Recently, the AI community has been abuzz with news of Meta’s union with Manus. Yet the two earliest and most credible sources reporting the deal presented strikingly different narratives: the WeChat public account “LatePost” ran a headline titled “Meta Acquires Manus for Billions of Dollars; Xiao Hong to Become Meta VP,” implying a massive acquisition. In contrast, ZhenFund—the early investor in Manus—published an article titled “The Manus Team Joins Meta: A Decade of Young Talent.”

This subtle difference in wording is far from accidental. In today’s climate of heightened antitrust scrutiny, whether a deal is framed as a corporate acquisition or a team hiring directly determines whether it sails through—or gets bogged down in prolonged regulatory and legal battles. If Meta truly attempted to acquire Manus as a company, it wouldn’t just face a bill—it would confront the real risk of both U.S. and Chinese antitrust authorities simultaneously hitting pause.

The “Acquihire” Model: How to Sidestep Antitrust Review

My assessment is that, to avoid antitrust scrutiny, Meta and Manus most likely employed the “acquihire” model—a strategy now prevalent among Silicon Valley giants in the chip and AI sectors. Recent U.S. deals in these fields have deliberately avoided elements that trigger anticompetitive concerns, such as acquiring control or exclusive rights to technology. Instead, they’ve adopted a carefully crafted “three-no” transaction structure: no purchase of business operations, no exclusive acquisition of intellectual property, and no dissolution of the target company’s legal entity. This approach successfully bypasses the two core triggers of antitrust law: large-scale asset transfer and change of control. The evasion strategy consists of three key components:

First, large-scale, long-term technology licensing. Rather than buying IP outright, acquirers like NVIDIA pay substantial licensing fees. Legally, this is classified as a commercial partnership—not an asset acquisition—requiring no regulatory filing and avoiding market concentration thresholds.

Second, systematic hiring of core teams. Through high salaries and senior titles, the acquiring company transfers the startup’s entire R&D team. Legally, this is treated as employment—not a merger. Financially, compensation is paid out over time rather than as a lump sum, minimizing immediate cash outflows and balance sheet impact, while allowing founders and key staff to capture most of the deal’s value through employment contracts.

Third, the target company continues to exist and operate independently. The acquirer may purchase a minority stake—or none at all—but the startup remains a legally independent entity.

Each of these actions, viewed in isolation, is lawful: tech licensing is standard business practice, and talent mobility reflects market freedom. Moreover, since acquired firms are typically early-stage startups with recently launched products, low market share, and minimal competitive impact, regulators struggle to apply traditional market-share-based frameworks to assess harm. This has created a de facto regulatory blind spot for such transactions.

The U.S. Perspective: FTC’s Logic and Strategic Concerns

Although the U.S. court dismissed the Federal Trade Commission’s (FTC) 2025 lawsuit alleging Meta’s acquisitions of Instagram and WhatsApp violated antitrust laws, the FTC’s complaint revealed its underlying stance: dominant platforms systematically weaken competitive threats by acquiring nascent or potential rivals, thereby entrenching their ecosystem dominance.

The FTC argued that Instagram and WhatsApp represented distinct, innovative product trajectories that, if allowed to grow independently, could have evolved into serious competitors to Meta. Thus, the acquisitions altered the long-term competitive landscape—not just short-term pricing or market share. Internal Meta emails, including statements from CEO Mark Zuckerberg explicitly framing acquisitions as a strategy to neutralize emerging threats, provided behavioral evidence of anti-competitive intent.

Therefore, if Meta’s deal with Manus involved equity purchase or exclusive licensing of core technology, it would likely face intense FTC scrutiny and potential litigation. Even if ultimately deemed legal—as in Microsoft’s protracted battle over Activision Blizzard—the resulting delays could be devastating in the fast-moving AI industry, where timing is everything.

Evolution of the Model: Silicon Valley’s Regulatory Game

The acquihire playbook was first fired by Microsoft. In 2024, it paid AI startup Inflection 650millionfornon?exclusivelicensingofitsmodelsandhiredmostofitsteam—includingco?founderMustafaSuleyman,formerlyofDeepMindandGoogle—whileInflectiontransitionedintoanonprofit.FTCChairLinaKhanpubliclyquestionedthedeal,launchinganinvestigationintowhetherthe650 million was truly a licensing fee or disguised acquisition payment.

