He Lost 1.9 Billion Yuan on One EV. The Real Reason He’ll Never Invest Again Isn’t What You Think.

September 15, 2026, Zhuzhou, Hunan. On the sidelines of the Fifth BeiDou Scale Application International Summit, a reporter from Xiaoxiang Morning Herald · Morning Video asked Zhou Hongyi whether he would still invest in new-energy vehicles. Zhou Hongyi answered without pausing: “I’ve already taken a loss once. I definitely won’t invest again. Right now, the opportunities in AI and embodied intelligence are probably greater.”

The reporter did not ask what that “one loss” specifically referred to. Everyone present knew.

A RMB 1.9 billion tab

In 2021, 360 announced an investment in Hozon New Energy, the parent company of Neta Auto. It planned to contribute a total of RMB 2.9 billion. Once fully invested, it would hold 16.5940%, becoming the largest shareholder outside the management team. It initially invested RMB 1.9 billion.

Zhou Hongyi had specific expectations for this investment. He hoped to put 360’s software and security capabilities into Neta’s hardware: smart cockpits, connected-car security, data protection. He personally stepped in as “product manager,” publicly criticized Neta’s marketing for “always self-indulgent,” and told then-CEO Zhang Yong to “learn from Lei Jun” in product presentations. In 2022, Neta Auto topped the annual sales ranking of new EV makers with over 150,000 deliveries.

Behind the sales crown were the financials. From 2021 to 2023, Neta Auto’s cumulative net loss exceeded RMB 18.3 billion. It burned about RMB 16 million a day on average. Its main models were concentrated in the below-RMB 100,000 market. Battery packs accounted for about 40% of vehicle cost. With R&D amortization, channel building, and marketing expenses added on, gross margins were negative for a long time. Only with annual sales above 100,000 units could it approach break-even. Neta traded price for volume in the price war, and losses continued to widen.

360’s financial reports recorded the cost. In 2022, it lost nearly RMB 600 million from the Neta investment. In 2023, it lost another nearly RMB 700 million. Over two years, the losses totaled about RMB 1.3 billion.

In June 2022, 360 announced it would transfer, for zero yuan, a 3.532% stake corresponding to RMB 1 billion of unpaid capital, and waive the remaining capital contribution obligation. After the transfer, it retained an 11.4266% stake and would no longer increase its investment. Zhou Hongyi’s explanation at the time: “Neta Auto is a startup. The founding team should still lead the company’s development, not capital entirely.”

What the announcement did not say: continuing to invest would not help both companies and would drag down the financial health of 360’s core business. 360’s core revenue came from cybersecurity and internet advertising. A cash outflow of nearly RMB 2 billion meant a huge opportunity cost. The zero-yuan transfer was a timely stop-loss.

The stop-loss protected the downside. It did not change the outcome.

In June 2025, the Intermediate People’s Court of Jiaxing, Zhejiang, ruled to accept the bankruptcy reorganization case of Hozon New Energy. In March 2026, Neta Auto founder Fang Yunzhou and former CEO Zhang Yong were listed as dishonest judgment debtors subject to enforcement. The first creditors’ meeting in the bankruptcy reorganization disclosed that the company had only about RMB 15 million left in book funds, confirmed claims of about RMB 5.1 billion, and owed wages and compensation of about RMB 460 million to more than 5,000 employees.

The RMB 1.9 billion Zhou Hongyi invested initially will most likely be fully impaired.

Why NIO could hold on, and Neta could not

Despite similarly huge losses, NIO, XPeng, Li Auto, and Leapmotor all survived. Neta did not.

Financing capability is the obvious answer. The structural difference is a set of survival rules.

Before its IPO, NIO completed at least five rounds of financing, raising about RMB 15 billion, with investors including Tencent, Baidu, JD.com, Hillhouse Capital, and Sequoia China. Li Auto completed nine rounds before its IPO, raising more than RMB 12 billion, with investors including Ming Capital, Matrix Partners China, Wang Xing, Meituan, and ByteDance. Hozon New Energy, the parent of Neta Auto, completed 10 rounds since 2017, raising RMB 22.844 billion in total.

By total financing, Neta was not shabby. RMB 22.8 billion exceeded the pre-IPO financing scale of Li Auto and NIO.

The difference lies in the nature of the financing. Pre-IPO private financing is one-time. Once the money is spent, the company must start a new round of financing negotiations. Each round carries the risk of a down valuation, tighter terms, or even investors refusing to continue. Post-IPO public-market financing provides a continuous funding channel. The company can issue additional shares, issue bonds, and use secondary-market liquidity for more flexible capital operations.

For an unlisted automaker, cash is fixed. For a listed automaker, cash is renewable.

This difference is not obvious during an industry upcycle. Money in the primary market is easy to get. During a downturn, the difference is fatal. After 2023, the logic of the capital market shifted from “betting on who can come out on top” to “money only goes to the leaders.” Primary-market capital closed its door to second-tier new forces. For automakers without a listing channel, there was almost no other way to obtain funds in the cold winter.

Neta’s predicament is a footnote. In June 2024, Hozon New Energy submitted a listing application to the Hong Kong Stock Exchange. The Hong Kong IPO window did not open at the critical time, and the listing process made no substantive progress. In February 2025, Neta Auto said it was conducting a Series E financing, with a planned scale of about RMB 4 billion. By May 2025, when Neta Auto was petitioned for bankruptcy review, the financing had still not arrived.

Listing is not a cure-all. Without listing, you do not even qualify to take the medicine.

