Trade War with China

(cont)
Plus, the trade issues have rattled China. “China is kind of backing off that 2025 goal. Part of that might be due to the trade issues. This ‘Made in China’ campaign and these aggressive targets really put foreign governments on alert with China,” IC Insights’ McClean said.

Still to be seen, however, is whether China will continue to pursue its aggressive goals. Amid the geopolitical issues, demand for chips is slowing in China. This is impacting a growing number of vendors, such as Apple, Intel, Texas Instruments and TSMC.

Citing a slowdown in memory and other factors, chip sales in China are expected to grow by only 3% in 2019, compared to 21% growth in 2018, according to IC Insights. In total, the worldwide chip market will grow by 2% this year, the firm noted.

There are other ominous signs. Pure-play foundry sales in China reached $10.69 billion in 2018, up 41% over 2017, according to IC Insights. In 2019, though, foundry sales in China are expected to slow to about 10%. “Some of the Chinese fabless companies are doing well. HiSilicon is one example. They are basically the main supplier for Huawei. And Huawei is doing well in the smartphone business,” McClean said. “In the foundry space, however, the growth in China is a little misleading. Especially in the first half of 2018, a lot of that growth was driven by cryptocurrency designs. Bitcoin took a tremendous hit last year. That’s taken away a lot of the steam in the China market.”

Still, China continues to build up foundry capacity. “Out of the 30 fabs and lines starting construction in 2018 and beyond, 13 are for foundry,” said Christian Dieseldorff, an analyst at SEMI. “As of the end of 2018, our data showed an installed foundry capacity of 1.3 million wpm in 200mm equivalents in China. Worldwide is about 7 million wpm. This includes IDM foundry capacity.”

The figures include both 200mm and 300mm fabs. In total, there are six 200mm facilities in the works in China, according to Dieseldorff.

From the equipment perspective, meanwhile, it’s a mixed picture. Last year, SEMI reduced its projections for fab equipment spending in China for 2019, from $17 billion to roughly $12 billion. So, in 2019, fab equipment spending in China is expected to reach $11.96 billion, down 2% over 2018, according to SEMI.


“We revised our forecast for 2019 and we still expect pretty healthy growth for China in 2020. We scaled down some of the capacity ramp-up plans in the next two years. But still, there are a lot of memory projects in China, such as Samsung, SK Hynix and even Intel. The investment in China is still strong,” SEMI’s Tseng said. “SK Hynix will invest more this year in China compared to Samsung. This is because they have a new fab in Wuxi.

“The reason why we scaled down the forecast is the China part. We don’t think they will ramp up as fast as they announced. There is a potential that the DRAM project in Fujian will essentially stop. The Hefei project has been very low profile. Overall, the progress is slower than what they announced. It’s also slower than what we expected,” Tseng said. The SEMI analyst is referring to DRAM hopeful JHICC, which is based in Fujian. Innotron, another China DRAM vendor, is located in Hefei.

Others see similar patterns. “The domestic semiconductor companies have been spending quite a bit. And the business was up at all major equipment suppliers in 2018,” said Risto Puhakka, president of VLSI Research. “Now, the indications are that it’s going to slow down like the rest of the industry. In China, you have two different groups. One, you have the domestic Chinese companies. Then, you have the multinationals. The multinationals, in a very large way, won’t be increasing their capital spending. Then, you have the domestic companies. Their projects are not finished by any measure. They will spend some, but it looks like they obtained a fair amount of equipment last year, roughly $5 billion. They have plenty of tools for now.”

Most see flat equipment growth for 2019. “I don’t expect the overall wafer fab equipment market from China to change significantly between ’18 and ’19,” said Oreste Donzella, senior vice president and chief marketing officer at KLA. “The mix is different. We see more foundry and less memory. We see more foreign investment and less local.”

Regarding the IC market as a whole, Donzella said: “For memory, what we are seeing right now is a definite slowdown. I see a CapEx decrease in ’19 for DRAM after an incredible year. In NAND, it will be modestly down in ’19. We believe foundry will go up. The question is how much will foundry go up.”

Multinational foundry landscape
There are several multinational foundry vendors in China. Last year, TSMC had a big year, thanks to a boom cycle in cryptocurrency. TSMC makes chips for the cryptocurrency systems firms in China and elsewhere.

Now, the company is suffering from the bust in Bitcoin. In its fourth-quarter results, TSMC posted mixed results with a weak outlook due to a slowdown in smartphones and cryptocurrency.

It’s unclear if this will impact TSMC’s fab plans in China. For years, the company has operated a 200mm fab in Shanghai.

Last year, TSMC began ramping up 16nm finFETs in a new fab in Nanjing. The 300mm fab is producing 10,000 wpm with plans to move to 20,000 wpm by year’s end.

