We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon
Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Chinese automotive semiconductor company Black Sesame said Wednesday it has reached a valuation of almost $2 billion after receiving an investment from Beijing-based Xiaomi ( OTCPK:XIACF).
Black Sesame didn't give an exact amount of the Xiaomi ( OTCPK:XIACF) investment, but said it was worth "hundreds of millions of dollars."
Xiaomi ( OTCPK:XIACF) is one of China's largest consumer electronics companies, and operates across its home country, as well as India and parts of Europe. The investment comes amid an ongoing shortage of semiconductor components that has had a particular impact on the automotive industry.
Among leading chip companies, Intel (NASDAQ: INTC), Advanced Micro Devices (NASDAQ: AMD), NVIDIA (NASDAQ: NVDA), Applied Materials (NASDAQ: AMAT) and KLA (NASDAQ: KLAC) all saw their shares rise in pre-market trading Wednesday.
Hai Robotics, a Shenzhen, China-based warehouse robotics startup that develops autonomous, case-handling robotic (ACR) system, announced it has raised $200 million. The funding comes from two separate equity rounds, a Series C and Series D, which are being disclosed simultaneously.
Its Series C funding was led by 5Y Capital, with participation from Sequoia Capital China, Source Code Capital, VMS, Walden International and Scheme Capital while the Series D round was led by Capital Today along with existing investors including Sequoia Capital China, 5Y Capital, Source Code Capital, Legend Star and 01VC.
The latest fundings, which come after its $15 million Series B+ round in March this year, will be used to bolster its robot fleet with technological upgrades, penetrate further to the global market, hire talent and support its supply chain management.
Hai Robotics, founded in 2016, has five overseas subsidiaries in Hong Kong SAR, Japan, Singapore, the U.S. and the Netherlands with international customers in more than 30 countries. The company entered Australia in this year by deploying HAI’s warehouse robots at Australia’s largest online bookstore Booktopia to handle packing and dispatch orders.
“Our major future orientation will center on expanding the overseas market and localizing our service,” said Richie Chen, Hai Robotics co-founder and CEO.
Hai Robotics develops an autonomous case-handling (ACR) system that provides warehouse automation solutions via robotics technology and AI algorithms for the logistics warehouse industry. The company said it has more than 400 global patents in areas like robot control and warehouse management, based on its statement.
The global warehouse automation market is projected to increase to $30 billion by 2026, at a CAGR of 14%, from $15 billion in 2019, based on a report by Logistics IQ.
There are a number of startups and larger technology companies building warehouse robotics these days — industrial applications more generally being one of the earliest and most successful use cases for robots. Others include Fetch Robotics, NextShift Robotics and Alog Tech.
Chen also said the company differs from other global peers in many aspects including project costs and customization capability. Hai Robotics has been running over more than 200 projects globally with 2,000 ACR robots deployed. Its ACR system can be applied in various industries like apparel, e-commerce, retail, manufacturing, 3PL, electronics, pharmaceutical, energy and automotive. It has partnered with global logistics and supply chain firms including LG CNS, MUJIN, BPS Global, and Savoye.
In 2015, the company launched its first ACR system (called HAIPICK A42), which can pick and place totes and cartons on storage shelves up to 5-7 meters high and carry up to nine loads to conveyors.
Hai Robotics has several other models in that line.
“The case-handling robot is riding on the market trend that shifts toward smaller workflows, such as from pallet-picking to totes-picking. We’re very pleased to see the company’s fast growth with good innovation,” Guo Shanshan, a partner of Sequoia Capital China said.
China is unlikely to okay Baidu's $3.6 billion purchase of JOYY's YY Live
Reuters | HONG KONG | Last Updated at September 24 2021
China's antitrust regulator is unlikely to approve Baidu's $3.6 billion acquisition of JOYY Inc's video-based domestic live streaming business YY Live, two sources with direct knowledge of the matter told Reuters.
This follows Beijing's move to rein in gaming-related businesses and corporate expansion via deals, they said.
The Baidu-JOYY deal would be the latest multi-billion dollar transaction to stumble amid China's broad crackdown on private companies, notably those in the internet sector, as Beijing seeks to control big data and break down monopolistic practices.
