Investing 37 billion but sighing about "leaking water": What exactly are Ma Huateng and the giants afraid of?

"It turned out that a year ago, we thought we'd boarded a ship, only to find it was leaking. Now it feels like we're on it, but we still can't sit down comfortably. We still hope the ship can pick up speed."

At the shareholders' meeting on May 13th, Ma Huateng used this metaphor to describe Tencent's recent AI progress.

The first-quarter report released on the same day showed that Tencent's capital expenditure in the first quarter was 37 billion yuan, mainly used to support AI-related investments.

On one hand, there are the expensive "tickets" bought with heavy investment. On the other hand, there is a bumpy and unclear business direction. Tech giants are experiencing the real pains after boarding the new "ship."

The cruel part of the technology industry is that such moments keep repeating. If we rewind a few years, another giant ship labeled as the gateway to the "next-generation Internet" - the metaverse - also triggered a panic rush of entry on a similar scale.

However, looking back now, that grand departure didn't lead to the promised land. Instead, it was a voyage that ebbed early.

In January 2026, the Internet world presented a strange puzzle: On one side, WeChat, regarded as a relic of the "classical Internet," quietly celebrated its 15th birthday and has long become an infrastructure in daily life, just like water and electricity. On the other side, the flagship metaverse product once labeled as the "next-generation Internet" announced its end at this time.

On January 16th, Meta quietly posted a notice on the homepage of its virtual workspace, Horizon Workrooms: "As of February 16th, 2026, Workrooms will no longer provide services." Due to the already low traffic, this farewell hardly attracted any attention. It wasn't until a few days later when The Verge reported the news that people realized that the metaverse product highly promoted by Meta was exiting the stage of history.

Compared with the fanfare at the opening, this ending was particularly desolate. Looking back to October 2021, Mark Zuckerberg announced at the Facebook Connect conference that the company was changing its name to Meta and declared with high spirits that it would fully enter (all in) the metaverse. Subsequently, Meta launched products such as Horizon Workrooms, Horizon Worlds, and Horizon Venues in one go, trying to make a comprehensive layout in areas such as office, entertainment, and culture and sports, vowing to build a new metaverse world.

As one of the core products, Horizon Workrooms was given high hopes. Meta not only hoped it would become a new-generation meeting tool to replace Zoom but also tried to cultivate it as an important fulcrum for its metaverse strategy and invested a large amount of resources in it. Conservatively estimated, from 2022 to 2024, Meta's investment in this project should be over one billion US dollars.

However, despite the huge investment, Workrooms' performance has never met expectations. On the contrary, this product showed obvious congenital deficiencies from the very beginning. When it was first launched, limited by computing resources and motion capture capabilities, Meta directly omitted the lower body in the virtual character design. So, some people joked, "Put on a heavy VR headset and discuss the company's plan with a group of 'legless monsters'." Even after the form was later completed, the grotesque appearance and stiff expressions of the characters were still widely criticized. Discussing serious matters among such strangely shaped characters was bound to be off-putting.

These experience defects have made it difficult to expand the user base of Workrooms. Even at its peak, the total number of users was only about 300,000, far from comparable to Zoom. In this sense, Workrooms had actually lost its vitality at the product level, and this shutdown was just Meta's official confirmation of this result.

Workrooms is not the only failure. Other products launched around the same time also performed poorly. Horizon Venues was quickly shut down due to insufficient traffic, and its main functions were incorporated into Horizon Worlds. Although Horizon Worlds is still in operation, the number of active users has dropped to more than 100,000. It can be said that Meta suffered a systematic defeat in its metaverse layout.

As a pioneer in the metaverse wave, Meta's failure has also become an important symbol of the metaverse concept's second bankruptcy (the reason for saying "second" is that it had already experienced a round of failure before, which will be explained later). So, why did this once-popular concept ultimately fail to escape the fate of "rising abruptly and falling suddenly"? Where exactly did companies like Meta go wrong? What experiences and legacies has the metaverse left for us? Is there still a possibility for it to make a comeback in the future? Let's analyze these questions one by one.

The Paradox of the Metaverse

Since the science fiction writer Neal Stephenson proposed the concept of the "metaverse" in his novel Snow Crash in 1992, building a virtual space parallel to the real world has always been a dream in people's hearts, especially among the geek community.

To achieve this vision, people in different eras have continuously tried to build virtual worlds with the most cutting-edge technologies of the time. As early as the early days of the popularization of personal computers, various simulation games began to appear, trying to replicate the trajectory of life and the logic of social operation. After entering the Internet era, massively multiplayer online role-playing games (MMORPGs) quickly emerged, highly simulating life, occupations, and even economic systems in the virtual space.

Take Second Life, a representative of such simulation games, as an example. In the game, players can not only work and live, get married and have children, and travel around the world in the virtual world but also successfully establish a connection with the real world, enabling people to earn real money through transactions, auctions, and providing services. As early as 2007, a player named Anshe Chung became a millionaire by engaging in virtual real estate transactions in Second Life.

Due to its high restoration of the real social structure and its real economic functions, many well-known companies, including IBM, once set up virtual office spaces in it and required employees to work there. Even in September 2007, IBM employees launched a strike in Second Life to protest against the company's salary cut.