Fig. 2: Major IC manufacturers in China Source: IC Insights

FinFETs represent a big leap for China. For years, chip production within China has been limited to traditional planar transistors at 28nm and above. FinFETs are used at 16nm/14nm and beyond. In finFETs, the control of the current is accomplished by implementing a gate on each of the three sides of a fin.

TSMC’s Nanjing fab is in the first phase with three other phases in the works. Originally, the company planned to move 7nm into a proposed second fab in Nanjing at some point, according to sources. Now, TSMC is rethinking its plans, targeting 16nm or 12nm for the second phase, sources added.

Not all processes are at the leading edge. For example, UMC has been producing chips in a 200mm fab in Suzhou for several years. UMC plans to expand this fab amid a worldwide shortage of 200mm capacity.

“That’s still on schedule,” said Jason Wang, co-CEO of UMC, in a recent conference call. “And we still have confidence in terms of our 8-inch outlook.”

Meanwhile, in 2017, UMC moved into production in a 300mm fab venture in Xiamen. The fab is ramping up 40nm and 28nm.

Demand for 40nm is steady, but 28nm is engulfed in an oversupply mode for all foundry vendors. “28nm will remain flat. 28nm may experience (an) overcapacity situation for the next years,” Wang said.

Meanwhile, GlobalFoundries is building a 300mm fab in Chengdu. Previously, the goal was to develop 180nm/130nm processes in the fab. Last year, though, the company changed its plans, targeting the fab for its 22nm FD-SOI technology, dubbed FDX. It has not provided a timeline for the completion of the fab.

GlobalFoundries also is ramping up 22nm FD-SOI in a German fab. Nonetheless, the company sees growing interest for FD-SOI in China. “FDX technology is particularly well-suited to the China market, and we continue to see strong potential for its up-take in segments, such as 5G, IoT and edge computing,” said Tom Caulfield, chief executive of GlobalFoundries.

In addition, contract manufacturing giant Foxconn is in talks to build a 300mm fab in Zhuhai, according to reports. The fab, a joint venture between Foxconn and Sharp, would be used for both captive and foundry purposes. Foxconn has yet to make an official announcement.

(cont)
Domestic foundry vendors
China also has more than a half-dozen domestic foundry vendors. SMIC is the largest one. The Huahong Group is another foundry player in China. Other vendors include ASMC and CSMC.

2019 is a major test for SMIC, the most advanced domestic foundry in China. SMIC’s most advanced technology is 28nm, although it has been struggling with its yields in this arena.

In comparison, TSMC introduced 28nm a decade ago. The foundry currently is ramping up 7nm, with 5nm and 3nm in R&D.

GlobalFoundries, Samsung and UMC offer 28nm, and they are ramping up 14nm. For now, though, GlobalFoundries and UMC have stopped development beyond 14nm and/or 12nm.

Samsung is ramping up 7nm and other nodes, with 3nm in R&D. And Intel continues to ramp up 14nm, with 10nm in the works. (Intel’s 10nm is roughly equivalent to 7nm at commercial foundries.)

China is behind in process technology. So in 2015, SMIC, Huawei, Imec and Qualcomm formed a joint R&D chip technology venture in China with plans to develop 14nm finFETs by 2020.

With the technology, SMIC will shortly enter the 14nm finFET market. “We are walking towards risk production in the second half of (2019),” said Liang Mong Song, co-CEO of SMIC, in a recent conference call.

Still to be seen, however, is whether SMIC can produce finFETs in volume with good yields. “With more knowledge and process module recipes available on finFETs, it is only a matter of time when SMIC will get it developed,” said Samuel Wang, an analyst with Gartner. “Time to market is critical. If early, they can be successful. If it’s too late, then it’s more challenging.”

If SMIC can produce 14nm finFETs in volume, it would put the company on par with several foreign rivals. Near-term, though, SMIC’s 14nm capacity will remain small.

SMIC isn’t standing still. With funding from the China government, it plans to develop more advanced processes in a new 300mm fab. “SMIC is getting $10 billion to build capacity for 14nm, 10nm and 7nm. They will have capacity for 70,000 wafers a month by Q4 in 2021,” said Handel Jones, chief executive of International Business Strategies (IBS). “The building is huge. They have bought some equipment, but nothing significant yet.”

SMIC faces some challenges. 14nm finFETs are difficult to develop, but 7nm represents a quantum leap. At each node, the technical challenges and costs escalate, and there are fewer customers that can afford these nodes due to soaring IC design costs. The processes are also becoming more complex, making it difficult to find the killer defects that impact yield. In addition, patterning remains challenging. And at some point during 7nm, some hope to insert extreme ultraviolet (EUV) lithography, a move that brings more risk into the equation. That’s just the tip of the iceberg.

