Meta buys AI agent Manus – The strategic purchase that is reshaping the AI ​​industry

A quiet takeover boom

On December 30, 2025, Meta announced its acquisition of AI agent provider Manus for more than two billion dollars. The news came as a surprise to many, as reports indicated that negotiations had taken only ten to fourteen days. But this was no ordinary transaction. It was the third-largest acquisition in Meta's history, surpassed only by WhatsApp and the $14.3 billion investment in data annotation specialist Scale AI in June 2025. The speed and scale of the deal reveal a deeper truth about the current dynamics of AI: The window of opportunity for acquiring field-proven agent technology is rapidly closing, and Meta didn't want to be left behind.

The acquisition underscores a crucial phenomenon in the tech industry in 2025. While experts are still debating the definition of generative AI and the limitations of large language models, autonomous AI agents have already established themselves as a fundamentally new product category. Manus is the epitome of this paradigm shift: Instead of simply answering questions, these systems execute complex tasks fully and autonomously. They book flights, write code, analyze stock markets, and create customized travel itineraries—all without human intervention. This isn't about a chatbot with an enhanced user interface, but rather the automation of complex processes controlled by natural language.

Meta's strategic move reflects a technological arms race gripping the tech giants. OpenAI, Google, and Microsoft had already mobilized massive investments in their AI infrastructure, but Meta lagged behind in translating this into practical applications. Sheer computing power and proprietary models like Llama were of little use as long as the company couldn't quickly deliver the right product form. With Manus, Meta acquired more than just software. The company secured proof of a functioning product-market fit in the fastest-growing AI segment—and at a time when other desirable targets like Perplexity were also considered acquisition candidates.

Not just a price or usage problem

The numbers surrounding Manus are breathtaking. The company officially launched in March 2025. Eight months later, at the end of 2025, it reported an annual revenue rate of over $125 million—roughly $100 million in strictly defined subscription revenue. This makes Manus the fastest AI application ever to reach the $100 million mark. For comparison, Cursor, the popular AI-powered code editor, took about eighteen months to reach the same milestone. Manus did it in eight.

The growth rate is as remarkable as it is sustainable. Since the release of Manus 1.5 in October 2025, the application has consistently grown by more than 20 percent per month. At this rate, revenue doubles every four months. This isn't the growth of an application that solves a niche problem. This is the growth of a tool that people use to transform the way they work. To grasp the magnitude of this success, Manus has processed 14.7 trillion tokens in less than a year—an unfathomable measure of actual human computing use.

Manus's commercial drive is also reflected in its pricing. The product operates on a three-tiered subscription model. Basic users pay $19 per month for limited daily tasks. The mid-tier pays $39 for advanced capabilities. The premium tier costs $199 per month—a price only the most dedicated or commercial users would pay. Despite this pricing model, which explicitly discourages intensive use, Manus has built an organic customer base willing to pay significantly. This contradicts a long-held assumption in the AI ​​industry: that users wouldn't pay for specialized AI agents. Manus has proven otherwise.

Manus' technological architecture is both conservative and innovative. The system relies on a multi-model approach, meaning it doesn't build on a single foundation, but on several. Manus uses Anthropic's Claude 3.5 Sonnet model as one of its core tools, combined with Alibaba's Qwen models and an ensemble of techniques drawn from the broad spectrum of modern AI research. This architecture is not a flaw; it is intentional. By blending models from different providers, Manus achieves a robustness and reliability that a single model alone cannot provide. It is a non-trivial feature of an older approach—ensemble methods—combined with the new paradigm of large-scale models.

The ability to plan and execute is its core feature. If a user tells Manus to plan a vacation in Japan in April, the AI ​​automatically breaks this task down into smaller sub-goals: searching for destinations, comparing flight prices, finding accommodations, securing travel visas, budgeting, and generating an itinerary. Each of these steps is executed in an iterative agent loop, where the system checks the result of each step, validates it against the original requirement, and decides whether adjustments are necessary. This is true autonomy—not simulated by a large prompt, but through genuine architectural decisions deep within the system's machinery.

The performance comparisons are also in Manus's favor. In the GAIA benchmark—a standardized test of agent capabilities that drives the real-world AI evaluation business—Manus outperforms OpenAI's Deep Research feature in several categories and difficulty levels. This is no small feat. OpenAI has become one of the most respected AI institutions since it introduced ChatGPT to the world. For a startup that's been around for less than a year to beat these metrics sends a clear signal to the market.

Meta's puzzle in the larger competition

Meta's position in the AI ​​industry is complex. The company has invested immense resources in fundamental research. The Llama model series is the open-source counterpart to OpenAI's GPT family. These models have been downloaded by more than 650 million people, a sign of genuine adoption by developers. Meta has also made available an intelligent assistant product called Meta AI, which has approximately 600 million monthly active users—a sign of market penetration.