Google later refined this model in its deal with chatbot firm Character.AI, targeting core founders Noam Shazeer and Daniel De Freitas—both ex-Googlers, with Shazeer being a co-inventor of the Transformer architecture foundational to modern LLMs. Unlike Microsoft’s Inflection deal, Google also bought out early Character.AI investors’ shares, signaling that part of the payment compensated for equity, not just IP. Later, when a young user died by suicide after excessive interaction with Character.AI’s bot, the mother sued—and named Google as a co-defendant, arguing it effectively controlled Character.

In 2025, Google executed a similar transaction with AI code-generation startup Windsurf: it hired the core team but left Windsurf’s remaining assets to be acquired by another firm, Cognition—deliberately avoiding control over the original company’s assets.

Also in June 2025, Meta acquired a 49% stake in data-labeling firm Scale AI for $14.3 billion and appointed its CEO, Alexander Wang, to lead Meta’s AI efforts. To preempt controversy, Scale AI remained independently operated, and Wang retained his board seat.

Just a week ago, NVIDIA announced it would pay approximately $20 billion to license technology from AI chip startup Groq and hire its core team. The deal is particularly notable: NVIDIA dominates the GPU market, while Groq’s founder, Jonathan Ross—a former Google engineer and inventor of the TPU—developed Groq’s LPU (Language Processing Unit) architecture specifically for AI inference, positioning it as a potential challenger to NVIDIA in that segment.

But before Groq could scale, its team and key tech were absorbed by NVIDIA. For end users—AI firms, cloud providers, and internet platforms—Groq’s existence meant more choice and bargaining power in AI accelerator procurement. Its absorption by NVIDIA further weakens that leverage, potentially leading to higher costs.

China’s Review: Substance Over Form

Let’s examine China’s antitrust framework. Manus originated in China; its founder Xiao Hong is the legal representative of Beijing Butterfly Effect Technology Co., Ltd., which operates in the Chinese market. Its Monica intelligent dialogue-generation algorithm has been registered with the Cyberspace Administration of China (CAC).

Earlier reports indicated Manus had relocated to Singapore—likely because its AI Agent services rely heavily on U.S. models like Anthropic’s Claude. Nevertheless, if Chinese regulators determine the Meta-Manus deal could disrupt competition in China’s AI services market, they retain authority to review it.

Under China’s revised “Provisions on Thresholds for Notification of Concentrations of Undertakings” (effective early 2024), a transaction requires antitrust filing if: (1) the combined global turnover of all parties exceeds RMB 12 billion; (2) their combined China turnover exceeds RMB 4 billion; and (3) at least two parties each have China revenue exceeding RMB 800 million. However, most Chinese AI startups—including Manus—likely fall below these thresholds.

Crucially, Article 26 of China’s Anti-Monopoly Law states: even if a concentration falls below statutory thresholds, the State Council’s antitrust authority may require notification if there is evidence it “has or may have the effect of eliminating or restricting competition.” Thus, SAMR (State Administration for Market Regulation) can initiate review sua sponte.

Per Article 4 of the “Provisions on Review of Concentrations of Undertakings,” SAMR assesses whether Meta has obtained “control or decisive influence” over Manus by considering factors including:

(1) The purpose and future plans of the transaction (clearly aimed at acquiring technology and talent to eliminate competitive threats);

(5) Appointment of senior management (Manus’s core team joining Meta, with Xiao Hong becoming Meta VP);

(7) Existence of significant commercial relationships or cooperation agreements (LatePost reported the deal value at “billions of dollars,” Meta’s third-largest acquisition ever—surpassed only by WhatsApp and Scale AI).

Chinese regulators consistently emphasize substance over form. In complex VIE-structured deals, they’ve repeatedly demonstrated the ability to “pierce the veil” and assess economic reality. Given these indicators—even without equity transfer—Meta has arguably exerted decisive influence over Manus. Therefore, if regulators deem it necessary, they could apply a “substance-over-form” principle to classify the deal as a notifiable concentration.

The greatest antitrust risk lies in vertical foreclosure. Although Meta’s Llama models still trail behind top-tier systems like Gemini, ChatGPT, and Claude in capability and market share, Chinese regulators evaluate competitive effects beyond current market position—they consider barriers to entry, technological control, and dependency across the supply chain. By integrating with an AI Agent firm, Meta gains access to cutting-edge inference technology and additional data, strengthening its algorithm development. Combined with Meta’s existing compute power and app distribution reach, such vertical integration could negatively impact market competition.

In conclusion, the Meta-Manus deal exemplifies the maturing acquihire playbook in Silicon Valley. When technological iteration outpaces legal reform, such borderline strategies become the unspoken shortcut for tech giants. While U.S. and Chinese regulators increasingly recognize the risks, crafting effective oversight that curbs anti-competitive harm without stifling innovation remains an unresolved challenge.

