(The audio version of this post can be found on the Interconnected YouTube channel):

China’s semiconductor ambition just had its first ponzi scheme fully exposed: Wuhan Hongxin Semiconductor Manufacturing (HSMC).

The HSMC ponzi scheme was led by a trio of characters, who have zero expertise in semiconductor (or anything tech related), but are experts in manipulating local government subsidies, construction contractors, a renowned but gullible former TSMC executive, and China’s desperate need for homegrown chips to pull of a heist so large it makes Theranos look amateur.

Details of this scheme, documented and exposed in this 36Kr article (in Chinese; hat tip: Jordan Schneider), illuminates a bigger structural problem in China, where the central government’s desires, and the local governments’ desire to satisfy those desires, can produce ripe opportunities to scammers to steal and profit on a monumental scale.

(Note: during the course of writing this article this week, the 36Kr report was deleted.)

The Heist

Let’s first summarize the major elements of the HSMC heist that unfolded between late 2017 and early 2021 (which may read like a movie script, not real life):

Part I -- the Trap:

  • Throughout 2017, a man by the name of Cao Shan traveled across China looking for a local government to invest in his semiconductor scheme. (“Cao Shan” is actually a fake name this person uses, because his real name is already tainted by the scams he used to do back in his hometown.)
  • Cao eventually found an accomplice, Mr. Long Wei, who worked his connections to get the City of Wuhan’s East-West Lake District Government to provide land and investment.
  • Long brought another close friend into the fold: Ms. Li Xueyan, a small business owner who has opened restaurants and sold Chinese rice liquor.
  • The trio -- Cao, Long, Li -- formed the board of directors of what became HSMC.

Part II -- the Money:

  • Throughout 2018, the trio worked to secure two sources of “income”: direct subsidies from the East-West Lake District Government (aka taxpayer money) and deposits from construction contractors who want to build the HSMC factory.
  • Sourcing both government subsidies and contractor deposits is a strategy for scam factory projects to increase the amount of money to be scammed.
  • The East-West Lake District Government decided to invest in HSMC partly because of its jealousy of a local rival district, which attracted and incubated a successful flash storage manufacturer.
  • To make themselves look important and powerful, the trio would spread false rumors about their personal background. Long was rumored to be the grandson of some high-level official, while Li would pretend to be the sister of some other political figure.
  • By May 2019, HSMC has received 6.5 billion RMB (~$1 billion USD) of investment from the district government. Cao and Long have quit the board, giving Li and her cronies full control. Cao began going to other provinces to set up similar ponzi schemes.

Part III -- Chiang Shang-Yi, TSMC, and ASML:

  • By June 2019, the trio targeted and successfully persuaded Chiang Shang-Yi, the legendary founding CTO of TSMC, to join HSMC as its CEO.
  • To convince Chiang, HSMC spread false rumors publicly that it has already attracted 100 billion RMB (~$15 billion USD) of investments. They also took advantage of Chiang’s gullibility and professional insecurities. (At the time, Chiang was a consultant at SMIC, China’s largest chip foundry, with relatively little influence.)
  • Using Chiang’s aura, HSMC started aggressively poaching engineers from TSMC with salary packages worth 2 to 2.5x more than what they were earning.
  • Chiang also used his industry reputation to convince ASML, the Dutch company and world’s leading manufacturer of lithography equipment, to sell one DUV equipment to HSMC. (DUV is not the most advanced equipment, but this is still a huge coup given heavy US pressure at the time on the Dutch government to not sell to China.)
  • By December 2019, the ASML equipment was delivered to HSMC amidst huge fanfare. The company also secured more investment due to this accomplishment from the district government, totaling 15.3 billion RMB (~$2.4 billion USD).
  • One month later, the same equipment was offered as collateral to a local Wuhan commercial bank for a 580 million RMB loan (~$90 million USD) -- another new source of “income” for the heist.

Part IV -- HSMC Collapses, Heist Completed:

  • During the first half of 2020, while Wuhan was ravaged by the coronavirus, the trio began siphoning HSMC money away.
  • One of its primary methods was conducting employee training programs with a company run by Li’s younger brother.
  • HSMC also refused to pay its construction contractors money, owing tens of millions of dollars.
  • By July 2020, it became clear that HSMC was a scam. Chiang left the company. The Wuhan city government started leaking news that HSMC is running out of money.
  • By November 2020, Li was pushed out of the company and the East-West Lake District Government took full ownership of HSMC.
  • By January 2021, Chiang rejoined SMIC. HSMC furloughed all its employees for 40 days and reduced salaries across the board.