A failure of the Baidu deal could cast a shadow over a separate planned transaction to take Nasdaq-listed JOYY private that would have valued it at up to $8 billion, said the sources and two other sources who are familiar with the matter.
Chinese search engine giant Baidu announced in November it would buy YY Live from social media firm JOYY, paying cash to help diversify revenue sources.
When contacted by Reuters, Baidu said it had not received messages from Chinese regulators that the deal was unlikely to be approved.
The State Administration for Market Regulation (SAMR) - China's antitrust regulator - and JOYY did not respond to requests for comments.
The deal has been awaiting SAMR approval to reach a final close. JOYY was to receive the remaining proceeds of about $1.6 billion from Baidu after the regulatory approval.
SAMR said in a report in early September that it is conducting anti-monopoly reviews of 11 transactions including Baidu's acquisition of YY Live.
But one of the sources, who is familiar with the regulator's thinking, said SAMR was unlikely to give the green light because allowing Baidu to acquire and further invest in a video game-related live streaming business amid the recent crackdown on the gaming industry could send the wrong message to the market.
Another source, who was also familiar with the situation, said SAMR had already indicated to at least one of the deal parties its stance and added that the approval process will likely keep dragging on until the application lapses.
"SAMR hopes the companies can withdraw the application," the source said.
SCRUTINY ON VIDEO GAMING
Beijing has expressed its concern over the influence video games have over the country's minors and last month introduced new rules that limit the amount of time under-18s can spend on video games.
It has also criticised the entertainment industry for "polluting" society and youth and recently published new guidelines for how artists and livestreamers should behave.
Two months ago, SAMR formally blocked Tencent Holdings' $5.3 billion plan to merge the country's top two videogame streaming sites Hula and DouYu on antitrust grounds.
A failure of the Baidu-JOYY deal would be a blow to JOYY's top two shareholders, Chairman David Li and Xiaomi founder Lei Jun, as the dynamics of their plan to take JOYY private would change completely, according to the sources.
They were teaming up for the deal as they believe the Chinese social media company is undervalued in the U.S. market, Reuters reported last month.
Li did not respond to a request for comment made via JOYY. Lei did not respond to a request for comment either.
Founded in 2005, JOYY went public in 2012. It also runs Singapore-based live streaming platform Bigo Live and owns a 16% stake in Huya.
YY Live largely counts on games and entertainment content for its revenue and user traffic. Its average mobile monthly active users reached 42 million in the fourth quarter of 2020, up 1.9% year-on-year, according to the parent's 2020 annual report.
However, its paying users dropped by 1.1% over the same period, which JOYY attributed to the impact of COVID-19 on people's spending.
(Reporting by Julie Zhu; Editing by Muralikumar Anantharaman)
Baidu Releases PLATO-XL: World’s First 11 Billion Parameter Pre-Trained Dialogue Generation Model
It has always been a great challenge for AI bots to conduct coherent, informative, and engaging conversations as human beings. For robots to serve as emotional companions or intelligent assistants it is essential that they develop high-quality open-domain dialogue systems. As pre-training technology further promotes models’ ability to learn from large-scale unannotated data, mainstream research is focusing on making more efficient and full use of massive data to improve open-domain dialogue systems. To this end, Baidu releases the PLATO-XL with up to 11 billion parameters, achieving new breakthroughs in Chinese and English conversations.
· PLATO-XL: Exploring the Large-scale Pre-training of Dialogue Generation
In recent years we have witnessed constant progress in the field of open-domain conversation, from Google’s Meena and Facebook’s Blender to Baidu’s PLATO. In DSTC-9, the top dialog system technology challenge, Baidu PLATO-2 broke a record by winning the first place in five different dialogue tasks.
Now Baidu PLATO-2 has been upgraded to PLATO-XL. Over ten billion parameters make it the world’s largest Chinese and English dialogue generation model. Achieving superior performance in open-domain conversation, PLATO-XL raises our expectation of what hundred-billion or even trillion parameter dialogue models could do.
