Falsifying Data Can’t Save You From Severe Competition: Why We Believe Pony AI Inc. is the Worst of the Robotaxi Hype
- Pony AI Inc (NASDAQ:PONY) is a hyped robotaxi company that went public in November 2024. Recent rumors around interest by former Uber CEO Travis Kalanick in PONY’s U.S. business have driven excitement around the company.
- Our research, including on-the-ground testing in one of the major cities in China and expert interviews, indicates that PONY has, in fact, very little to offer and is more akin to a smoke and mirrors show.
- There are serious allegations from an apparent insider that PONY actively falsified data for the algorithm of its self-driving software. Management is allegedly aware of the issue and was covering it up. We find this reckless and dangerous, especially for a company that develops autonomous driving software.
- While PONY likes to portray itself as an international player, the reality is that the company’s permit to conduct autonomous vehicle testing without a driver in California was revoked a few years ago after a crash. Pony had a similar incident in May 2025 in China, which reportedly led to the temporary suspension of the service in that district. It appears that PONY currently only has a permit to conduct autonomous vehicle testing with a driver in the U.S.
- We tested the company’s robotaxis in available districts in China. Based on our experience, it appears PONY has the least pick-up spots, longest waiting time, and overall worse customer experience than the industry leader there. In our opinion, PONY is clearly losing against Chinese peers such as Baidu Apollo and WeRide.
- Expert interviews confirm the issues we saw in our on-the-ground due diligence. PONY is lacking data, a high-definition map, and would need enormous funding and time to catch up with peers, according to industry experts.
- The economics of PONY currently appear dire. Financials worsened after the IPO and the company heavily relies on its related-party transactions. A big portion of PONY’s total revenue comes from an entity directly related to the Chinese military.
- PONY’s close ties to the Chinese Government caused Senators to call for a delisting of PONY from U.S. stock exchanges. We think these issues themselves make a potential acquisition of the U.S. business difficult despite management claiming a separation of the U.S. and Chinese businesses.
- PONY is trading at record valuations, and the recent expiration of the lock-up restrictions is poised to put pressure on the company’s share price.
Introduction
Pony.ai Inc. (Nasdaq: PONY) is a China-based autonomous vehicle (AV) technology company specializing in software deployment and maintenance, vehicle integration, and road-testing services for original equipment manufacturers (OEMs) and transportation network operators. In 2018, they launched PonyPilot—an autonomous ride-hailing service that enabled users to book robotaxi rides in Guangzhou.
Pony.ai made early inroads into the U.S. market, securing a 3-month autonomous vehicle testing permit in California as far back as 2018. However, by late 2021 the California DMV suspended a driverless testing permit for Pony.ai after a vehicle hit a road center divider and a traffic sign in Fremont after turning right. The incident also prompted an investigation by the NHTSA, culminating in a formal recall.
PONY likes to tout its international business and advanced software. In June 2025, The New York Times first reported that former Uber CEO Travis Kalanick is allegedly interested in purchasing PONY’s U.S. business with Uber potentially financially backing the transaction.
After extensive research including testing of AV services in China, expert interviews, and consultation with former employees, we believe that PONY is in fact a dishonest company with severe issues that is currently losing the AV race against better funded competition. With the recent expiration of lock-up restrictions, we believe PONY will see mounting pressure on its business and its stock.
Praying for a Buyout
In June 2025, the New York Times reported that Uber and its previous CEO Travis Kalanick might be in talks to help fund his acquisition of PONY’s U.S. subsidiary. The article stated the company “has permits to operate robot taxis and trucks in the United States and China.” The report does not mention the specific valuation of PONY’s U.S. subsidiary and states that the “talks are preliminary.”
We have seen unfounded excitement around buyout rumors before, recently in Sigma Lithium Corporation which we criticized in a critical report in March 2023. We believe a transaction is in this case not only ill-advised for Uber and Travis Kalanick, but also unlikely to positively surprise investors if it indeed happens.
