its own EUV machines, and it's hard to say if it'll work.
That said, Huawei already exceeded expectations by making 7nm chips at all, and now its challenge is just making them in large quantities and of a high enough quality.
Either way, legacy semiconductor fabrication will seemingly continue to be an important part of China's tech industry – and as we said, these older process nodes are still needed for a large part of the chip world – but the report notes that an excess of capacity will probably lead to lower prices, which will in turn make it harder to turn a profit.
On the other hand, it could also push international competitors [16]out of the legacy silicon business , paving the way for Chinese dominance in a part of the market that's still relevant - even if it's not at the cutting edge. ®
Get our [17]Tech Resources
[1] https://www.theregister.com/2024/03/18/national_strategy_microelectronics_research/
[2] https://www.semiconductors.org/wp-content/uploads/2024/05/Emerging-Resilience-in-the-Semiconductor-Supply-Chain_report.pdf
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2Zjv1@ajzz7xYKXEm3JkX6QAAAAk&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[4] https://www.theregister.com/2024/02/26/us_chips_funding/
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Zjv1@ajzz7xYKXEm3JkX6QAAAAk&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Zjv1@ajzz7xYKXEm3JkX6QAAAAk&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Zjv1@ajzz7xYKXEm3JkX6QAAAAk&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[8] https://regmedia.co.uk/2024/05/08/handout_sia_wafer_capacity.png
[9] https://www.theregister.com/2024/04/29/teardown_confirms_huaweis_pura_70/
[10] https://www.theregister.com/2023/09/19/huaweis_advanced_chip_sparks_us/
[11] https://www.theregister.com/2024/04/11/huawei_developing_chipmaking_equipment/
[12] https://www.theregister.com/2023/09/19/huaweis_advanced_chip_sparks_us/
[13] https://www.theregister.com/2024/04/29/teardown_confirms_huaweis_pura_70/
[14] https://www.theregister.com/2024/05/08/huawei_licenses_revoked/
[15] https://www.theregister.com/2024/04/11/huawei_developing_chipmaking_equipment/
[16] https://www.theregister.com/2024/01/17/chip_wars_china_oversupply/
[17] https://whitepapers.theregister.com/
Did anyone say what 'advanced' means?
In semiconductors 'more' has always equated to 'better' simply because our software technology has failed to keep up -- we're still pretty much stuck with second generation software which needs brute force to overcome its numerous limitations. This is why you now need what was a decent sized supercomputer of a decade or so ago to just look at web pages. If you point this out to anyone then they just recite the mantra "Moore's Law" as its supposed to somehow fix the problem rather than just papering over the cracks, the reality being that power consumption and resulting thermal stress are not good for the environment or the long term reliability of systems (which, according to marketing, we're supposed to throw away regularly regardless of whether they're usable or not). I don't expect any software breakthroughs but should they occur it will knock the bottom out of the high performance microcircuit market -- they're too expensive, too fragile and too power hungry.
Anyway, experience in other technologies has repeatedly shown us that while the fattest individual profits are to be made from making a relative handful of 'the latest' the real money's in commodity. Who cares is China's "only making 2% of advanced chips" if they're making 98% of everything else?
It looks the main US targets are Taiwan and South Korea.
Despite all the talk of China, it looks like projected US growth is primarily at the expense of Taiwan and South Korea. I'm not sure those two countries are just going to sit there and take it, or whether they will come out with a response of their own.
Projected US growth is almost entirely in two areas, DRAM and logic < 10nm. With all the subsidies being thrown at the industry, I won't be surprised to see massive over capacity and bankruptcies in the DRAM business making those projections obsolete.
As for US growth in logic < 10nm, that seems to depend heavily on Intel being successful in making chips for third parties, something they haven't demonstrated much luck with so far. Given the fading relevance of x86, I wouldn't be willing to bet money on Intel still being in the chip making business in 2032, let alone a power there. They may very well decide to go fabless to avoid the high capital expenditure and associated financial risk.
The winners in 10 years time may be the countries that avoided throwing massive subsidies at the industry.
I suppose it depends what you mean by "advanced".
A lot of the most interesting developments at the cutting edge are smaller, less complex chips cleverly networked together, witness the story earlier about the SpiNNaker2. If we saw a massive switch in the industry between now and 2032 (which I think *very* likely) then I think a lot of the countries like China could stand to make some inroads.