But there was a significant gap between the raw material and the finished product. OpenAI had ChatGPT, arguably the most popular AI application of all time. Google had Gemini, its answer to ChatGPT, integrated with its comprehensive services. Microsoft leveraged its partnership with OpenAI and integrated that technology into Microsoft 365 and Azure—a strategy that has earned Microsoft billions in licensing revenue. Meta, on the other hand, had vast computing resources and high-quality models, but only limited, user-unseen applications to offer.

The acquisition of Scale AI in June 2025 was an attempt to close this gap. Meta invested $14.3 billion for a 49 percent stake in Scale AI, the data labeling specialist. The deal came with an additional component: Alexandr Wang, the CEO and founder of Scale AI, left his position to lead a new superintelligence research lab at Meta. This was not a passive acquisition. It was a move to control the data infrastructure that is crucial for modern AI training.

But the Scale deal wouldn't have solved the agent problem. Scale is an infrastructure layer, not a consumer layer. Manus is different. It's a productive end product that people use directly. It's the interface between Meta and the global AI agent market, which is projected to grow to $200 billion to $236 billion by 2034. With a 45 percent annual growth rate, this is the largest market the AI ​​industry has generated.

Even after the Llama-4 model—which Meta released in April 2025, months ahead of OpenAI's GPT-5—it still faces some nuances. While Llama-4 is remarkably powerful, it has been shown to lag behind the best OpenAI and Google models in domains such as reasoning and coding performance. This isn't fatal, but it does mean that Meta can't simply climb to the top through sheer model quality. It needed other avenues—and agents are one where it could have gained an early advantage.

Here, the strategic brilliance of Meta's Manus acquisition becomes particularly clear. Instead of waiting two years to build and refine its own agent technology, Meta acquired an operational application that already had a proven track record. The team was integrated with Meta. The technology was bound to Meta's infrastructure. The customer base—millions of paying users—was reassembled across Meta's platforms, from WhatsApp to Instagram to Facebook. That's the scale of the reach: While other providers are still debating how to get agents to the masses, Meta aims to reach billions of users in hours.

The geopolitical trap

But beneath the surface of this transaction lies a geopolitical tension that intertwines the dynamics of the world in the 2020s. Manus was founded by Xiao Hong, a Chinese entrepreneur with over a decade of experience in various AI and productivity software ventures. Before Manus, Xiao Hong built Monica, a browser plugin agent that acquired 10 million global users. ByteDance attempted to acquire Monica for $30 million—a respectable sum—but Xiao declined.

Then, in June 2025, Manus moved from Beijing to Singapore. This wasn't just a change of location. It was a signal: Manus would become a global product, built for the rest of the world, not for China. The company completely disappeared from the Chinese market. Its core models – the Anthropics Claude – are not available in China. This was a deliberate strategic move.

This makes Meta's acquisition far more geopolitically fascinating. In recent years, the US government (under Biden and now again under Trump) has imposed aggressive export controls on China—on its best semiconductors, advanced chips, and computer equipment. The theory is that delaying Chinese AI development will maintain US dominance in AGI. In March 2025, the US controlled 75 percent of global AI computing capacity. China's share had plummeted from 37.3 percent in 2022 to just 14.1 percent.

But here's the paradox: The best spyware product currently on the market was developed by a Chinese entrepreneur. This fact doesn't directly harm China, because the founder has opted out and subjected his technology to American control. But it reveals something questionable about US strategy: Export controls cannot prevent innovative talent from leaving China. They cannot prevent architectures from being fully developed outside of China. And they cannot prevent American companies from buying up and absorbing these innovations.

Xiao Hong will now report directly to Meta's Chief Operating Officer. His entire team will be integrated with Meta's infrastructure. The geopolitical reality is that China has lost talent, innovation, and in this case, an entire operational function to the United States. This is a win for the US—but it's also a sobering sign that talent and innovation outflows cannot be stopped by national barriers. The race for AI is not simply a competition between systems. It's a race for talent absorption.

Here you will learn how your company can implement customized AI solutions quickly, securely, and without high entry barriers.

A Managed AI Platform is your all-round, worry-free package for artificial intelligence. Instead of dealing with complex technology, expensive infrastructure, and lengthy development processes, you receive a turnkey solution tailored to your needs from a specialized partner – often within a few days.

The key benefits at a glance:

⚡ Fast implementation: From idea to operational application in days, not months. We deliver practical solutions that create immediate value.

🔒 Maximum data security: Your sensitive data remains with you. We guarantee secure and compliant processing without sharing data with third parties.

💸 No financial risk: You only pay for results. High upfront investments in hardware, software, or personnel are completely eliminated.

🎯 Focus on your core business: Concentrate on what you do best. We handle the entire technical implementation, operation, and maintenance of your AI solution.