Recently, the AI community has been abuzz with news of Meta’s union with Manus. Yet the two earliest and most credible sources reporting the deal presented strikingly different narratives: the WeChat public account “LatePost” ran a headline titled “Meta Acquires Manus for Billions of Dollars; Xiao Hong to Become Meta VP,” implying a massive acquisition. In contrast, ZhenFund—the early investor in Manus—published an article titled “The Manus Team Joins Meta: A Decade of Young Talent.”

This subtle difference in wording is far from accidental. In today’s climate of heightened antitrust scrutiny, whether a deal is framed as a corporate acquisition or a team hiring directly determines whether it sails through—or gets bogged down in prolonged regulatory and legal battles. If Meta truly attempted to acquire Manus as a company, it wouldn’t just face a bill—it would confront the real risk of both U.S. and Chinese antitrust authorities simultaneously hitting pause.

The “Acquihire” Model: How to Sidestep Antitrust Review

My assessment is that, to avoid antitrust scrutiny, Meta and Manus most likely employed the “acquihire” model—a strategy now prevalent among Silicon Valley giants in the chip and AI sectors. Recent U.S. deals in these fields have deliberately avoided elements that trigger anticompetitive concerns, such as acquiring control or exclusive rights to technology. Instead, they’ve adopted a carefully crafted “three-no” transaction structure: no purchase of business operations, no exclusive acquisition of intellectual property, and no dissolution of the target company’s legal entity. This approach successfully bypasses the two core triggers of antitrust law: large-scale asset transfer and change of control. The evasion strategy consists of three key components:

First, large-scale, long-term technology licensing. Rather than buying IP outright, acquirers like NVIDIA pay substantial licensing fees. Legally, this is classified as a commercial partnership—not an asset acquisition—requiring no regulatory filing and avoiding market concentration thresholds.

Second, systematic hiring of core teams. Through high salaries and senior titles, the acquiring company transfers the startup’s entire R&D team. Legally, this is treated as employment—not a merger. Financially, compensation is paid out over time rather than as a lump sum, minimizing immediate cash outflows and balance sheet impact, while allowing founders and key staff to capture most of the deal’s value through employment contracts.

Third, the target company continues to exist and operate independently. The acquirer may purchase a minority stake—or none at all—but the startup remains a legally independent entity.

Each of these actions, viewed in isolation, is lawful: tech licensing is standard business practice, and talent mobility reflects market freedom. Moreover, since acquired firms are typically early-stage startups with recently launched products, low market share, and minimal competitive impact, regulators struggle to apply traditional market-share-based frameworks to assess harm. This has created a de facto regulatory blind spot for such transactions.

The U.S. Perspective: FTC’s Logic and Strategic Concerns

Although the U.S. court dismissed the Federal Trade Commission’s (FTC) 2025 lawsuit alleging Meta’s acquisitions of Instagram and WhatsApp violated antitrust laws, the FTC’s complaint revealed its underlying stance: dominant platforms systematically weaken competitive threats by acquiring nascent or potential rivals, thereby entrenching their ecosystem dominance.

The FTC argued that Instagram and WhatsApp represented distinct, innovative product trajectories that, if allowed to grow independently, could have evolved into serious competitors to Meta. Thus, the acquisitions altered the long-term competitive landscape—not just short-term pricing or market share. Internal Meta emails, including statements from CEO Mark Zuckerberg explicitly framing acquisitions as a strategy to neutralize emerging threats, provided behavioral evidence of anti-competitive intent.

Therefore, if Meta’s deal with Manus involved equity purchase or exclusive licensing of core technology, it would likely face intense FTC scrutiny and potential litigation. Even if ultimately deemed legal—as in Microsoft’s protracted battle over Activision Blizzard—the resulting delays could be devastating in the fast-moving AI industry, where timing is everything.

Evolution of the Model: Silicon Valley’s Regulatory Game

The acquihire playbook was first fired by Microsoft. In 2024, it paid AI startup Inflection 650millionfornon?exclusivelicensingofitsmodelsandhiredmostofitsteam—includingco?founderMustafaSuleyman,formerlyofDeepMindandGoogle—whileInflectiontransitionedintoanonprofit.FTCChairLinaKhanpubliclyquestionedthedeal,launchinganinvestigationintowhetherthe650 million was truly a licensing fee or disguised acquisition payment.