This is a timeline (in Chinese) of the major events put together by the 36Kr team:

Central and Local

While the HSMC saga may read like a movie script and an isolated incident, it’s a symptom of a structural problem. This symptom is amplified by the high-profile nature of China’s need to produce its own semiconductors.

From the outside, a common stereotype and myth about China is that Beijing’s central government is omnipresent and omnipotent. In reality, the dynamic is more of a call-and-response; the central government calls with directives, local governments respond with implementation plans.

I described the importance of understanding this dynamic last year in Part II of my “Open Source in China” series. I emphasized the role that provincial governments play in aligning and allocating resources according to the central government’s desires, in order to compete for star projects and build new companies. Local officials are motivated to deliver results for political advancement.

In a recent piece by Matt Sheehan of the think tank MarcoPolo, he also illustrated this relationship when analyzing China’s plans around “5G+ Industrial Internet.” Four months after the central government issued its “5G+ Industrial Internet” guidance, the Guangdong provincial government issued its implementation plan, and the city of Huizhou in the Guangdong province soon issued its own plan to align with the provincial government’s plan. What tangibly resulted was Guangdong (the province) offering a 30% public cloud discount to push businesses to move to the cloud, while Huizhou (the city) pledging 100 million yuan (~$14 million USD) in subsidies to attract so-called “intelligent manufacturing projects” to set up shop there.

HSMC is the archetypal child of this system -- a new shiny company with a rockstar legendary CEO that would’ve done wonders for the political future of officials in Wuhan and the East-West Lake District, if it had succeeded. The allure was so strong that no one from the district government bothered to scrutinize the outlandish promise that Cao pitched -- HSMC will start producing 14nm chips right away, then quickly advance to 7nm chips and compete directly with TSMC and Samsung. (A more realistic timeline would be to start with 65nm chips, move to 40nm, and then progressively to denser chips; a process that could take at least a decade.) Cao knew nothing about semiconductor production, but he did know that the smaller the chip, the more advanced it is, and the more investment he could scam by promising it.

Herein lies a more vexing problem.

The same reasons that make semiconductor manufacturing hard also make it easy to scam. Few people have the knowledge to oversee and pinpoint why a project failed, which party should be responsible, and how to prevent similar failures. This ignorance extends all the way to the top. When the National Development and Reform Commission, the central government regulator, learned of HSMC’s failing, its official response was “谁支持,谁负责” (whoever supported the project is responsible). There’s not much more it could do.

Furthermore, such a failure in the high-profiled semiconductor space is so embarrassing publicly for the local government that there’s no incentive to litigate the truth, because an honest investigation would inevitably uncover gross incompetence by local officials. As a case in point, according to 36Kr’s report (now deleted), after the district government took over ownership of HSMC, it did not spend time inspecting wrongdoings, but instead traveled to Shanghai and other places to try to sell what remains of HSMC to salvage this failed investment (so far without any success).

For professional scammers like Cao, Long, and Li, semiconductor projects are ripe targets -- low risk, high reward, too complicated for most people to understand, and too embarrassing to investigate if projects fail.

That’s why Cao, instead of sitting in jail, has formed several other “semiconductor factory projects” in cities like Jinan and Zhuhai. He is still out there, trying to replicate other HSMC’s.

Shooting Yourself in the Foot

Last October, I wrote a column “What Can $1.4 Trillion Buy” for The Wire China, where I identified the two main chokepoints to China’s semiconductor ambition: access to advanced equipment, hiring technical talent. The greed of the “HSMC trio” has managed to shoot China in the foot in both areas, permanently damaging the prospect of other semiconductor upstarts that may actually be trying to make chips.

Regarding advanced equipment access, by luring Chiang and using his reputation to buy ASML’s DUV equipment (only to loan it off for more cash), ASML may never sell to another Chinese semiconductor startup, with or without sanctions. ASML’s reputation is damaged by this transaction. It doesn’t sell to any company that has the money to buy; it evaluates the buyer’s technical capabilities first to make sure the buyer can make proper use of its product. ASML toured HSMC prior to agreeing to sell its equipment, and lauded HSMC’s team as the best it has seen in Mainland China, mostly on the strength of Chiang’s reputation and the engineers he was able to attract. ASML now must feel like a fool. But the bigger fool here is China. ASML had a record 2020, selling 258 lithography equipment, primarily to TSMC. ASML doesn’t need the Mainland China market; China urgently needs ASML’s products.

