Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/how-chinas-chip-production-boomed-in-2023-despite-sanctions.19341/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021770
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

How China's chip production boomed in 2023 despite sanctions

Daniel Nenni

Admin
Staff member

Huawei played a central role in the US-China tech war​


235f3620-a00f-11ee-8bf6-bef5e3710522

REUTERS/Aly Song
It's been an interesting few years for Huawei. After the Chinese giant's initial struggle with the US trade sanctions, it would end up with a surprise mobile resurgence featuring homegrown processors — ones that are just two generations behind the competition. Not only that, the Chinese government has since allocated billions of dollars to boost its silicon industry, so much that Huawei is already working towards a self-sufficient chip network. It's as if former President Donald Trump's earlier attempts to starve Huawei of vital inputs eventually accelerated China's semiconductor development.

Trump's first strike on Huawei was the declaration of a national emergency in May 2019, which saw the Commerce Department add the company to its Entity List, citing surveillance concerns and links to the Chinese state security. As such, Google could no longer provide Android support to Huawei, thus causing the Mate 30 series and later models to miss out on Google apps (they would eventually adopt Huawei's Android replacement, HarmonyOS, two years later).

In November 2019, the FCC banned carriers from buying Huawei and ZTE networking gear with government subsidies.The following March, Trump signed a bill that would reimburse the replacement of Chinese gear — even if it meant spending an estimated $1.8 billion. Huawei attempted to sue the FCC over these restrictions, but the court sided with the regulator.

The tech war heated up rapidly in May 2020, when the US further restricted Huawei's access to American equipment and software. This meant Taiwan Semiconductor Manufacturing Company (TSMC), the world's leading fab, would have to stop producing HiSilicon chips for Huawei — its then second-largest customer, after Apple. Likewise, Samsung and SK Hynix had to stop selling chips to the Chinese brand by the September 15, 2020 deadline. As Bloomberg's teardown of the latest Huawei smartphones revealed, the company didn't have a problem stockpiling these Korean memory chips.

42613020-9f36-11ee-ade6-cd7761101753


For processors, Huawei had no choice but to rely more on local chip makers, namely Semiconductor Manufacturing International Corporation (SMIC) and Shanghai IC R&D Center. That meant a significant downgrade, though: SMIC had just started mass-producing 14nm chips for Huawei then, whereas TSMC reached 5nm later that year and supplied Kirin 9000 processors for Huawei's Mate 40. That would be the final "high-end" Kirin chip, Huawei's mobile boss Richard Yu said at the time.

Qualcomm was eventually allowed to supply 4G chips to Huawei as of November 2020, but that's four G, and market share figures don't lie. The once-leading brand in China dropped to just 16 percent locally in January 2021 (and then down to a mere 6 percent in Q2 2022), as noted by Counterpoint. Huawei's global market share has been negligible since 2021. According to both Counterpoint and Statista, though, since Huawei sold the Honor brand in November 2020, the spin-off has been able to claim one of the top China quarterly chart positions all this time.

China's chip investment finally paid off when SMIC made a 7nm breakthrough in August 2022 — a leap from 14nm in just two years — faster than it took TSMC or Samsung, according to TechInsights. What's more, this achievement was apparently done without using the most advanced lithography equipment, which were largely exclusive to the likes of ASML and Nikon. It wasn't until earlier this year that the US convinced the Netherlands and Japan to restrict China’s access to advanced chipmaking machinery.

As Bloomberg would later find out in a lengthy investigation, this might have been the fruition of a Shenzhen city government investment fund from 2019 that helped Huawei build "a self-sufficient chip network." Through a network of enterprises, Huawei could stealthily gain access to lithography tech while exchanging experts to work on each others' turfs, without raising any flags. Huawei apparently even managed to hire several former ASML employees, which was likely key to reaching the 7nm node process for its latest processor (the 5G-capable HiSilicon Kirin 9000S, fabricated by SMIC). Benchmarks indicate that this chip's performance is on par with Qualcomm's Snapdragon 888 from late 2020, thus suggesting that it's around two generations behind the leading competition.