In terms of the integrity of the virtual world and the degree of connection with reality, Second Life was undoubtedly a successful attempt. However, its development trajectory did not continue this glory. After a brief period of popularity, this virtual society quickly declined, and a large number of players gradually returned to real life.

In subsequent research, some scholars interviewed players who quit the game to understand the reasons for their departure. The response of one player was quite representative: "I feel that Second Life is becoming more and more like real life. It's already tiring to live one life. Why do I need to live another?"

This response reveals the core paradox in the process of building the metaverse. In essence, the metaverse is not only a technical project but also a complex social project. To create a sense of realism, designers must reconstruct social relations, organizational structures, and institutional systems with a high degree of refinement. However, the real society itself is extremely complex. Once any systematic design has a deviation, it may lead to overall disorder.

In this context, the safest option is often to directly copy and simulate the rules of the real society. But the problem is that when the virtual world becomes more and more similar to the real world, the meaning of its independent existence will gradually disappear. Moreover, under the current technological conditions, the digital system's simulation of reality often only presents a version with limited functions and a degraded experience, which will further weaken users' willingness to participate.

If we look back at the new wave of metaverse enthusiasm that emerged in 2022 with this logic, whether it's Meta's Horizon series or the once-popular Decentraland and Sandbox, their paths are almost the same. These platforms try to build a complex virtual social system relying on VR, blockchain, and 3D modeling technologies but still fall into the same dilemma as Second Life: If they try to completely reconstruct social rules, their technology and governance capabilities are still difficult to support; if they choose to copy the real structure, they are difficult to surpass the real world in terms of experience.

Before this paradox can be effectively solved, it will still face insurmountable structural obstacles for the metaverse to truly mature.

Where Did Workrooms Fail?

If the above paradox restricts the realization of the metaverse at the overall level, then at the product level, the reasons for its failure are more specific and clear. Taking Horizon Workrooms as an example, its dilemma does not stem from a single functional defect but is a concentrated manifestation of a series of systematic mismatches.

First of all, the experience brought by Horizon Workrooms does not match its design goal. Meta tried to create a stronger sense of immersion through VR technology to improve communication and collaboration efficiency. However, practice has proved that the sense of immersion has not been converted into efficiency. Instead, it has magnified the physical and cognitive burden. The discomfort caused by wearing the headset for a long time and the structural misalignment between the perception system and the virtual environment make users need to spend extra energy to "maintain presence" instead of focusing on the discussion itself. Coupled with the stiffness and strangeness of the digital avatar's image, this high-stimulus but low-meaning environment constantly distracts users' attention.

More importantly, the virtual space has not reduced the communication cost. Instead, it has introduced new frictions. Social skills that are automatically completed in reality, such as operating the avatar, judging the speaking opportunity, and reading emotional feedback, have become tasks that need to be deliberately learned in the virtual environment. Workrooms has not eliminated the "noise" in communication but has transformed the implicit interaction ability into an explicit interaction burden.

Secondly, Meta overestimated the organizational ability of the space itself when designing Workrooms. It implicitly assumes that as long as a realistic enough shared space is provided, collaboration will naturally migrate into it. However, the operation of real organizations never only relies on formal meetings but more on the trust structure, informal information flow, and stable expectations accumulated over a long time. These relationship dimensions will not be automatically generated because of the virtualization of the space. Instead, they are often weakened. Digital avatars can simulate the body but are difficult to reproduce the social relations in reality; virtual meeting rooms can copy the layout but cannot truly replicate the on-site atmosphere.

Thirdly, Workrooms also faces significant adjustment cost problems. For enterprises, this is not a low-cost attempt but a systematic investment involving hardware procurement, employee training, and process reconstruction, and its efficiency benefits are always difficult to quantify. Virtual meetings have neither significantly shortened the decision-making cycle nor significantly improved the quality of collaboration. In the absence of a clear return on investment, it is difficult to upgrade from an experimental tool to an infrastructure. When enterprises start to cut non-core investments, such high-cost and low-certainty projects are often the first to be abandoned.

Finally, Meta underestimated the impact of virtual office on the organizational institutional boundaries. Embedding organizational activities in a space controlled by the platform means that data ownership, rule setting, and visibility distribution all need to be mediated by the platform.

This is not just a tool selection problem but a problem of the transfer of governance rights and sovereignty. When virtual office is envisioned as a normal structure, these conflicts will inevitably surface. Enterprises can use cloud services and collaboration software but are difficult to accept embedding the operation of the organization itself in a space defined by an external platform.

Looking at these factors together, the failure of Workrooms is not accidental but the result of a systematic mismatch: It requires individuals to bear higher experience costs but fails to provide equivalent efficiency returns; it tries to carry the organizational structure but lacks the mechanism to generate trust and institutional stability; it hopes to become an infrastructure but cannot overcome the economic and governance uncertainties. These problems reinforce each other, ultimately keeping virtual office at the conceptual level and difficult to enter daily practice.