Regarding hiring technical talent, a chilling effect will permeate the industry, where experienced engineers from Taiwan will think twice about joining a mainland startup, since even the reputation and judgment of Chiang cannot protect you from scam artists. This situation cuts off the best and most obvious source of talent, given the cultural and language similarities between Mainland China and Taiwan. Poaching talent from South Korea and the United States would be much harder. And training domestic talent will take much (much) longer.

HSMC isn’t the only semiconductor startup that failed in 2020. Last May, a startup in Chengdu failed after two years of operation and supposedly attracted $10 billion USD in investment. Last July, another startup based in Nanjing filed for bankruptcy after four years of operation, with a supposed investment size of $3 billion USD. We don’t know yet why these startups failed.

And that’s the key question: why?

According to a report from Caixin (in Chinese), in just the last three years, the number of semiconductor startups in China have increased by more than 50%. Moreover, the amount of investment has increased 6-times, from $946 million USD in 2018 to $6.16 billion USD in 2019. For the first half of 2020, investment has reached $8.46 billion USD.

Source: https://finance.sina.com.cn/tech/2021-01-04/doc-iiznezxt0546437.shtml

Undoubtedly, not all of these investors know what they are investing in; most of them don’t. Similarly, not all of these startups will succeed; most of them won’t.

Failures are inevitable for something as hard as semiconductor manufacturing. Learning about why failures happen is an important necessity to a new industry’s growth and maturity. But what did China’s fledgling semiconductor industry learn from the HSMC experience?

Based on 36Kr’s calculation, the “HSMC trio” stole in total 12.4 billion RMB (~$2 billion USD) in three years from the three-course meal of district government investment, contractor deposits, and bank loans. (In comparison, Theranos evaporated roughly $1.4 billion USD of its gullible investors’ money in about 15 years.)

The HSMC heist has managed to make Theranos look somewhat benign. If HSMC isn’t treated as a criminal act (like Theranos) but as just another failed startup in a frothy market, it will likely not be China’s only “semiconductor Theranos.”

If you like what you've read, please SUBSCRIBE to the Interconnected email list. To read all previous posts, please check out the Archive section. New content will be delivered to your inbox (twice per week). Follow and interact with me on: Twitter, LinkedIn, Clubhouse (@kevinsxu).

中国的 "半导体Theranos":弘芯

国内红红火火的半导体行业里的第一个大骗局最近被赤裸裸的暴露了:武汉弘芯半导体(“弘芯”)。

弘芯骗局背后的“三人帮”,不懂半导体,甚至不懂任何与科技有关的东西,但很懂怎么操纵地方政府的补贴、承包商的保证金、一位著名但容易上当的前台积电高管以及对国产芯片的迫切需求而产生的各种泡沫。这场骗局的规模,让在美国臭名远扬的Theranos看起来很小儿科。

这篇36氪的深度报道(感谢司马乔丹的分享)记录并揭露整场骗局的细节,也揭示了在国内更大的结构性问题。中央政府的欲望和地方政府想满足那些欲望的欲望,给投机分子带来了诱人的“盗窃”机遇。

(注:这几天在写这篇文章的过程中,36Kr的报道被删了。)

盗窃细节

我们先来总结一下从2017年底到2021年初在弘芯发生的主要事件(读起来可能像电影剧本,而不想真实事件)。

第一步 -- 下套:

  • 2017年,一个叫曹山的人全中国到处跑,寻找地方政府投资他想下的“半导体套”。("曹山"都是不是这个人的真名,是用来行骗的假名,因为他的真名已经被以前在老家做的各种骗局而玷污了)。
  • 曹山最终找到了一位叫龙伟的帮凶,他动用关系,说服了武汉市东西湖区政府提供土地和投资。
  • 龙伟又把另一位好友也拉进局,李雪艳。她的创业履历是:开过餐馆、卖过烈酒和其他小生意。
  • 这三人 -- 曹、龙、李 -- 组成了弘芯的董事会。

第二步 -- 圈钱:

  • 2018一整年,“三人帮” 圈了两种 "收入"来源:东西湖区政府的补贴和投资(也就是纳税人的钱)和建造弘芯工厂的承包商的押金。
  • 政府补贴和承包商押金 “两边吃”,是做工厂项目欺诈行为的一套策略。
  • 东西湖区政府决定投资弘芯,部分原因是嫉妒当地另一个区成功吸引并孵化的一家Flash存储芯片制造厂。
  • 为了夸大自己的影响力和神秘感,“三人帮” 故意散布关于个人背景的虚假谣言。龙某被传是某高官的孙子,而李某则假装是某政治人物的妹妹。
  • 2019年5月,弘芯已获得区政府65亿元人民币(约合10亿美元)的投资。曹、龙退出了董事会,控制权完全让给李和她的亲信们。曹开始到其他省份设立类似的“芯片骗局”。