Besides SMIC, meanwhile, the Huahong Group is also a player in China. HHGrace and Hua Hong Semiconductor, two foundry vendors within this group, provide 200mm foundry services. Hua Hong also is building a new 300mm fab for mature processes in Wuxi.

Shanghai Huali, another member of the Huahong Group, has a new 300mm fab in the works with plans to ramp up 28nm. At some point, the company is targeting 14nm.

One of the newer foundry players is CanSemi Technology, which is building a 300mm fab in Guangzhou. At one time, CanSemi was led by Richard Chang, formerly president of SMIC.

CanSemi’s future is unclear as its leader has departed. “CanSemi seems to be in existence,” Gartner’s Wang said. “Richard Chang has left to start a new company.”

Put in perspective, China’s IC industry has been dynamic with sizzling growth. Now, though, China faces an economic slowdown, not to mention a trade war, so perhaps the IC party will be postponed.

List of semiconductor making machinery

SMEE (lithography)
AMEC (etch, deposition)
Piotech (deposition)
ACM and Kingsemi (surface conditioning)
Tianjin Hwatsing (flattening equipment CMP)
Raintree (inspection)
Wuhan Jingce Electronic (testing equipment)
NAURA (thermal processing, etch, deposition, materials, cleaning tool)
PNC Process Systems (cleaning tool)
CETC (ion implanters, inspection, materials, laser, CMP, lithography, deposition, cleaning, packaging and testing)
ASM Pacific, Hangzhou Changchuan Technology (Chip assembly, testing and packaging equipment)
Zhejiang Jingsheng (semiconductor materials)
HGTECH (Laser equipment)
ACKEN Optoelectronics (Laser equipment, semiconductor materials)
Konfoong Materials (materials)

Max Demian said:

This depends on the industrial sector. For companies like Huawei, ZTE and the like there is little recourse but outsource the manufacturing of critical IC components. Some sources claim that over 90% of ICs consumed are manufactured by foreign chipmakers.

An effort is underway to change the balance, but thus far the results have been underwhelming. They need to step up investments big time.

Click to expand...

Huawei doesn't outsource, but they're in the minority. Most companies from food to electronics assemble outsource to many different actors.

However there's ample evidence that this is turning around. Cost is increasingly not the only variable due to rising world tension.

Maybe more importantly, the state takes the issue of maintaining industrial primacy seriously. They're experimenting with ways to circumnavigate inherent cost, space and population limitations, such as AI, 5G based communication, remote controlled factories and unmanned transports.

I think with the leisurely pace nearly every other country is advancing industrial tech, China can tread slowly and make sure it gives time to it's companies to properly adapt while making sure new technology is well functioning and not rushed. After all, China already holds major if not dominant market share in all important goods, with outright monopoly on critical extraction technologies. So it is more important to make sure the new industry gets a good roll out.

If the state subsidizes too much, it could risk causing a great leap forward type disaster where companies overspend using state funds. But if it provides too little perks, companies will keep outsourcing and seeking short term profit rather than helping the state build up the industry of 2025.

That is a good article. Chinese financial interests not only tried to purchase Micron but also other companies like Cypress Semiconductor.
A lot of foreign companies are currently building semiconductor fabs in China because of the generous funding of local governments as part of the Chinese government's push towards increased production of such products in China. While the article talks about the lack of organic drive for the growth of these sectors it makes sense for China to make these massive investments. A lot of them might fail but this is the fastest way to move production into China.
Eventually the Chinese government needs to fund more dedicated research programs in these areas however. I think for example the Chinese government should have made a specific initiative around memory products (NAND, DRAM, MRAM, etc). This could have been done decades ago.

Also another issue with the current program is that it is highly capital intensive since the most expensive item, the chip manufacturing tools, are imported. This means a massive drain in foreign capital reserves.

Biscuits said:

Huawei doesn't outsource, but they're in the minority. Most companies from food to electronics assemble outsource to many different actors.
...
.

Click to expand...

This part of your statement is false. Huawei doesn't manufacture its chips. It outsources that to foundries such as TSMC and others. Moreover, Huawei through Hisilicon purchases licences for CPU designs from ARM.

Max Demian said:

This part of your statement is false. Huawei doesn't manufacture its chips. It outsources that to foundries such as TSMC and others. Moreover, Huawei through Hisilicon purchases licences for CPU designs from ARM.

Click to expand...

Exactly. I have read that a few times that Huawei mainly design their CPU based on ARM licenses (Hisilicon) and mainly TSMC manufactures them for Huawei. Very dangerous. Taiwan will do a blanket ban with one visit of American diplomat and Huawei, ZTE, BKK electronics, Hikvision, Dahua and many more are done for it. I don't understand why Chinese politicians/rulers have made their country / industry be so dependent on the West and their vassalages. They know the current hegemon is not so friendly to them? A very grave mistake and some heads should be rolling. I am death serious. They have created with all these dependence so many headways for American blackmail. When Xi Jinping gives them here a concession and some there, they will be emboldened and next time they will demand more extreme concessions.