Five Wrong Assumptions of the Metaverse

It should be noted that the problems exposed in the product design of Horizon Workrooms are not an isolated case. If we compare the metaverse products in recent years, we will find that their mistakes in design concepts are highly consistent. This shows that the repeated setbacks of the metaverse are not so much due to implementation-level mistakes as they are rooted in a set of mutually reinforcing but overall distorted cognitive assumptions.

The first wrong assumption is to equate "immersion" with "realism." In the technological narrative of the metaverse, immersion is often regarded as the only way to achieve authenticity, as if as long as the visual, auditory, and even motion capture are realistic enough, people will naturally have a sense of presence and restore the quality of interaction in reality. However, this reasoning ignores a basic fact: Realism does not come from the intensity of sensory stimulation but from the stability of the situation and the continuity of meaning. The "real" in reality is not because we see more clearly but because we know who the other person is, what the relationship is, and what the consequences of our actions will be. The metaverse can magnify the senses but is difficult to automatically generate this social context. As a result, the stronger the immersion, the emptier the meaning.

The second wrong assumption is to mistake "connection" for "relationship." Metaverse products generally emphasize the density of connection: more people online at the same time, more diverse interaction methods, and a more three-dimensional coexistence space, which almost constitute their common pursuit. However, in reality, social relations are not a function of connection but a function of time. Trust, tacit understanding, and a sense of responsibility come from stable expectations in long-term interactions, rather than high-intensity synchronous coexistence. Taking Workrooms as an example, what it provides is more of a form of instant coexistence but lacks a mechanism to transform this coexistence into relationship accumulation. Connections are constantly created, but relationships are difficult to precipitate, and collaboration becomes more fragile.

The third wrong assumption is that "digitalizing space" can naturally lead to "digitalizing society." The metaverse largely understands society as the result of spatial organization, as if as long as the space is redesigned, social behavior will change accordingly. However, modern society is no longer an organizational form centered on space. Enterprises and institutions rely more on rules, processes, and role division rather than physical coexistence. Virtualizing the office space will not automatically reconstruct these abstract structures. Instead, it may cover up the process problems that really need to be redesigned in remote collaboration because of the "return to the same space" in form.

The fourth wrong assumption is to equate "virtual identity" with "social identity." In many metaverse products, digital avatars are given high symbolic meaning, as if they can become substitutes for real identities. However, social identity has never been a visual presentation but a set of rights, responsibilities, and expectations recognized by others, which are based on reputation, relationship networks, and institutional certifications. Virtual identities can change the appearance but cannot compress these social processes. As a result, they either become decorative interfaces or are difficult to assume the stable identity function required for organizational operation.

The fifth and most fundamental wrong assumption is that the platform's technical architecture can replace the institutional system of the real society. Many metaverse products not only hope to provide virtual spaces but also try to organize interactions and allocate resources through platform rules, algorithms, and interfaces. However, real institutions do not originate from overall design but are formed in long-term evolution to stabilize expectations, restrict power, and accumulate trust. Platform rules may be able to improve coordination efficiency in the short term, but once they touch the core areas of organization and governance, they will expose their vulnerability due to the lack of social embeddedness.

Putting these five assumptions together, we can see the overall outline of the metaverse narrative: It tries to make up for social complexity with technological intensity, replace institutional evolution with space design, and cover long-term relationships with instant connections. This idea seems elegant at the engineering level but is highly fragile at the social level. It is in these cognitive misalignments that metaverse products, even with a large number of new technologies, still find it difficult to truly solve users' real pain points.

Reflecting on the Failure of the Metaverse from the Success of WeChat

Almost at the same time when Horizon Workrooms announced its premature end, an application regarded by metaverse enthusiasts as a product of the previous era was celebrating its 15th birthday. This product is the well-known WeChat. Although in the public narrative, WeChat has never claimed to be a metaverse product, from a functional perspective, it has actually built a digital world in a certain sense: It has a stable digital identity system, uses relationship networks as the basic organizational unit, and undertakes communication, collaboration, transactions, content dissemination, and social interactions, gradually forming a complete economic and service ecosystem. Users work, socialize, consume, and obtain information in it, and its importance has long exceeded the scope of a tool application. If we use "a digital space that bears social activities" as a broad definition, then WeChat can also be regarded as a digital parallel system of the real society.

If the failure of metaverse products exposes the limits of a technological utopia, then the success of WeChat shows a completely different technological path. The difference between the two does not lie in who has more advanced technology or who has a larger capital investment but in their fundamentally different ways of understanding the "relationship between technology and society."

First of all, WeChat's growth path follows a continuously evolving product logic rather than a top-down overall design. It initially only solved the most basic instant communication needs and then continuously expanded its functions under the impetus of user behavior and market scenarios. Group communication was to meet the needs of organizational collaboration, voice and video were to adapt to mobile scenarios, the payment system was to meet the demand for digital transactions, and the mini-program ecosystem was to further carry services and business activities. This series of expansions did not originate from a pre-planned "virtual social blueprint" but gradually formed a complex system through continuous trial and error and adjustment. In this sense, WeChat did not try to build a complete world at once but let the world grow naturally under the traction of real needs.

Secondly, We