第三步 -- 蒋尚义、台积电、ASML:

  • 到了2019年6月,“三人帮” 成功说服了台积电的传奇创始CTO,蒋尚义,加入弘芯担任CEO。
  • 为了说服蒋,弘芯公开散布谣言,称已经吸引了1000亿人民币(约150亿美元)的投资。他们还利用了蒋的轻信和对自己职业状况的不满。(当时,蒋是中芯国际的一位资深顾问,但没什么影响力。)
  • 利用蒋的光环,弘芯开始积极从台积电挖工程师,薪资待遇比他们的收入高出2到2.5倍。
  • 蒋还利用自己在业界的名声,说服荷兰的ASML(全球领先的光刻设备制造商),向弘芯出售一台DUV设备。(DUV并不是最先进的设备,但考虑到当时美国对荷兰政府施压,要求不要卖产品给中国,能搞到一台DUV还是非常不易的。)
  • 到了2019年12月,ASML设备隆重的交付给弘芯,公司也因此获得了区政府更多的投资,总额达153亿人民币(约24亿美元)。
  • 一个月后,同一台设备被抵押给武汉当地的一家商业银行,拿到5.8亿元人民币(约合9000万美元)的贷款,变成了“三人帮”一条新的 "收入" 来源。

第四步 -- 弘芯崩垮,盗窃完成:

  • 2020年上半年,在武汉被疫情肆虐的同时,“三人帮” 开始从弘芯的账户中套钱。其主要手段之一是与李某的弟弟开的公司合作做大量的员工培训。弘芯还拒绝支付给承包商钱,欠下数千万美元。
  • 到2020年7月,很明显,弘芯是个骗局。蒋离开了公司。武汉市政府开始泄露弘芯缺钱的消息。
  • 到2020年11月,李某被赶出公司,东西湖区政府全面接管弘芯。
  • 到2021年1月,蒋重新加入中芯国际。弘芯将给所有员工放40天大假,全面降薪。

以下是36氪团队整理的重大事件的时间线:

中央和地方

虽然弘芯的故事读起来像电影剧本,像一座孤立的事件,但其实是个结构性问题的症状之一。由于中国急需生产本土半导体,这种症状也被高调地放大了。

从外界来看,很多西方人士对中国的一个常见偏见是,北京的中央政府无所不在、无所不能。事实上,更准确的是一种号召和响应的关系;中央政府用指令号召,地方政府用实施计划响应。

去年,我在 "中国的开源世界" 系列文章第二部中描述了需要理解这层关系的重要性。我强调省级政府在跟随中央的意愿时,调整和分配资源所扮演的角色,从争夺明星项目,创建新企业,到落实出成果。这都是地方官员为了升官而积累业绩的重要环节。

在智库MarcoPolo的Matt Sheehan最近的一篇分析中国 "5G+工业互联网" 计划的文章中也讨论到这一点。在中央政府发布 "5G+工业互联网" 指导意见四个月后,广东省政府就发布了实施方案,而广东省惠州市也很快发布了自己的落实计划,响应省政府。最终的结果是,广东省提供30%的公有云优惠,以推动企业上云,而惠州市则承诺提供1亿元人民币(约1400万美元)的补贴,以吸引所谓的 "智能制造项目" 去当地落户。

弘芯其实是这一体系的典型 “产品” -- 一家新新闪亮的公司,配上一位有传奇色彩的CEO,如果成功了,将为武汉和东西湖区大小官员赢下显赫的政绩,保证今后步步高升。诱惑力如此之大,以至于区政府没有人去仔细审视曹某说出的与行业现实完全脱钩的承诺:弘芯一开始就生产14nm芯片,然后迅速推进到7nm,直接与台积电和三星竞争。(更现实的时间表是,从65nm芯片开始做,然后升级到40nm,逐步向更高密度的芯片发展;整个过程一般至少需要十年时间。) 曹某虽然对半导体生产一窍不通,但他知道,芯片越小,越先进,他能骗取的投资就越多。

这里面就呈现出了一个更麻烦的问题。

制造半导体背后难的原因,同时也是它容易变成欺诈目标的缘由。很少有人足够的专业知识来监督和审视一个项目为什么失败,应该哪一方负责,以及如何防止未来类似的失败。这种“无知”一直延伸到最高层。当国家发改委得知弘芯的各种问题时,官方姿态仅是 "谁支持,谁负责"。最高监管也做不了什么。