First of all, PONY’s license to conduct autonomous vehicle testing without a driver in California has been revoked by the California Department of Motor Vehicles after a crash a few years ago. Currently it only holds a license to conduct autonomous vehicle testing with a driver. Therefore, we do not believe the statement in the article that the company “has permits to operate robot taxis and trucks in the United States and China” is entirely correct.
The currently existing U.S. operations of PONY are tiny. According to the company’s annual report, it only generated $321K in revenues from the U.S. in 2023 and $604K in 2024. PONY’s entire business had a negative free cash flow of over $120m in 2024 and is in dire need of funding to realize its ambitions. In terms of PONY’s mileage on the road in California, according to the website of the California DMV, the company only had about 51.6K autonomous miles from December 2023 to November 2024.
Uber’s interest should not be overinterpreted either. Uber has stated that it has a platform strategy that involves partnering with a range of AV companies.
While PONY announces many strategic partnerships in different countries, you can only access their AV service in China where it seems to lose the race against competitors.
Without a driverless testing permit in the U.S., or any meaningful operations or financials, the value in acquiring a U.S. business would mostly lie in the technology, especially the software Pony has developed. Here is where we uncover possibly the most concerning issues. Allegedly, PONY not only falsified data for its AV software, but management was also aware of the issues and was actively covering them up.
The second big issue with the technology is in our opinion PONY’s proximity to the Chinese Government. PONY is a Chinese company, and a PLA affiliated company is in fact PONY’s biggest customer and supporter. PONY allegedly forked the source code of its technology in 2022, and therefore the AV technology developed over the past 2 years was developed in the United States, not China. We distrust this statement and apparently the leading U.S. government voices agree with us. Chairman Moolenaar of the House Select Committee on China and Chairman Rick Scott of the Senate Committee on Aging sent a letter to Securities and Exchange Commission (SEC) Chairman Paul Atkins urging the Commission to begin delisting Chinese companies that pose serious national security and investor protection risks. The letter explicitly mentioned PONY as checking all 3 boxes in terms of military-civilian fusion, forced labor, and party branch affiliation.
Given these issues and other problems we raise in this report, we believe a transaction as portrayed by the New York Times article is not only unwise, but also unlikely to ever happen.
Alleged Company Insider Exposes PONY’s Data Falsification and Other Issues
Our research led us to a Chinese website where an apparent company employee posted in 2023 very concrete allegations exposing that PONY’s autonomous driving technology involves data falsification and other issues. The text below is the paraphrased translation from the original Chinese posting.
“Recently, many WeChat public accounts have been reposting PR articles from Pony.ai, which are practically hyping up their IPO using language that borders on deceptive advertising. I’ve had enough and decided to expose some of the truths behind Pony.ai.
Straight to the point:
1) The technical manager responsible for path planning algorithms at Pony.ai was exposed by a colleague for data falsification. Rather than addressing the issue, senior management tried to delay and cover it up, even going so far as to use the investigation as a way to identify the whistleblower. Pony.ai itself doesn’t even know the real safety statistics of its algorithm. Despite knowing the algorithm is unreliable, they still continued real-world vehicle operations, completely disregarding passenger safety and seriously endangering public safety.
2) The so-called L4 business that Pony.ai promoted in fundraising rounds before Series D was never feasible. They had already abandoned the L4 commercialization path, which has inherent technical flaws. Even achieving scalable L3 is difficult. With weak productization capabilities, they have no competitive edge in the highly competitive L2 market. Without the L4 business, it’s questionable whether they’re even worth a fraction of their D-round valuation of 8.5 billion. The valuation is grossly inflated, making it virtually impossible to raise more capital. Given how fast the autonomous driving industry burns cash, this spells a dead end. Management likely already knows the IPO may fail, and amid a deep economic downturn, they are just trying to help senior executives and employees cash out by buying back shares.