Huawei then took a rather unusual approach to launch its Kirin 9000S smartphones at the beginning of September this year. Without any launch event or teaser, the company simply announced on Weibo that the Mate 60 and Mate 60 Pro were immediately available. This surprise stunt coincided with the US Commerce Secretary Gina Raimondo's visit to China, which led many to believe that Huawei received special orders from certain authorities to hastily launch these 5G devices ahead of schedule. This was quickly followed by the China's announcement of a $40 billion fund to further boost its chip industry, as well as the launch of two more phones, the Mate 60 Pro+ and the Mate X5 foldable, a week later.

People look at Huawei Mate 60 series smartphones displayed at a Huawei flagship store in Beijing, China September 25, 2023. REUTERS/Florence Lo

People look at Huawei Mate 60 series smartphones displayed at a Huawei flagship store in Beijing, China September 25, 2023. REUTERS/Florence Lo (REUTERS / Reuters)

While this may seem a temporary win for China, the country actually saw 10,900 chip-related companies close down in 2023 (as of December 11) — a staggering 90-percent year-on-year increase, which is a sign of a bad economy, according to TMTPost. On the flip side, 65,700 new chip-related companies registered in the same period, which is a 9.5 percent increase year-on-year. The report added that the China-made RAM chips and processors on Huawei's Mate 60 series are an indication of the growing reliance on the local supply chain, which will continue to drive the long-term development of the Chinese semiconductor industry.

As much as the US government wants to limit China's access to high-end tech, the truth is western companies still want to tap into the big market in the east. NVIDIA is a prime example, as it's still in talks with the authorities on the specifications of AI chips that it can sell to China, without breaching US export rules. "What we cannot allow them to ship is the most sophisticated, highest-processing power AI chips, which would enable China to train their frontier models," Raimondo told Reuters. Of course, failing that, China may eventually come up with an AI chip that's just as impressive, if not more — like its recent claim of a light-based chip that is apparently 3,000 times faster than NVIDIA's A100.

The US-China tech war isn't just limited to chips, either. The Biden administration is proposing to cut tax credits on electric vehicles that contain Chinese components — especially batteries, as an attempt to wean local car brands off Chinese components. The trade-off here is always the cost savings (as is the idea behind Ford and CATL's Michigan battery plant), as well as the US market missing out on potential breakthroughs on power density or output, namely the upcoming 150kWh battery demoed in Chinese EV manufacturer Nio's ET7, which reached a range of around 650 miles. Who knows, maybe someday Huawei may want to sell its Aito or Luxeed electric cars in the US, too — if it's allowed to enter at all.

 
It's an exceptionally poor insight.

They provide zero info on what's actually booming in China. It has little to do with smartphone application processors, as the author strongly pushes.

Smartphone application processors are just the most visible tip of the iceberg.
 
It's an exceptionally poor insight.​

They provide zero info on what's actually booming in China. It has little to do with smartphone application processors, as the author strongly pushes.

Smartphone application processors are just the most visible tip of the iceberg.

Just putting this on the table https://www.digitimes.com/news/a20231225PD200.html?mod=2

In a few years, SMIC alone may get within a striking distance of TSMC's legacy offerings by capacity, which is insane, and that wasn't on anybody's radar just a year ago.

SMIC bet yolo big on 200mm by the time almost every scale player left it. COVID chip crunch rewarded them hugely, and now they double down.

Why it's near impossible to compete with mainlanders on 200mm?

1. No new equipment made in more than a decade. No big player in the West will ever attempt building a brand new fab with second hand equipment, because they can't even imagine that such a thing is possible.
2. Obviously, everything semiconductor is heavily subsidised in China
3. Mainlanders became experts at vacuuming second hand equipment off the market
4. Western institutional investors/creditors don't get it at all. No bank will underwrite a credit for such an attempt
5. The ecosystem effect is not strong on old nodes
6. Point 5 is synergistic with 3rd world design outsourcing shops

Remember, starting off as a volume oriented bottom feeder was how TSMC first got its dominance. Now, Chinese are copying that because nobody in Taiwan had balls for "scorched earth" strategy, to strategically deny close followers profits on legacy nodes.

When TSMC is too preoccupied with the high profitability bleeding edge, their mass market high volume service is not so extremely competitive as it was 10 years ago.




The worst held secret in the industry: Net sales value of even very old node services has been growing for years! This was not only COVID

This is the most natural consequence of the world just consuming more electronics, and using more of it in lower value, disposable goods.