Introduction to PLATO-XL
PLATO-XL adopts the unified transformer architecture that allows simultaneous modeling of dialogue understanding and response generation, which is more parameter efficient. A flexible mechanism of self-attention mask enables bidirectional encoding of dialogue history and unidirectional decoding of responses. In addition, the unified transformer architecture proves to be efficient in the training of dialogue generation. As we know, given the variable lengths of conversation samples, a large amount of invalid computations is caused by padding in the training process. The unified transformer is able to greatly improve the training efficiency through effective sorting on the input samples.
To alleviate the inconsistency problem in multi-turn conversations, the multi-party aware pre-training is carried out in PLATO-XL. Most of the pre-training data used are collected from social media, in which multiple users are involved to exchange their ideas. The learned models tend to mix information from multiple participants in the context, thus having difficulties generating consistent responses. To tackle this problem, PLATO-XL introduces the multi-party aware pre-training, which helps the model distinguish the information in the context and maintain the consistency in dialogue generation.
The 11 billion parameter PLATO-XL includes two dialogue models, one for Chinese and the other for English. 100 billion tokens of data are used for pre-training. PLATO-XL is implemented on PaddlePaddle, the deep learning platform developed by Baidu. To train such a large model, PLATO-XL adopts the techniques of gradient checkpoint and sharded data parallelism provided by FleetX, PaddlePaddle's distributed training library. PLATO-XL is trained on a high-performance GPU cluster with 256 Nvidia Tesla V100 32G GPU cards.
Superior performance on various conversational tasks
For the comprehensive evaluation, PLATO-XL is compared with other open-source Chinese and English dialogue models. As shown in the following figure, PLATO-XL outperforms Blender, DialoGPT, EVA, PLATO-2, etc. Besides, PLATO-XL proves significantly better performance than the current mainstream commercial chatbots.
In addition to open-domain conversation, PLATO-XL strongly supports knowledge grounded dialogue and task-oriented conversation with proven superior performance. The PLATO series covers dialogue models of different sizes, from 93M to 11B parameters. The figure below shows a relatively stable positive correlation, suggesting that the increase of model size has a significant impact on model performance improvement.
PLATO-XL is able to have logical, informative and interesting multi-turn conversations with users in English and Chinese.
China is a leading player in artificial intelligence (AI), with several public companies such as Baidu, Hikvision, iFlytek, Tencent, and Alibaba.
It also has a strong AI start-up ecosystem, which is evident from the large number of AI unicorns (privately held start-up company valued at $1bn or more). China has the highest number of AI unicorns in the world, with a total of 45 in China as of 31 August 2021. These 45 unicorns are collectively valued at $43.5bn.
The Chinese AI dragons are getting bigger
SenseTime, Megvii, Cloudwalk, and Yitu Technology are collectively called China’s four “AI dragons”.
SenseTime specializes in image recognition software, computer vision, and deep learning technologies. Its solutions are used in autonomous vehicles, smart cities, edtech, and healthtech. Its technology is also useful in the emerging metaverse, which promises to mix reality and a virtual world.
It is currently the most valuable of all the AI unicorns, with a valuation of $12bn, having received total funding to date of $2.6bn. Key investors include Tiger Global Management, Softbank China Venture Capital, and Qualcomm Ventures.
Cloudwalk is a developer of facial recognition software for fintech, public security, and aviation sectors. It is valued at $3.3bn with total funding of $0.5bn. Its key investors include Haier Capital, Industrial and Commercial Bank of China, and Atlas Capital Group.
The two other Chinese AI dragons are Megvii and Yitu Technology. Megvii, a facial recognition software developer for smart cities, is valued at $4bn with total funding of $1.4bn. It is backed by investors such as Alibaba Group Holding, Foxconn Technology Group, and Boyu Capital Consultancy.
Yitu Technology offers an AI computing ecosystem, AI chips and AI algorithm software for smart cities, smart fintech, smart healthtech, and smart retail. It has achieved a valuation of $2.2bn and total funding of $0.4bn. Investors like Sequoia Capital, Hillhouse Capital Group, and Gaorong Capital backed the unicorn with funding.
Apart from these four, Horizon Robotics is another notable unicorn that provides edge AI platforms, AI chips, and AI-based face recognition cameras for smart vehicles, smart cities, and smart retail. It is valued at $5bn and has received a total funding of $3.1bn. Sequoia Capital, Citic Group, Hillhouse Capital Group, and Intel are some of its notable investors.
