$14 Billion for One Person, Meta's Market Value Surges by Over $200 Billion: If Tencent Steps Into Manus, Can It Replicate the Next Muse?

In June 2025, Mark Zuckerberg made a move that raised collective eyebrows on Wall Street: he acquired a 49% non-voting stake in data labeling firm Scale AI for approximately $14.3 billion, and brought its 28-year-old founder Alexandr Wang on board to head the newly established "Superintelligence Lab".

The announcement was met with widespread skepticism — "Spending $14 billion just to buy one person?" "How can a data labeling company be worth $29 billion?"

Yet over a year later, those doubts were shattered by a set of figures. As of September 23, 2026, Meta's total market capitalization stood at around $1.94 trillion, representing a cumulative increase of over $200 billion since the close of trading on September 5, the day before Muse was launched. The consideration Meta paid for the 49% stake in Scale AI was roughly $14.3 billion — even by conservative estimates, the market value return from this transaction has exceeded 10 times.

From "No AI Products" to the First Tier of the AI Race

Prior to Muse, Meta had long lacked a competitive product in the AI track. It had Llama — the pioneer of open-source large models, but the fiasco of Llama 4 and the subsequent stagnation in its development left Meta gradually falling behind in the 2025-2026 AI competition. Google had Gemini and DeepMind, OpenAI had ChatGPT, Anthropic had Claude, and it was not until the arrival of Muse that Meta finally secured a product capable of breaking into the first tier of the AI industry.

The turning point came on September 8, 2026, when Meta officially launched Muse, its personal AI agent application. According to Meta, this is a "digital assistant" that runs on an independent cloud virtual machine — each user has their own Secure VM, equipped with a separate browser, file system and terminal. With user authorization, Muse can open web pages, fill out forms, process emails, complete shopping and make reservations. As Meta explained, Muse can run tasks continuously, and will keep executing in the background even if the user closes the application.

The market reaction was nothing short of frenzied. Data from Sensor Tower showed that Muse exceeded 2.5 million downloads roughly two weeks after its launch; Apptopia figures indicated that it reached 2.8 million global downloads within 12 days, topping the overall free app chart on the U.S. App Store and pushing ChatGPT, which had long held the top spot, down to second place. On September 19, Muse set a new single-day download record of 264,000 in the U.S. market.

The capital market responded even more dramatically. On September 21, Meta's stock price surged 11.43% in a single day to close at $741.25, marking its largest one-day gain since April 2025 and adding roughly $192 billion to its market capitalization in a single day. Mark Zuckerberg's personal net worth increased by about $25 billion that day, reaching $253.6 billion. As of September 23, Meta's share price rose further to $761, with a total market capitalization of around $1.94 trillion.

Analysts at JPMorgan Chase believe that Meta's social ecosystem covering 4 billion users is a "significant competitive advantage", and judge that Meta is in the early stage of rolling out AI products to consumers, with "significant upside potential". Those previous voices questioning Mark Zuckerberg's capital allocation capacity have all fallen silent in the face of Muse's impressive performance.

The Underrated "Shovels Seller" and the "Most Valuable Person"

Looking back, Mark Zuckerberg's logic was in fact very clear.

Scale AI is engaged in the least glamorous yet most indispensable work in the AI industry — data labeling. Large models require massive amounts of high-quality labeled data for training, and these labeling tasks involve professionals such as PhDs, lawyers and engineers. Scale AI's clients once covered almost all leading AI models including Google's Gemini and OpenAI's ChatGPT.

More importantly, what Mark Zuckerberg wanted was not just Scale AI's data capabilities, but Alexandr Wang himself. Wang's resume is a Silicon Valley legend: his parents are immigrants who worked at Los Alamos National Laboratory; he dropped out of the Massachusetts Institute of Technology at 19, founded Scale AI after receiving a $120,000 investment from Y Combinator; and became the world's youngest self-made billionaire at 24. Under this transaction, the value corresponding to Wang's personal shareholding is approximately $5 billion.

What Mark Zuckerberg bought for roughly $14.3 billion was not just a 49% stake, but a person capable of redefining Meta's strategic direction in the AI competition. Subsequent facts proved that Muse's product logic — "personal superintelligence", moving beyond the chat box to an independent Agent, connecting emails, calendars, payments and shopping — was almost entirely driven by the Superintelligence Lab led by Wang.

As Ram Ahluwalia put it, Mark Zuckerberg replaced Turing Award winner Yann LeCun with Wang, "It was not an easy decision for Mark Zuckerberg, because it was tantamount to admitting that he had made a wrong call. But that's life. Instead of trying to manipulate public opinion like most companies would do, Mark Zuckerberg chose to solve the problem."

China's Version of "Acqui-Hire": Tencent's Manus Wager

Turning to China, a highly similar scenario is unfolding at Tencent.

Manus — the company known as the "world's first general-purpose AI agent" — rose to overnight fame in March 2025 when its invitation code was speculated to sell for as high as 100,000 yuan. In December 2025, Meta completed a lightning acquisition of Manus for $2 billion in just 10 days. However, by April 2026, the transaction failed to close due to regulatory approval procedures.