Appix said:

Exactly. I have read that a few times that Huawei mainly design their CPU based on ARM licenses (Hisilicon) and mainly TSMC manufactures them for Huawei. Very dangerous. Taiwan will do a blanket ban with one visit of American diplomat and Huawei, ZTE, BKK electronics, Hikvision, Dahua and many more are done for it. I don't understand why Chinese politicians/rulers have made their country / industry be so dependent on the West and their vassalages. They know the current hegemon is not so friendly to them? A very grave mistake and some heads should be rolling. I am death serious. They have created with all these dependence so many headways for American blackmail. When Xi Jinping gives them here a concession and some there, they will be emboldened and next time they will demand more extreme concessions.

Click to expand...

Well the west has huge lead when it come to semiconductor It is push and pull factor until recently there is no demand for semiconductor The main user of semiconductor are smartphone, PC, server industry etc which does not even exist in China 15 to 20 years ago So why build semiconductor industry when there is no demand or very small demand

Another thing is with 4G Qualcomm and other monopolize the technology you have to use their chip and their technology That is why this 5G is so important for China because now Huawei own most of the technology.Even if Huawei is banned most Telco will still use their technology
It provide an opening for China just like conventional engine China can never beat the European or Japanese because the technology developed over hundred of years But with electric car China is on the same starting point just like everybody So chip is the the same suddenly we have AI now the level playing field is even

The other thing is legacy the west and japan has huge lead starting with Fairchild industry and Texas instrument who built the first semiconductor for Apollo program. Steve Jobs, Wozniak, Noyce all work there. so China is 50 yeas behind add to that Cultural revolution and other political upheaval
Heck University doe not even it door until early 80's

plawolf said:

Surely Trump would have encountered MOUs in the business world before? Or did he just bully everyone into called it something else all his life?

Anyways, it looks increasingly likely a trade deal will be struck, and part of that will be a massive spending spree by China.

While on the face of it, that might looks like paying up ‘protection money’ to US extortion, and there is indeed a great deal of truth to that, such purchase will not be without condition either.

China was already buying all it wants/needs from the US on the open market; but there just wasn’t that much China wanted to buy, hence the massive trade deficit.

America only has so much soya beans and natural gas, if America wants to take that 1tn+ order, they need to put things on the table that China wants and needs.

That may mean easing of export restrictions, or else China can just give them a 1tn shopping list, and if America wants to reject most of that, then that’s America not sell instead of China not buying.

In addition, I see this as a way for China to de-couple itself from the USD by paying for those purchase with some of its US T Bills reserves.

Such a one-off shopping spree would be a perfect way to offload a massive amount to T Bills without freaking out the market into a selling frenzy and massively devalue the T Bills China still hold.

I think this will be remembered as the beginning of the end of the USD dominance.

Click to expand...

China --- and Japan --- are already beginning to offload their Treasury reserves, gradually and in small amounts. In the meantime, US Federal debt has ballooned by an additional trillion to $22 trillion. By 2025, the amount of interest the US has to pay would exceed its defense budget.

The US needs more buyers of Treasuries to not only make up for the additional yearly one trillion, but also to compensate for the decreasing number of traditional buyers. And yes, FDI to the US is going down thanks to its increasingly anti-immigrant atmosphere while simultaneously foreigners are beginning to pull out their holdings from the US.

This is happening in the background as the US Millennials are more in debt than ever, with over $1 trillion in student debt alone, and due to that, are buying less houses and cars. The US consumer is in ever greater danger of defaulting on their debts. What happened to the "robust" economy? Was it all just a lie?

I don't think China is caving in on anything. Despite weaker numbers, China actually is going to be stronger on the rebound of next credit cycle. It's only natural for China to dove-tail her domestic reforms with US's 141-point demand. China already gave them 3 large baskets where one is approved items like buying more US goods, another like IP and tech which will be work in progress and the last but not least is the basket of non-negotiable items like structural changes and industrial policies. Trump wants a deal now. He doesn't know nor care much about the nitty gritty details. He just wants a headline that says he's a winner. As he rushes in to close a deal, some similar elements of USMCA and KORUS are copied in and replicated. Rumor has it there will be a currency pact agreement, offering China to a currency swap line to help her preserve hard currency in times of volatility. In turn China has to agree PBOC will work closely with Fed on any and all interventions, just like in KORUS. Competitive devaluation will be off the table, which China is not doing anyways for her own valid reasons. That currency pact will affect 40 % of global GDP, less than Plaza accord's G5 percentage wise, but still a significant number. It would help China reform to open up capital account but it will lose some independence of action to Fed. So let's wait and see if it's a miniature Plaza or not.