Google later refined this model in its deal with chatbot firm Character.AI, targeting core founders Noam Shazeer and Daniel De Freitas—both ex-Googlers, with Shazeer being a co-inventor of the Transformer architecture foundational to modern LLMs. Unlike Microsoft’s Inflection deal, Google also bought out early Character.AI investors’ shares, signaling that part of the payment compensated for equity, not just IP. Later, when a young user died by suicide after excessive interaction with Character.AI’s bot, the mother sued—and named Google as a co-defendant, arguing it effectively controlled Character.

In 2025, Google executed a similar transaction with AI code-generation startup Windsurf: it hired the core team but left Windsurf’s remaining assets to be acquired by another firm, Cognition—deliberately avoiding control over the original company’s assets.

Also in June 2025, Meta acquired a 49% stake in data-labeling firm Scale AI for $14.3 billion and appointed its CEO, Alexander Wang, to lead Meta’s AI efforts. To preempt controversy, Scale AI remained independently operated, and Wang retained his board seat.

Just a week ago, NVIDIA announced it would pay approximately $20 billion to license technology from AI chip startup Groq and hire its core team. The deal is particularly notable: NVIDIA dominates the GPU market, while Groq’s founder, Jonathan Ross—a former Google engineer and inventor of the TPU—developed Groq’s LPU (Language Processing Unit) architecture specifically for AI inference, positioning it as a potential challenger to NVIDIA in that segment.

But before Groq could scale, its team and key tech were absorbed by NVIDIA. For end users—AI firms, cloud providers, and internet platforms—Groq’s existence meant more choice and bargaining power in AI accelerator procurement. Its absorption by NVIDIA further weakens that leverage, potentially leading to higher costs.

China’s Review: Substance Over Form

Let’s examine China’s antitrust framework. Manus originated in China; its founder Xiao Hong is the legal representative of Beijing Butterfly Effect Technology Co., Ltd., which operates in the Chinese market. Its Monica intelligent dialogue-generation algorithm has been registered with the Cyberspace Administration of China (CAC).

Earlier reports indicated Manus had relocated to Singapore—likely because its AI Agent services rely heavily on U.S. models like Anthropic’s Claude. Nevertheless, if Chinese regulators determine the Meta-Manus deal could disrupt competition in China’s AI services market, they retain authority to review it.

Under China’s revised “Provisions on Thresholds for Notification of Concentrations of Undertakings” (effective early 2024), a transaction requires antitrust filing if: (1) the combined global turnover of all parties exceeds RMB 12 billion; (2) their combined China turnover exceeds RMB 4 billion; and (3) at least two parties each have China revenue exceeding RMB 800 million. However, most Chinese AI startups—including Manus—likely fall below these thresholds.

Crucially, Article 26 of China’s Anti-Monopoly Law states: even if a concentration falls below statutory thresholds, the State Council’s antitrust authority may require notification if there is evidence it “has or may have the effect of eliminating or restricting competition.” Thus, SAMR (State Administration for Market Regulation) can initiate review sua sponte.

Per Article 4 of the “Provisions on Review of Concentrations of Undertakings,” SAMR assesses whether Meta has obtained “control or decisive influence” over Manus by considering factors including:

(1) The purpose and future plans of the transaction (clearly aimed at acquiring technology and talent to eliminate competitive threats);

(5) Appointment of senior management (Manus’s core team joining Meta, with Xiao Hong becoming Meta VP);

(7) Existence of significant commercial relationships or cooperation agreements (LatePost reported the deal value at “billions of dollars,” Meta’s third-largest acquisition ever—surpassed only by WhatsApp and Scale AI).

Chinese regulators consistently emphasize substance over form. In complex VIE-structured deals, they’ve repeatedly demonstrated the ability to “pierce the veil” and assess economic reality. Given these indicators—even without equity transfer—Meta has arguably exerted decisive influence over Manus. Therefore, if regulators deem it necessary, they could apply a “substance-over-form” principle to classify the deal as a notifiable concentration.

The greatest antitrust risk lies in vertical foreclosure. Although Meta’s Llama models still trail behind top-tier systems like Gemini, ChatGPT, and Claude in capability and market share, Chinese regulators evaluate competitive effects beyond current market position—they consider barriers to entry, technological control, and dependency across the supply chain. By integrating with an AI Agent firm, Meta gains access to cutting-edge inference technology and additional data, strengthening its algorithm development. Combined with Meta’s existing compute power and app distribution reach, such vertical integration could negatively impact market competition.

In conclusion, the Meta-Manus deal exemplifies the maturing acquihire playbook in Silicon Valley. When technological iteration outpaces legal reform, such borderline strategies become the unspoken shortcut for tech giants. While U.S. and Chinese regulators increasingly recognize the risks, crafting effective oversight that curbs anti-competitive harm without stifling innovation remains an unresolved challenge.