Expect four to go public soon – despite regulatory challenges
CloudWalk, Yitu Technology, Megvii, and SenseTime were put on the US entity list by the US government over allegations that they had violated human rights and posed a threat to US national security.
SenseTime has filed for an initial public offering (IPO) in Hong Kong in 2021. It did not disclose the size of the deal but is pushing ahead with the planned listing. It plans to invest 60% of the amount raised in IPO to improve research and development (R&D).
Cloudwalk has also filed for an IPO of $0.5bn in the Shanghai Stock Exchange which is now confirmed and is expected to go public soon.
Megvii is also in the IPO pipeline for the Shanghai Stock Exchange’s STAR Market. Approval was given for a $0.9bn IPO on the Shanghai Stock Exchange and expects to go public soon. Their first attempt to list was in February 2020, when it’s Hong Kong IPO filing lapsed because of the impact of the coronavirus pandemic.
Yitu Technology is also considering IPO in Hong Kong after failing to win regulatory approval in Shanghai and seeks a valuation of about $4bn in Hong Kong. Like other AI unicorns, Yitu Technology will invest the capital raised from the IPO in R&D and for market expansion.
The Chinese AI ecosystem will become even stronger as these companies list and are able to invest the significant amounts of new capital raised. This will further boost the broader AI start-up ecosystem in China.
AI, Robotics and Automation board gets refresh! This year the board has been much more active after I became interested in AI a year ago. Several months ago the board got a new name and new moderator (Glenn Petersen), who is revamping the Introduction header, which was focused on an extensive discussion of the Singularity. He has started with a new logo for the board. I like it.
Three years ago, Tang Xiao’ou joked to an audience at Massachusetts Institute of Technology that it was his company, SenseTime, rather than Google, that instantly came to mind when people talked about artificial intelligence.
Today, SenseTime, one of China’s brightest hopes for its plan to lead the world in AI, is preparing for a $2bn public listing in Hong Kong that may come before the end of the year.
Known for its expertise in facial recognition, SenseTime has raised more than $3bn from investors including SoftBank, Alibaba, Tiger Global and Silver Lake, making it one of the four most hyped AI companies in China, together with its peers Megvii, Yitu and CloudWalk.
All four have grown rapidly with Beijing’s backing, which has allowed them to gather and process swaths of data on the country’s huge population, are now trying to list, and have ambitions to expand overseas.
But all four companies have also suffered concerns over the use of their technology and have been hit by sanctions from the US for allegedly aiding human rights abuses in the western region of Xinjiang, a charge the companies have denied.
Since then, the companies have had to tread a careful path to maintain good relations with their most visible and most important customer: the Chinese government.
SenseTime is “caught in a conundrum”, said Jeffrey Ding, postdoctoral fellow at Stanford’s Center for International Security and Cooperation. “Western investors might want them to be more forthright in disavowing [ethnic profiling]?.?.?.?but this would lose them business in China.”
Documents filed by the start-up ahead of its IPO reveal the extent of its reliance on Beijing. Its Smart City business, which includes facial recognition and predictive policing, contributed 40 per cent of its sales last year. Its five largest customers, which made up almost a third of its Rmb3.4bn ($525m) of sales are mostly what SenseTime calls “system integrators”, who serve city administrators.
Since 2008, at least 500 Chinese cities have tried to install so-called Smart City technology, working with private companies like SenseTime, Alibaba, Hikvision and Huawei to find digital solutions to urban problems.
SenseTime’s software has been deployed in 119 cities, the vast majority of which are in China. In one city, its systems are being used to detect people not wearing seat belts in cars, with a claimed precision rate of 94 per cent. It can also spot drivers who are using mobile phones with 86 to 96 per cent accuracy.
In another top-tier city in southern China with 17m people, SenseTime’s traffic management system logged traffic violations by people on mopeds. Police tracked down 50,000 such incidents, and the drop in monthly abuses by mopeds dropped by more than half in just a couple of months. The number of drivers choosing to wear a helmet, meanwhile, rose from under half to 94 per cent.