NIO is the counterexample. At the end of 2019, after the ES8 spontaneous combustion controversy escalated, NIO’s share price once fell to $1.4 per share, and its book cash flow could only last five months. Li Bin was named by the media “the most miserable person in the auto industry in 2019.” NIO was already listed. It could continue to raise funds through the public market, bring in strategic investors, and maintain a presence in the secondary market until a turnaround appeared. In 2023, when NIO lost tens of billions and once again fell into a trough, it secured a strategic investment from Abu Dhabi investment institution CYVN.

NIO lost countless rounds. It won one round and came back. The mechanism produced that outcome.

The essence of listing: a license to make mistakes

The auto industry has a cruel feature: between input and output lies a long “valley of death.” Factory construction, technology R&D, channel rollout, and delivery operations each require continuous capital injection. Revenue may only appear after products are delivered at scale. Toyota took nearly 20 years from its founding in 1937 to achieving stable profitability. Hyundai Motor started by assembling Ford models in 1967 and remained in losses for nearly eight years until it launched its own model in 1975.

In this “valley of death,” what a company needs is not a single large financing but a channel through which it can continuously obtain funds. Under the Japanese and South Korean chaebol systems, auto businesses could obtain long-term funding support from group internal banks and cross-shareholdings. This was an institutional fault-tolerance mechanism.

Chinese new-energy automakers do not have such a chaebol system. They rely on the capital market. The fault-tolerance mechanism provided by the capital market has a key feature: it is open, but the threshold is listing.

For listed automakers, fault tolerance manifests at three levels. First, they can continue operating while losing money. The capital market allows a company to pay for “today’s losses” with “future profits,” as long as the story and growth remain. Second, they can make mistakes repeatedly. Product iteration failures, strategic missteps, supply-chain pitfalls: as long as credit is not completely destroyed, the capital market usually gives a second or third chance. Third, they can endure cycles. When the industry winter comes and primary-market financing closes, the secondary market still exists, and companies can raise funds through additional issuance, convertible bonds, and other means.

For unlisted automakers, none of the above three holds. Cash flow is limited, the chance to trial and error is one-time, and there is no buffer when the cycle turns. One product mistake, one supply-chain pitfall, one cold market, and the cash flow chain may break. There is no second chance.

The separation between Neta and NIO, XPeng, and Li Auto came down to one thing: whether there was an institutional fault-tolerance mechanism to absorb the cost of trial and error.

What Zhou Hongyi lacked

Zhou Hongyi did not pick the wrong track, and he understood new energy. He lacked the ability and timing to push Neta to list.

Neta did try. The June 2024 submission of a Hong Kong IPO prospectus showed that both the management team and investors understood that listing was the only way out. The Hong Kong IPO window in the second half of 2024 did not open for second-tier new forces like Neta. The logic of the capital market had already changed. Money only went to the leaders, and Neta was not among them.

Zhou Hongyi stopped further investment in 2022 and chose the route of “founding team leadership.” That decision was financially rational. Strategically, it meant 360 gave up the possibility of deep intervention through capital power and forcefully pushing the company toward listing. He did not join forces with multiple capital parties or bind with partners possessing top-level resources to forcibly push Neta to list.

Zhou Hongyi did not create that “fault-tolerance card” for Neta.

This cannot be entirely blamed on him. Pushing an automaker that had lost RMB 18.3 billion to list required not only capital but also a combination of industry cycle, policy window, and capital-market sentiment. Those conditions were not present in 2024 and 2025.

The case shows how elimination races work. In the elimination race of new-energy vehicles, the answer to “who can live longer” is often half-written before the race even begins. The core question on that half of the exam paper is: Do you have an institutional arrangement that allows the company to make mistakes continuously, lose money for a long time, and endure cycles?

If yes, you qualify to wait for the inflection point. If no, you are out.

From betting on hardware to managing software

Zhou Hongyi’s pivot is a response to the above logic.

The new direction he gave was AI and embodied intelligence, and he proposed the concept of “silicon-based leadership”: “In the future, everyone will have a bunch of Agents to manage. These capabilities require you not only to operate Agents but also to build Agents and help enterprises build knowledge bases.”

From a financial perspective, vehicle manufacturing is a typical heavy-asset, long-cycle industry with extremely high cash-flow requirements. As an internet security company, 360’s core strength lies in software algorithms, not hardware manufacturing and supply-chain management. Continuing to invest in this track would not help both companies and would keep dragging down the main business. Pulling resources out of the “bottomless pit” of vehicle manufacturing and reinvesting them in AI software and security services, which have higher gross margins and decreasing marginal costs, was an inevitable choice to improve profitability.

From a strategic perspective, Zhou Hongyi’s pivot reflects a broader trend: tech companies are shifting from “getting into carmaking themselves” to “becoming intelligent suppliers.” Compared with bearing the heavy-asset risks of vehicle manufacturing, providing solutions such as smart cockpits, connected-car security, and autonomous-driving algorithms yields more certain commercial returns and is closer to 360’s capability boundary.

The fault tolerance of AI and embodied intelligence is not necessarily higher than that of new-energy vehicles. The large-model track is currently in a stage of high R&D investment, the commercialization loop has not been fully closed, and the maturity of the Agent tools and services still needs time to be verified. For 360, the cost structure of trial and error in this track is different. Software mistakes can be quickly iterated and corrected, while hardware mistakes often mean chain reactions across production lines, inventory, and channels.

For 360, the better fit was a structure where software errors can be corrected faster than hardware errors.

Epilogue

Neta’s story is not over. In September 2026, the fourth creditors’ meeting of Hozon New Energy’s bankruptcy reorganization disclosed that Zhejiang Taiyi Shenglian Enterprise Management Partnership planned to invest RMB 3 billion to take over and acquire about 70.62% of Hozon New Energy. Whether Neta can “revive” is still unknown.

Zhou Hongyi has turned the page. His next sentence at the summit was about AI Agents. The reporter did not ask about Neta again.