The Full Story of Alibaba's Investment in Moonshot AI

@Chen Zhiyan

Elsewhere previously covered Kimi's Series C round — the IDG-led deal that stands as one of venture capital's most spectacular bets made just before dawn. But looking back at Kimi's funding story, what truly put Zhilin Yang at the large model table was an earlier, far larger check from Alibaba.

In November 2023, when Alibaba first invested in Kimi, it put in $30 million at a $600 million valuation. Three months later, on Lunar New Year's Eve 2024, Alibaba suddenly raised its bet to $800 million, pushing the post-money valuation past $2.5 billion.

That money made Alibaba Kimi's largest shareholder to date, and instantly turned Kimi into an unexpected player at China's large model table.

Three years on, Kimi's valuation has surged to $50 billion. But inside Alibaba at the time, the decision was contentious. Some called it a "PowerPoint company" — when Alibaba first met the team, there wasn't even a model yet; after internal technical due diligence, there was a clear conclusion: don't invest.

But Alibaba ultimately did invest, and it became the only large model company backed by Alibaba Group's strategic investment arm, with six consecutive follow-on rounds. Barring surprises, this will also be one of Alibaba's most lucrative investments on paper in the AI era.

Investors love to say: bet on the game changer. Conversely, investment itself sometimes plays a similar role. And Alibaba's bet on Kimi was precisely that — a capital allocation that changed the game.

A Near Miss

Kimi and Alibaba nearly missed each other.

The year was 2023. Regulatory aftershocks still lingered, and like other internet CVCs, Alibaba Group's strategic investment arm had scaled back considerably, with remaining deals mostly tied to retail and e-commerce. Large model investments were explicitly handed to Alibaba Cloud.

The logic was straightforward at the time. Large models were first and foremost Alibaba Cloud's business: invest in a company, hold equity in the left hand, gain a customer in the right.

So by year-end, Alibaba Cloud as the investment vehicle had already deployed capital into Zhipu AI, Baichuan, and 01.AI. By March 2024, it added MiniMax.

Alibaba Cloud took the safest approach, which in retrospect also looked like the smartest: accumulate multiple options. But even so, Kimi was not among them.

Alibaba Cloud had looked at Kimi early on, but the company was still at the PowerPoint stage. For a company that would need to continuously purchase compute with no visible product or revenue in the near term, this deal was difficult to justify through the cloud business's investment framework. Ultimately, Alibaba Cloud passed.

As the absolute main force in AI investing, with Alibaba Cloud declining Kimi, Alibaba missing out on Kimi early on had become all but certain.

Then a woman named Jiang Shanshan appeared.

As an investor in the strategic investment department, she had already spent five years within the Alibaba ecosystem. She had tried adding Zhilin Yang on WeChat several times without success, until she enlisted the help of Song Yao, a fellow Tsinghua-affiliated founder, to make the connection.

Several AI founders have mentioned to me that Jiang Shanshan was one of the few investors who could actually talk technical details. She studied mathematics and computer science for her undergraduate and graduate degrees, with her graduate research focused on optimization algorithms. And according to those who have worked with her, she carries a cool, reserved demeanor, has been vegetarian for years, and wins people over through professional competence rather than people skills.

After speaking with Zhilin Yang, she pitched the project to Hu Xiao — then head of Alibaba Group's investment department. After Hu Xiao's meeting, he recommended it to Eddie Wu. Following Eddie Wu's conversation with Zhilin Yang, there was virtually no hesitation: he decided to invest.

In November 2023, Alibaba's strategic investment arm, alongside Ant Group, each committed $30 million to Kimi at roughly a $600 million valuation.

The Origin of $800 Million

Compared to what followed, this was merely an appetizer.

The investment emerged within a historical inflection point. In September 2023, Alibaba completed a leadership transition: Eddie Wu took over as Group CEO. Upon returning to Alibaba, he reopened a window for the group's strategic investment arm: beyond serving existing businesses, it could also look at frontier technologies like AI and Embodied Artificial Intelligence.

This was the opening that allowed Alibaba's strategic investment arm to invest in Kimi.

$30 million was enough to get Alibaba on the shareholder register, but not enough to change anything. For a large model company, tens of millions or even hundreds of millions burn quickly.

During the closing process for this investment, on October 9, 2023, Kimi announced a breakthrough in "long context" capabilities, launching Kimi Chat — the first intelligent assistant product supporting inputs of up to 200,000 Chinese characters. This was Kimi's first genuine viral moment.

This set the stage for Alibaba's second round. As Elsewhere understands, model performance was the primary factor in Alibaba's willingness to continue investing in Kimi. According to multiple sources close to the deal, the direct push to raise the amount to $800 million came mainly from Eddie Wu's own judgment — a bet he personally greenlit.