Megvii, the oldest of the four AI start-ups, also counts the government as its biggest customer. The Alibaba and Ant Group-backed company last year derived 64 per cent of its revenue from smart-city technology. The facial recognition technology company was forced to abandon its Hong Kong listing when Washington blacklisted the company in February last year.
But international expansion has been hampered by concerns about Chinese influence. UK intelligence agencies have, for instance, been pushing for limitson whether local authorities can purchase the software, fearing that it might be used for espionage or surveillance.
SenseTime has over the past couple of years added offices and customers across south-east Asia, north Asia and the Middle East, where reservations over surveillance are also less pronounced.
But it does not give many details on how many overseas deals it has done, especially in the wake of its blacklisting in 2019.
A long-discussed project to detect fraud and cheating in casinos with the leisure group Genting, which operates Singapore’s casino on Sentosa island, has yet to result in any concrete public information. A research hub in Abu Dhabi to provide “AI capabilities” to local start-ups has not announced many deals since it was set up in 2019.
Meanwhile, other Chinese tech companies are investing in their own AI capabilities, in a potential hit to SenseTime’s “Smart business” enterprise arm.
“Unicorns and the other big technology companies increasingly have their AI in house,” said William Bao Bean, a Shanghai-based general partner at global venture capital firm SOSV. “Most Chinese corporates are not in the habit of paying for software, which is why a lot of clients for companies like SenseTime are government.”
One former SenseTime employee told the Financial Times that in the early days it won tech customers including ByteDance, the owner of TikTok, and Hikvision, which makes CCTV cameras.
But by 2019, ByteDance, was “heavily disengaging” from SenseTime in favour of building its own internal teams after working with it on AI facial filters. “It is faster as they had quicker access to the data, and they don’t need to pay SenseTime,” the person said.
A person close to the company acknowledged that it had lost some business but was continuing to expand its base of more than 2,400 customers.
SenseTime has tried to push more heavily into hardware and offer a broader package to clients. It wants to spend a significant amount of its IPO proceeds on research and development, which includes a new AI supercomputing data centre close to Shanghai to be completed in 2022. It is also developing AI chips, which are designed to process AI tasks faster and more efficiently.
Another area is software to support autonomous driving — a phrase mentioned 68 times in its documents, compared to just six mentions of “facial information”.
Even here, SenseTime has tough competition, said experts, and currently the company’s auto division contributes less than 5 per cent of its sales.
“Autonomous driving is a crowded market and SenseTime is a late entry. A lot of dance cards have already been filled by the likes of Baidu, Pony.ai and WeRide,” said Tu Le, founder of consultancy Sino Auto Insights. “I don’t think many would dual source or abandon current partners for SenseTime.”
But many investors are encouraged by the company’s support from the government. SenseTime received Rmb352.8m in government grants in 2020, an increase of 37 per cent on what it received in 2019.
Producing cutting-edge and agenda-setting technology is key to president Xi Jinping’s stated aim that China should be an AI leader by 2030. The success of SenseTime and other AI including Megvii, Yitu and CloudWalk feeds directly into that ambition. The AI sector has not received the degree of scrutiny from regulators that other Chinese internet companies have this year.
As one mainland-based investor in Megvii said: “Is it a bad thing right now if your biggest customer is the Chinese government, which is invested in making sure you do well? We don’t think so.”
According to the report, Apollo’s autonomous driving service has covered four cities including Beijing, Guangzhou, Changsha and Cangzhou, with an operating area of over 600 square km, securing over 410 test operation licenses, providing over 400,000 individual service trips and receiving a 95.3% rate of five star customer reviews.
In the second quarter of 2021, the Apollo autonomous driving service was commercially piloted in Cangzhou and Beijing’s Shougang Park. At the same time, Apollo began to participate in the second stage of unmanned ride testing in Beijing. In Shougang Park, it can also carry out trial operations of autonomous driving. Its unmanned technology has achieved a leading breakthrough. Since May, Apollo’s unmanned vehicles in Shougang Park have carried more than 5,000 passengers and accumulated a total travel distance of over 11,850 km.
On March 12 of this year, the platform officially obtained an autonomous driving test notice and autonomous driving demonstration operation notice issued by the Cangzhou Municipal Government, including a chargeable demonstration operation license. From April 22 to May 31, the platform officially opened its commercial charging function in Cangzhou, and completed more than 1,800 charged orders.