Decades old fab lines are regaining double digit profitability. And for the speciality service like mixed signal, HV, power, analog. and BiCMOS, the capacity is even tighter.
 
Last edited:
The worst held secret in the industry: Net sales value of even very old node services has been growing for years! This was not only COVID

This is the most natural consequence of the world just consuming more electronics, and using more of it in lower value, disposable goods.


Decades old fab lines are regaining double digit profitability. And for the speciality service like mixed signal, HV, power, analog. and BiCMOS, the capacity is even tighter.
Was this pasted in from the techovedas.com link which wants to install a browser extension to view their "insights"? I'm sure some 200 mm fully depreciated lines are profitable, but I doubt there will be significant growth in 200 mm silicon foundry revenue.
 
The thing the article seems to miss is that more recent “legacy nodes” can be run just as inexpensively per unit area (or even more inexpensively per transistor) by fully depreciated 300mm lines.
 
The thing the article seems to miss is that more recent “legacy nodes” can be run just as inexpensively per unit area (or even more inexpensively per transistor) by fully depreciated 300mm lines.

Not quite, depends on what you consider recent. 130nm on 300mm for example is rare, and was only done by few European fabs. Some 90nm fabs are 200mm, and most of niche process ones.

So, a lot of HV, BiCMOS, power, and analog capacity will be on 200mm, until new fabs for that will be built, or somebody will move proprietary processes onto an existing 300mm fab.

But a much more likelier is that nobody will "sacrifice" an existing 300mm fab still doing relatively profitable wafers for such a risky experiment. Oldest 300mm capacity is still very well loaded because low-end IC shipments are obviously growing, and will be growing long term.
 
But a much more likelier is that nobody will "sacrifice" an existing 300mm fab still doing relatively profitable wafers for such a risky experiment. Oldest 300mm capacity is still very well loaded because low-end IC shipments are obviously growing, and will be growing long term.
I don’t quite see it that way. TSMC, UMC and now even Intel are pretty good about expanding their product offering to include niche processes on their legacy nodes. TSMC offers an HV variant with BCD down to 40nm in former leading edge 300mm fabs. I think both TSMC and Intel are working on Automotive processes down to 14nm. The goal is to fill older fully depreciated 300nm fabs with new types of designs, by adding specialty variants to former core digital processes.
 
Last edited:
I don’t quite see it that way. TSMC, UMC and now even Intel are pretty good about expanding their product offering include niche processes on their legacy nodes. TSMC offers an HV variant with BCD down to 40nm in former leading edge 300mm fabs. I think both TSMC and Intel are working on Automotive processes down to 14nm. The goal is to fill older fully depreciated 300nm fabs with new types of designs, by adding specialty variants to former core digital processes.
How well does it go for them?
 
How well does it go for them?
Seems to be helping TSMC keep the legacy 300mm fabs full.

TSMC provides foundry's most comprehensive and competitive Bipolar-CMOS-DMOS (BCD) Power Management process technologies and is also the first foundry to adopt 300mm wafer production for the BCD Power Management process.

TSMC BCD Power Management process features higher integration, smaller footprint, lower power consumption, covering nodes from 0.6µm to 40nm. Customers chips produced by the process provide more stable and efficient power supplies that consume less energy, ideal for applications including consumer electronics, communication devices, and computers.

But it would be great if you had more specific info.
 
Seems to be helping TSMC keep the legacy 300mm fabs full.

TSMC provides foundry's most comprehensive and competitive Bipolar-CMOS-DMOS (BCD) Power Management process technologies and is also the first foundry to adopt 300mm wafer production for the BCD Power Management process.

TSMC BCD Power Management process features higher integration, smaller footprint, lower power consumption, covering nodes from 0.6µm to 40nm. Customers chips produced by the process provide more stable and efficient power supplies that consume less energy, ideal for applications including consumer electronics, communication devices, and computers.

But it would be great if you had more specific info.
I heard it's quite pricey for what could've been done at 2nd tier fabs, unless your volumes are in tens of thousands
 
I heard it's quite pricey for what could've been done at 2nd tier fabs, unless your volumes are in tens of thousands
That’s kind of what I thought. But if you want volume, then they have the 300mm to deliver.
 
Last edited:
Back
Top