This investment inclination toward "high uncertainty + technical genius + frontier technology" certainly had something to do with Eddie Wu's experience founding Yuanjing Capital in 2015 and fully immersing himself in VC investing. But on another level, a more pragmatic backdrop may have been: Alibaba hadn't fully secured its ticket in the Mobile Internet era, with Pinduoduo and Douyin attacking its e-commerce core.

But for the new leader known internally as "Mama Wu," there was also a subtle question: Alibaba already had Qwen, so why double down on an external model company?

Here's the timing gap: it was still far from the point where Qwen's capabilities would be validated. The earliest that would come was around June 2024 — after the successive releases of the Qwen2 and Qwen2.5 series. The true milestone might not arrive until September 2025, with the release of Qwen3's hybrid reasoning model.

Eddie Wu's message to the organization at the time was that Alibaba could continue building its own model while also investing in the external team most likely to succeed. Self-development was the card in Alibaba's hand; investment preserved another possibility, ensuring Alibaba would still be at the table if that team won.

In other words: we must build, but we can't only bet on ourselves.

And since we are betting, for large models, the stakes must be large enough. This is the origin of the $800 million. That sum could fund three VC firms.

An investor in MiniMax once told me that around this same time, MiniMax was also in negotiations with Alibaba and had felt confident about closing a deal. Until during a break in negotiations, Alibaba's team mentioned they had "another card" — this card referred to Kimi, then still unknown to many.

On February 9, 2024, Lunar New Year's Eve. The signing deadline for Kimi's A+ round was set for this day. Shareholder opinions went back and forth, and signatures weren't completed until around midnight, already into the first day of the Lunar New Year.

An investor close to the deal told me that Yutong Zhang also played an important role during this period — she understood what Kimi's founding team cared about and what investors wanted, helping both sides reach agreement on many specific points.

The final transaction documents included a provision: beyond the $800 million signed, Kimi could take an additional $200 million within the following 60 days. All previous funding Kimi had raised totaled less than $300 million, placing it in the lower-middle tier among the "Six Little Dragons," only above StepFun. Yet this single round alone exceeded $1 billion.

This was also the largest single round raised by a Chinese large model company at the time. After this battle, Kimi's post-money valuation surpassed $2.5 billion.

Delicate Times

Going back to 2023, early-stage large model funding wasn't actually that difficult.

After ChatGPT ignited the frenzy, investors feared missing the next technological wave, throwing in tens of millions for an entry ticket. Several institutions would pool together, each contributing a bit, taking a look at the cards first — what the investment world calls a Club Deal.

The hard part was the middle money.

When a company has produced its first model and its valuation reaches one or several billion dollars, what it needs is no longer an ordinary growth-stage financing round. A single new training run might cost hundreds of millions of dollars, yet investors still cannot determine whether the next model version will succeed, or how long before the company generates sufficient revenue. This was also where Kimi's fundraising process had gotten stuck.

And in recent years, as traditional dollar-denominated growth funds continued to contract and RMB funds were constrained by scale and investment horizons, this round became especially difficult.

Only the tech giants had money to put up.

This was also the most visible change in Alibaba Group's strategic investment arm after Eddie Wu's return: shifting from seeking business synergies around retail and e-commerce to frontier directions like AI, Embodied Artificial Intelligence, brain-computer interfaces, and nuclear fusion; once conviction was established, putting in enough money to directly alter a company's development trajectory.

But quite dramatically, after the $800 million went out, Kimi quickly became embroiled in a web of internal and external conflicts.

First came the controversy over founding team share sales. After the company's valuation reached approximately $1.5 billion, Kimi had provided liquidity to several founding team members through secondary share sales as stipulated in prior investment agreements. As I understand it, the actual amounts were shared among multiple people, with some used to buy apartments in Beijing.

Founding team members partially cashing out mid-journey is not uncommon in startups. But as a high-profile company, this was inevitably magnified.

The fire quickly spread to Alibaba, which had just made its heavy bet.

In mid-2024, Hu Xiao left Alibaba, and Jiang Shanshan was also moved out of her investment role. In some narratives, these were quickly stitched into the same story: the people who pushed this investment also had problems.

As we understand it, after leaving, Hu Xiao joined FirstLight Capital under Yong Zhang (Xiaoyaozi). Jiang Shanshan transferred to the Group Strategy Department for half a year. This was Eddie Wu's newly established CEO Office, where she was selected to work on Alibaba's AI strategy.

In any case, some internal friction was more or less inevitable, especially given the timing around Alibaba's core management transition. According to accounts from some within Alibaba's investment ecosystem, those involved in the Kimi investment were under considerable pressure during that period.

It wasn't until February 2025 that Jiang Shanshan was transferred back to the investment department to lead a new sub-division: the Frontier Technology Investment Department.