At present, Apollo’s service can achieve full-time coverage and meet diversified needs. The nighttime unmanned ride testing, initiated in Beijing, Guangzhou, Cangzhou and other places, has expanded the service time to 23:00 p.m.
According to national data in the report, in the first half of 2021, the average waiting time for calling cars on Apollo was 4.6 minutes. Cars took orders within 42 seconds on average, while passengers were picked up within 3.9 minutes on average. The data fully show that Apollo’s service booking platform is responsive.
In this year, the platform has received nearly 50,000 user evaluations – over 95% of which included five-star praise – and a relatively high test ride satisfaction level with 4.9 points.
Baidu researchers have proposed a novel reinforcement learning-based evolutionary foot trajectory generator that can continually optimize the shape of the output trajectory for a quadrupedal robot, from walking over the balance beam to climbing up and down slopes. Our approach can solve a range of challenging tasks in simulation by learning from scratch, including walking on a balance beam and crawling through a cave. To further verify the effectiveness of our approach, we deploy the controller learned in the simulation on a 12-DoF quadrupedal robot, and it can successfully traverse challenging scenarios with efficient gaits.
Countdown Starts on Chinese Company Delistings After Long U.S.-China Audit Fight
The three-year countdown will accelerate a decoupling of the world’s two largest economies
By Dawn Lim in New York and Jing Yang in Hong Kong Wall Street Journal Oct. 2, 2021 7:00 am ET
U.S. securities regulators have started a countdown that will force many Chinese companies to leave American stock exchanges, after a long impasse between Washington and Beijing over access to the companies’ audit records.
The action will accelerate the decoupling of the world’s two largest economies and affect investors that own securities in more than 200 U.S.-listed Chinese companies with a combined market value of roughly $2 trillion.
In late 2020, then-President Donald Trump signed a law that bans the trading of securities in foreign companies whose audit working papers can’t be inspected by U.S. regulators for three years in a row. The passing of the Holding Foreign Companies Accountable Act followed nearly a decade of failed attempts by regulators in the U.S. and China to resolve sharply differing expectations over how such audit inspections would be carried out.
The Securities and Exchange Commission is working out details of how the law will be implemented and is finalizing its associated rules. Its chairman, Gary Gensler, has said the clock started ticking this year.
The SEC expects that U.S. regulators could flag Chinese companies in 2022 if they don’t get access to audit work involving 2021 financials of those companies, a person familiar with the matter said.
Anticipating the outcome, some investors have exchanged their American depositary receipts in Chinese companies for shares that trade on Hong Kong’s stock exchange.
New York fund manager WisdomTree Investments in late 2020 swapped ADRs of Alibaba Group Holding Ltd. for the e-commerce giant’s Hong Kong-listed shares, in some exchange-traded funds. The firm is monitoring Hong Kong trading volumes to determine whether it should convert other companies’ ADRs, said Liqian Ren, a quantitative investment specialist.
Wim-Hein Pals, head of the emerging markets equity team at Netherlands-based asset manager Robeco, said he swapped all Chinese ADRs to Hong Kong-listed shares where possible between last year and early this year. Chinese ADRs now represent just 1.5% of his roughly $1.4 billion emerging-markets portfolio.
“We see liquidity moving gradually but consistently to Hong Kong over the next couple of years. More and more investors will go to the Hong Kong-listed names, and neglect their U.S.-listed shares,” Mr. Pals predicted.
Since Alibaba’s landmark secondary listing in Hong Kong in late 2019, 15 more U.S.-listed Chinese companies added so-called homecoming listings in the Asian financial hub, according to Hong Kong stock exchange data. Recent data shows most trading still occurs among Chinese ADRs.
For years, U.S. regulators said they never got the transparency they needed into the work of auditing firms on Chinese companies, because China wasn’t routinely handing over the papers they needed or negotiating in good faith.
The Chinese side, on multiple occasions, said it opposes “politicization of securities regulation,” and that it welcomes dialogue to find a solution.