📈 Future-proof & Scalable: Your AI grows with you. We ensure ongoing optimization and scalability, and flexibly adapt the models to new requirements.

More about it here:

• The Managed AI Solution - Industrial AI Services: The key to competitiveness in the services, industrial and mechanical engineering sectors

The bigger picture: Why agents now

The acquisition of Manus is related to a broader shift in the AI ​​industry. 2024 and the first half of 2025 were obsessed with questions of model capacity, reasoning, and scaling. Companies were asking: Can we build a model that's even smarter? The answer was yes, and all the big companies were still building bigger models. But by mid-2025, a new question became urgent: How do we actually use these models for something useful?

That's the agent question. An agent is an AI system that receives a goal, then autonomously plans and executes to achieve that goal. It's not a chatbot that sits on the side and waits. It's not a workflow engine written for programmers. It's something in between: intelligent autonomy, accessible to ordinary people.

The agent market is projected to grow to $200 billion by 2034, with an average annual growth rate of approximately 45 percent. This is faster than the convergence of the chatbot market, faster than the convergence of traditional cloud software, and even faster than the convergence of the mobile app market in the early 2010s. Agents are seen as the next platform layer, much like mobile smartphones displaced the desktop.

This market scale explains the urgency of Meta's acquisition. If agents prove to be the true platform—and the Manus numbers suggest they could—then early player ownership in this category is non-negotiable. Any other system, any other way to build agents, won't be operating on the same side of the universe. That's why Meta couldn't wait. That's why the purchase was so swift.

The perspective of investments and capital expenditure

To fully grasp the scale of Meta's AI bet, it's necessary to consider its capitalization. Meta will spend between $66 and $72 billion in 2025—double the $39 billion spent the previous year. This sum, which deserves to be repeated several times, exceeds the annual GDP of over 100 countries. It's an unparalleled capital commitment to a single product: artificial intelligence.

Meta's plan is physically colossal. The company is building Titan clusters—gigantic data centers built with thousands of the most powerful graphics cards currently available. The first of these clusters, Prometheus, is being built in Ohio and is expected to go online in 2026 with one gigawatt of computing power—that's 1,000 megawatts, enough to power one hundred million homes. The Hyperion cluster, in Louisiana, will scale up to five gigawatts over several years, a footprint the size of Manhattan.

This isn't speculation on future AI capabilities. This is capital tied up in real bricks, real cables, real graphics cards. It's a test of whether Meta believes AI agents are a real, lasting category or just a passing fad—and if the former, whether the risks of this capitalization are justifiable.

There's another aspect to this investment logic: the talent wars. Meta reports that it's recruiting leading AI researchers from OpenAI, Google, Apple, and other companies. This recruitment comes at a non-trivial price—individual packages are reportedly worth $200 million over four years. That's not a salary in the traditional sense. These are sums that represent an entire lifetime of a knowledge worker's career.

The reason Meta is willing to do this is that technology is moving too fast. If they wait two years to build talent internally, they've lost. OpenAI could have already achieved the next breakthrough. Google could have already built the next breakthrough. Anthropic could have shipped a product that makes theirs obsolete. Within this timeframe, buying talent—through acquisitions of entire companies or massive packages—is the only rational strategy.

The business implications for Meta

What will Meta do with Manus? The company has spoken of running Manus as a standalone service, but also integrating its technology into Meta AI, Facebook, Instagram, WhatsApp, and its smart glasses line. This is the standard integration playbook—the company keeps the external product alive while also incorporating the best parts of it into its internal systems.

The monetization strategy is less clear. Manus currently operates on a subscription model. Meta earns its power through advertising—companies pay Meta to place their ads on social networks. It's unclear whether Meta intends to tie a subscription service so closely to advertising. More likely, Meta will integrate Manus into its existing platforms and then leverage the AI ​​agent's capabilities to better target ad placement and gather more user data to refine ads. An AI agent that understands a user is planning an upcoming trip could lead to more deeply targeted ads.

This is not speculation. Meta has already communicated that AI-powered ad tools are driving revenue growth. In Q2 2025, Meta reported revenue growth of 22 percent, driven in part by AI-powered ad tools. The company uses AI for real-time translation for international users and automated video creation. The agent aspect will enhance these capabilities.

Advertising remains the primary commercial avenue. AI is the tool that makes advertising more focused, relevant, and ultimately more profitable. Manus will help Meta gain a deeper understanding of user intent, thereby enabling better ad placements.