But various voices never fully subsided. Later, as Qwen's capabilities and open-source influence gradually rose, a natural question emerged internally at Alibaba: if our own model has caught up, why spend so much investing in someone else?

Yet this investment had faced significant internal controversy from the very beginning. When the group's strategic investment arm pushed for the deal, internal technical assessments had produced evaluations like "not recommended," "PowerPoint company," and "weak team"; whenever Kimi's model performance fell short of expectations or controversies arose, these opposing voices would resurface.

Someone close to Alibaba's strategic investment department told me that it wasn't until this year's K3 received overwhelming positive reception that the various discussions around this matter gradually subsided.

Meituan Longzhu, ByteDance, and Others

At this point, there are a few more details behind this investment between Kimi and Alibaba worth noting.

One involves Meituan Longzhu. The earliest CVC to bet on Kimi was actually not Alibaba, but Meituan Longzhu.

After ChatGPT's debut, Meituan Longzhu's investment team began searching domestically for teams building decoder-only large models, with Zhilin Yang on their priority watch list. But like Jiang Shanshan's initial experience, they also repeatedly tried adding Zhilin Yang on WeChat without success. They pursued him for four months, only connecting after Kimi had completed its first funding round.

The debate at the investment committee was similarly intense. The core tension: large models require massive capital expenditure — would this be only a big tech game, with no room for startups?

At an impasse, Xing Wang had to leave for his next meeting. Before departing, he said: "I do believe the probability of a startup making a super model and super application is very low, but I'm willing to support it."

Ultimately, the Longzhu IC still voted through this contentious project.

The $600 million valuation was set by Meituan Longzhu partner Xinyu Wang and Zhilin Yang over a meal. When Xinyu Wang named the number, Zhilin Yang agreed immediately. This was also the valuation at which Alibaba's strategic investment arm and Ant Group entered in their first round.

In that round, Longzhu, Ant, and Alibaba's strategic investment arm came in at the same price, $30 million each. But Longzhu passed its IC in July, while Ant and Alibaba didn't clear until November. In those four months of difference, Kimi's first model came out.

However, Longzhu did not continue adding to its position as Alibaba did. The main reason was reportedly that the valuation had changed but the company hadn't changed enough. Xinyu Wang later reflected on this, calling it a lesson: "Don't lose your position on the most important projects."

It wasn't until the $6 billion valuation round at the end of 2025 that Longzhu re-entered, and added to every subsequent round. In March 2026, Longzhu even led a $200 million-plus investment at a $20 billion valuation. And Kimi became the project into which Longzhu has deployed the most capital since its founding in 2017.

Another rumor involves ByteDance.

An investor mentioned to Elsewhere a story: when Alibaba was leading the $800 million round, this was precisely when ByteDance most wanted to invest in Kimi. At that time, Doubao had not yet achieved mainstream recognition — "ByteDance was at the moment when it most needed an external model." But in the end, Yiming Zhang decided ByteDance would build its own model. This bears some resemblance to the story of ByteDance abandoning its investment in MiniMax at the last minute.

As of today, though Kimi has numerous investors, there aren't many seats at the board table.

Initially there was only HSG; Meituan Longzhu was added later; then Alibaba and Ant came on — for a period, these two rotated the same seat. After Alibaba's $800 million came through, a new board seat was added, while Ant and Longzhu exited, leaving only HSG and Alibaba as the two external investor seats.

After Alibaba's $800 million financing closed, Kimi's founding team retained super-voting rights, with portions of the investment placed in a co-managed account for purchasing Alibaba Cloud services. Alibaba became Kimi's largest external shareholder, but Kimi remained an independent company.

A recent prediction about the future landscape of large models suggests that China's foundation model market may ultimately be left with "two big players, two small players."

One possibility is that the so-called "two big players" of the future may not be today's big players. Today's small players like Kimi, DeepSeek, and Zhipu AI could also claim one or two of those positions.

But in the process of small players growing into big players, what they precisely need is the kind of money that only today's big players can put up.

This is the most subtle aspect of this investment: an AI startup wanting to become tomorrow's big player often must first take today's big player's money; and today's big players, fearing elimination by the next technological wave, are also compelled to invest in the very future challenger that may displace them.

Recently, various capital-market stories about Kimi have emerged in succession. The stories of its relationships with other investors — Tencent, Monolith, ZhenFund, 5Y Capital, Gaorong Ventures, Huiwen Wang, and others — we will publish in due course.

Elsewhere also understands that Jiang Shanshan left Alibaba's strategic investment department in August and plans to start a new fund.

Cover image: Adolph Menzel, Flute Concert of Frederick the Great at Sanssouci, 1850–1852, Alte Nationalgalerie