For data-heavy internet companies, which make up the bulk of U.S.-listed Chinese companies, audit working papers can contain raw data such as meeting logs, user information and email exchanges between companies and government agencies, among other things. In the U.S., the inspections are done by the Public Company Accounting Oversight Board, which the SEC oversees.
China has also said giving a foreign government access to such details for data-heavy tech companies could endanger state security. Earlier this year, Chinese officials wanted ride-hailing giant Didi Global Inc. to put off its New York listing until they could address the audit working paper issues, The Wall Street Journal previously reported.
U.S. officials, in turn, have said China has used the national security argument as a ruse to not open up companies’ books.
The audit standoff has long been a contentious point in cross-border relations between the two countries. For more than a decade, the PCAOB, which functions essentially as the auditor of auditors, has struggled to inspect China-based audit firms, as well as the mainland Chinese affiliates of the Big Four accounting firms.
In 2013, the U.S. and China had a brief breakthrough. Both sides agreed to allow the PCAOB to inspect work done by auditors of U.S.-listed Chinese companies that were being investigated by regulators.
The China Securities Regulatory Commission subsequently turned over the audit papers of four companies for PCAOB’s review. The 2013 pact also paved the way for both sides to talk about a broader set of inspection protocols.
In late 2015, officials from both countries met in Beijing to try to establish those protocols. After two weeks of negotiations, the talks broke down. One deal breaker: Chinese officials weren’t willing to let the U.S. inspect the audit papers of Alibaba and Baidu Inc., two of the most valuable Chinese companies listed on American exchanges.
Shaswat Das, the PCAOB’s chief negotiator at the time, said he understood from prior talks with the Chinese that access would be granted, and took their response—that they needed to consult with other ministries and the State Council first—as a sign they weren’t negotiating in good faith.
The Chinese side had expected the U.S. to eventually come around to “regulatory equivalence,” an arrangement that China has with the European Union, said Paul Gillis, professor of practice at Peking University’s Guanghua School of Management and a former member of the PCAOB Standing Advisory Group. “It basically means the U.S. would accept the work done by the Chinese regulator as if they had done it themselves,” he said.
That wasn’t acceptable to the U.S., people familiar with the SEC and PCAOB’s thinking said.
U.S. and Chinese officials tried to revive talks afterward, but they couldn’t agree on key issues. One sticking point was China’s restricting of information that U.S. regulators considered essential. In 2017, when the PCAOB attempted to inspect an audit of a China-based company, the Chinese didn’t produce the working papers the U.S. demanded and redacted others, according to an oversight board letter to government officials.
In the absence of a resolution, the Holding Foreign Companies Accountable Act was introduced in March 2019.
In April 2020, the CSRC proposed a joint inspection framework under which U.S. officials can conduct inspections and investigations in China with Chinese officials present and access audit papers of companies deemed relevant by the Chinese side.
Accounting irregularities at Luckin Coffee, a rival to Starbucks in China, hardened many politicians’ resolve to push through a bill to enforce tighter audit standards. PHOTO: MARK SCHIEFELBEIN/ASSOCIATED PRESS ---------------------------------------
The proposal was seen as imposing “critical limitations” on the PCAOB’s ability to conduct inspections, according to the oversight board letter.
Around the same time, Luckin Coffee Inc., an upstart rival to Starbucks Corp. in China, admitted to fabricating revenues and expenses. The accounting chicanery hardened many politicians’ resolve to push through a bill to enforce tighter audit standards.
Luckin’s implosion also caused embarrassment back home. The CSRC publicly criticized the company but stopped short of taking any regulatory actions, because Luckin is registered in the Cayman Islands and listed in the U.S.
The CSRC offered an amended proposal to the PCAOB in August 2020. It is unclear what discussions followed.
In June this year, the Senate passed another bill that, if enacted, would shorten the three-year timetable for delistings to two years.
In August, CSRC Chairman Yi Huiman said promoting China-U.S. cooperation on auditing oversight is one of the regulator’s top priorities for the remainder of this year.
The looming threat of delistings gives U.S. officials key leverage on the negotiating table against the Chinese side. “If the U.S. is going to have any success at the negotiating table, this legislation has got to be implemented,” said Mr. Das, who is now a lawyer at King & Spalding LLP in Washington.