According to reports from multiple media outlets, in July 2026, Tencent was leading a Chinese capital consortium, planning to repurchase 100% equity of Manus from Meta at a valuation of roughly $2 billion. A person familiar with the matter revealed to the media that Tencent will become the largest single shareholder while remaining a minority shareholder without holding controlling stakes, and Manus will continue to operate as an independent company based in Singapore. The repurchase price is on a par with Meta's original acquisition price, but at this point Manus' ARR has climbed from approximately $100 million at the time of acquisition to the range of $400 million to $500 million. Citing people with knowledge of the matter, The Information reported that calculated at $2 billion, the PS ratio has dropped from 20 times to 4 to 5 times.

Nevertheless, a considerable portion of this revenue growth came from channel diversion from Meta's advertising system, and it remains uncertain whether this momentum can be maintained after breaking away from the Meta ecosystem. In addition, reports say Manus is pushing for a new round of financing at a $4 billion valuation after the repurchase is completed, with potential investors including IDG Capital and Boyu Capital, while existing shareholders Tencent, HSG and ZhenFund all intend to participate in the follow-on investment, and the company is also evaluating the possibility of a Hong Kong IPO. None of the above transaction details have been officially confirmed by Tencent or Manus.

What Is Tencent Buying?

If Meta's Muse has proven the formula of "top talent + right direction = market capitalization multiplication", then the logic behind Tencent's acquisition of Manus has a clear reference frame.

Tencent has long adhered to the principle of "investing only, not acquiring" in the AI field, and has participated in investments in firms such as Moonshot AI, Zhipu AI and Baichuan Intelligence, reaping considerable returns from its investments in Zhipu AI and Enflame Technology. However, a clear gap remains: in the AI application layer closest to users, Tencent has never used a key acquisition to lock in the general Agent entry the way Meta did when it bought Scale AI.

Tencent is not without existing layouts. WorkBuddy ranks first in China's domestic PC-side AI-native office agent market with 20.97 million monthly visits, but its strength lies in "working on your own computer"; Manus follows the route of "running errands for you on the cloud", with tasks running in independent virtual machines that can continue executing even when the device is shut down. The two are complementary rather than substitutive. If Tencent takes over Manus, it will fill in the missing puzzle piece of "general-purpose task execution".

There are also risks. The seven months when Manus saw its fastest revenue growth was exactly the period when it operated as "part of Meta", and it is doubtful whether its ARR can maintain growth after losing Meta's traffic pipeline; in addition, it uses the Claude model, with a cost of about $2 per task and a gross margin of only roughly 50%.

Yet these challenges do not change the strategic judgment. Tencent has taken its Hunyuan large model as the foundation, connected scenarios including WorkBuddy, Yuanbao, ima and Tencent Docs, and formed an Agent matrix covering office work, conversation and knowledge management. The addition of Manus will give Tencent the ability to independently complete cross-application complex tasks on the cloud for the first time — this is a key step for Tencent's AI narrative to evolve from a "tool matrix" to a "general-purpose agent platform".

The Math of a Deal Centered on "People"

Going back to the controversies mentioned at the beginning, those who criticized Mark Zuckerberg for "spending $14 billion just to buy one person" confused a key point: in the AI era, top-tier talent itself is the most scarce asset.

Google DeepMind reached a talent recruitment and technology licensing agreement with Contextual AI for approximately $100 million, bringing in more than 20 researchers without involving equity acquisition; earlier, it "absorbed" the core team of Windsurf with a $2.4 billion license fee, hiring CEO Varun Mohan, co-founder Douglas Chen and the core R&D team to join DeepMind, without holding any equity in Windsurf. Elon Musk's SpaceX acquired Anysphere (the parent company of Cursor) for $60 billion in an all-stock deal, which was closed on August 14, 2026, making Cursor a wholly-owned subsidiary of SpaceX.

This is not madness, but sobriety. Meta spent roughly $14.3 billion to acquire Wang, in exchange for over $200 billion in market capitalization growth and a truly representative AI product. If Tencent takes over Manus for $2 billion, it will get a general-purpose Agent entry, a top-tier team, and a chance to replicate the Muse-style market capitalization effect in China's AI application layer.

As for those who questioned Mark Zuckerberg's capital allocation ability by citing the $88 billion loss of Reality Labs, Muse's performance has already given the answer — the failure of the metaverse lies in "going against human nature", while Muse is "betting in line with human nature".

For Tencent, if the Muse story proves the formula of "acquiring top talent + product catalysis = capital recognition", then the rumored Manus transaction is a strategically positioned move for Tencent in its AI narrative with controllable costs and huge potential returns. $2 billion in exchange for an entry that may define the landscape of China's AI application layer — this deal is a no-lose proposition by any calculation.

This article is from the WeChat official account "Tech Business", author: Tech Business, published with authorization from 36Kr.