The competitive reshaping

Meta's acquisition of Manus subtly but significantly shifts the balance of power in the competitive landscape. While OpenAI leads in this metric with 800 million weekly ChatGPT users, it hasn't yet widely rolled out its agent capabilities – Deep Research remains a niche product. Google also incorporates agent-based approaches within its major core models, but lacks a standalone product with the functionality of Manus. Microsoft integrates the capabilities of its partner OpenAI, but likewise doesn't offer a standalone agent application that's immediately usable by end users

Meta's acquisition of Manus means it's the first of the major tech companies to have direct control of a proven, massively popular agent application. This gives Meta a timing advantage. It can roll out this technology to billions of users. It can build agent capabilities directly into WhatsApp, directly into Instagram. It can build the ecosystem of agent apps on Meta's platforms, much like the Apple App Store.

This has implications for OpenAI. OpenAI will need to quickly build or acquire an agent product. The window for acquiring the best options is closing. Google will feel pressure to roll out its agent integrations faster. The world of AI agents is one where early adopters reap enormous benefits. Meta wasn't early in models—OpenAI was. But Meta could be early in agents. The acquisition of Manus is an attempt to close that gap.

The risks of betting

But a bet of this magnitude also carries risks for Meta. The first is operational: Can Meta truly integrate Manus without destroying what makes the company successful today? Manus's success relies heavily on its multi-model approach and its iterative agent loop. If Meta intervenes too much to make processes faster or cheaper, it could also destroy the magic behind them. This is a typical mistake large companies make during acquisitions.

The second question is technological: Will the current generation of models—Llama 4, Claude, Gemini—meet the requirements for truly arbitrarily generalized agents? Today, agents can perform specialized tasks well. But can they converge into a single agent that does everything on demand? The capabilities aren't there yet. This means that Meta is still four or five years away from truly realizing its full potential.

Third, there's the economic question: Can the models improve quickly enough to justify the multi-billion dollar expenditures? Meta spends an estimated $70 billion annually on AI infrastructure. Will revenues from improved advertising, intelligent agents, and the licensing of Llama models grow quickly enough to justify this investment? It's possible, but by no means guaranteed. Bain & Company estimates that annual depreciation on AI investments is $40 billion—higher than the revenues currently generated by all the major hyperscalers from AI.

Another, non-technological risk is regulatory in nature. The planned acquisition by Meta could trigger an antitrust review, as the US government is already investigating whether large tech companies are accumulating excessive market power. An acquisition of this magnitude ($2 billion) in a growth sector like AI agents would inevitably bring this issue to the attention of the authorities.

The larger shift

Meta's acquisition of Manus is not an isolated case. Rather, it is symptomatic of a broader shift in the tech industry. Large companies that previously relied on organic growth and internal research and development (R&D) are now turning to aggressive mergers and acquisitions (M&A) to remain competitive. Microsoft has a close partnership with OpenAI, Amazon is negotiating a multi-billion-dollar investment in Anthropic to develop new chips, and Apple is exploring a collaboration with Perplexity. All of these companies understand that the window of opportunity for accessing the most valuable AI resources—models, data, and talent—is rapidly closing.

This trend, however, is not limited to the US. China, restricted by export controls on semiconductors, is increasingly relying on domestic solutions. Companies like Baidu and Alibaba are driving the development of agent technologies there. Russia and other countries are also working on developing their own models. Agent technology is thus becoming globalized – not only in terms of its prevalence, but also in terms of the diversity of its implementations.

Meta's acquisition of Manus sends a clear message: The company is ready to act quickly and decisively. It demonstrates that Meta understands the "Year of Agents" isn't just approaching, it's already begun. The window of opportunity for a large enterprise to acquire an established AI agent application is rapidly closing. In six months, the best opportunities could already be gone – acquired by competitors – or prohibitively expensive. Meta has chosen to act now.

This will have profound implications for the structure of the AI ​​industry over the next two to three years. It's quite possible that we will look back on this week as the pivotal moment when Meta regained its competitive edge in the race for AI agents.

☑️ Our business language is English or German

☑️ NEW: Correspondence in your national language!

 

I would be happy to serve you and my team as a personal advisor.

You can contact me by filling out the contact form or simply call me on +49 89 89 674 804 (Munich) . My email address is: wolfenstein ∂ xpert.digital

I'm looking forward to our joint project.

 

 

☑️ SME support in strategy, consulting, planning and implementation

☑️ Creation or realignment of the digital strategy and digitalization

☑️ Expansion and optimization of international sales processes

☑️ Global & Digital B2B trading platforms

☑️ Pioneer Business Development / Marketing / PR / Trade Fairs

Xpert.Digital has in-depth knowledge of various industries. This allows us to develop tailor-made strategies that are tailored precisely to the requirements and challenges of your specific market segment. By continually analyzing market trends and following industry developments, we can act with foresight and offer innovative solutions. Through the combination of experience and knowledge, we generate added value and give our customers a decisive competitive advantage.

More about it here:

• Use the 5x expertise of Xpert.Digital in one package - starting at just €500/month