3) The IPO hype is all smoke and mirrors. Pony.ai’s PR claims “large-scale operation of fully driverless autonomous vehicles in busy urban areas of Beijing and Guangzhou.” The truth is, they can only operate in places like Beijing’s Yizhuang and Guangzhou’s Nansha, where there’s barely any traffic. Even deploying the same algorithm in another similarly low-traffic area would take over a year of tuning. So forget about “large-scale operation” in “busy urban areas.” For a company that claims to be technically driven, Pony.ai plays word games just like education-tech companies that rely solely on marketing. They even claim their “BEV perception algorithm” and “game-theory interactive planning algorithm” are self-developed and original. This is like a Chinese EV manufacturer claiming to have invented the electric vehicle. In reality, Pony.ai just follows top-tier company Waymo. When Waymo publishes a new method in a paper, they copy it—and even their replications are full of errors. Their tech lags Waymo by at least three years. Even Waymo’s technical director admits that the current tech stack can’t achieve large-scale deployment in busy cities.
4) The company culture is toxic. Management enables incompetence and pushes out talented employees. Many capable people have left, while those who remain are often unqualified and unethical. These people manipulate data and shift blame to others to protect their own interests. Even when exposed by colleagues, they face no consequences because management has formed its own self-serving interest group.
Supplementary Note (2023-01-18 12:05 +8:00):
I just received an email from Pony.ai’s legal counsel accusing me of “misrepresent to the public” “without evidentiary basis and fact,” also citing confidentiality clauses and threatening legal action. I replied all my statements are either based on sufficient evidence/facts, or based on publicly known information about autonomous driving technologies, shared as personal opinion, or self-derived general statement based on personal conversation with industry leader. All my technical comments about autonomous driving are things that anyone with experience in the field and who has contact with the VC community can easily conclude.
To avoid confusion, I add the following clarifications:
1) Data falsification refers to internal data falsification—plainly, deceiving colleagues to get one’s algorithm deployed. Before I reported this to Pony.ai’s executive Tiancheng Lou, senior leadership either didn’t know or pretended not to know. After reporting, HR replied with an email stating they are investigating, but there’s solid evidence they never conducted a real investigation.
2) “L4 business is unviable” refers to all domestic L4 companies. With their current heavily rely on rule approaches, it’s impossible to scale. This is not aimed solely at Pony.ai. I never worked in their planning team and don’t care what specific methods they use. What I do know is that Pony.ai CTO Lou is a “rule-based true believer”. He previously had a disagreement on direction with Andrew Ng while at Baidu—this was told to me by a former Baidu Apollo executive.
3) Pony.ai also accused me misrepresent to the public that the company has copied and infringed on Waymo’s technology. But in the autonomous driving field, the widely known fact is: most papers come from Waymo Research. When people see some good models, they would use them. I merely stated a basic fact. Pony.ai’s legal team twisted it into me claiming they “copied and infringed on Waymo’s technology,” which I never said. Anyone with basic understanding would take it this way. If reproducing a published paper constitutes “copied and infringed on Waymo’s technology, does that mean Waymo could go around suing every self-driving company and academic group out there? Waymo could sue every AV company and academic lab on the planet. As for the statement regarding the “Waymo tech director” —again, that’s based on my personal conversation with industry leader and a personal opinion based on public knowledge of AV tech paths and situations.
I also want to publicly warn Pony.ai. Do not use legal action as threats and intentionally twist my statements. If you infringe upon my legal rights, I will pursue personal legal action against the individuals responsible under the law.”
These allegations are in our opinion credible and extremely concerning. Autonomous driving can be extremely dangerous, as faulty software may cause crashes and potentially cost lives. We believe that while thankfully nobody has gotten seriously injured yet, PONY’s at best sloppy treatment of its software has already led to incidents and the revocation of one of the company’s most valuable licenses in the U.S.
Read the full article here by Grizzly Research