还有一个层面:一个半导体项目的失败会吸引很多注意力,无法轻描淡写,会让地方政府非常难堪,所以其实没有动力去诉讼真相,因为真相的背后肯定会揭发许多官员的严重失职。根据36氪(已经被删了)的报道,区政府在接手弘芯后,没有花时间去检查和追究责任,而是前往上海和其他地方试图出售弘芯剩余的资产,来挽救这笔失败的投资(至今还没有卖出去)。

对于曹、龙、李这样的职业骗子来说,半导体项目非常有吸引力 -- 风险低、回报高,对大多数人来说太复杂,无法理解,失败了也不好意思去调查。

所以曹某不但没有坐牢,还继续在济南、珠海等城市继续“下套”,试图复制下一个弘芯。

砸自己的脚

去年10月,我曾为The Wire China写过一篇专栏《1.4万亿美元能买到什么》(英文),其中指出了站在中国半导体雄心壮志前面的两大障碍:获得先进设备,招聘技术人才。弘芯 “三人帮” 的贪婪成功地在这两方面砸了中国自己的脚,永久地损害了其他可能真正在努力尝试制造芯片的半导体创业公司的前景。

在获得先进设备这一方面,通过引诱蒋并利用他的名声购买ASML的DUV设备(随后直接抵押成现金),可能导致ASML永远不会再卖给另一家中国半导体创业公司,不管有没有制裁。ASML的声誉已因这次交易受损。它通常就不会卖给任何一家有钱想买的公司,而首先会评估买方的技术实力,以确保能正确使用其产品。ASML在同意出售设备之前参观了弘芯,并称赞弘芯的团队是它们在中国大陆看到最好的团队,这主要是因为蒋的名声和他能够吸引到的工程师的实力。ASML现在一定觉得自己上当受骗了。但这里更惨的被骗对象是中国整个半导体行业。ASML在2020年拿下了创纪录的业绩,卖出了258台光刻设备,主要是卖给台积电。ASML不需要中国大陆的市场,而中国急需ASML的产品。

在招聘技术人才方面,行业内将弥漫着寒蝉效应,来自台湾的资深工程师会在加入大陆创业公司之前更加三思而后行,毕竟连蒋的声誉和判断力都无法保护自己不上当受骗。鉴于中国大陆和台湾在文化和语言上的接近,去台湾挖人本来是最佳选择。要想去韩国和美国挖人可难得多。培养国内自己的人才将需要更长的时间。

弘芯并不是唯一一家在2020年失败的半导体创业公司。去年5月,成都格芯在运营两年后失败,据说吸引了100亿美元的投资。去年7月,南京德科码在运营四年后申请破产,据说投资规模为30亿美元。我们还不知道这些公司为什么失败。

而这就是最关键的问题:为什么?

根据财经的报道,在过去三年里,中国的半导体创业公司数量增加了50%以上。此外,投资金额也增加了6倍,从2018年的9.46亿美元暴涨到2019年的61.6亿美元。2020年上半年,投资额已达84.6亿美元。

来源:https://finance.sina.com.cn/tech/2021-01-04/doc-iiznezxt0546437.shtml

毫无疑问,并不是所有这些投资人都知道自己在投什么,绝大多数都不知道。同样,并不是所有这些创业公司都会成功,绝大多数都不会。

对于像半导体制造这种难事,失败是不可避免的。搞清楚为什么失败,而从中学习是一个新产业成长和成熟过程中非常重要也必要的。那对于中国刚刚起步的半导体产业,从弘芯的经验中学到了什么?

根据36氪的计算,弘芯 “三人帮” 在三年内从区政府的投资、承包商的押金和银行贷款这三道大餐中共窃取了124亿人民币(约合20亿美元)。再看看Theranos,它在15年左右烧掉了大约14亿美元投资。

弘芯这个“套”somehow把Theranos变得看起来也没那么可恶。如果弘芯一案不被当作犯罪行为处理(像Theranos一样),而之当作泡沫中又一家失败的创业公司,那它很可能不会是中国唯一的 "半导体Theranos"。

如果您喜欢所读的内容,请用email订阅加入“互联”。要想读以前的文章,请查阅《互联档案》。每周两次,新的文章将会直接送达您的邮箱。请在TwitterLinkedIn、Clubhouse(@kevinsxu)上给个follow,和我交流互动!