• Meta acquisition: Meta Platforms agreed to buy Manus, a Singapore-based AI startup originally tied to China, for $2.5 billion with a $500 million retention pool for employees.
• Manus capabilities: The AI tool handles complex tasks like long research reports, slideshows, or websites using Anthropic and other foreign models.
• Strategic repositioning: Manus founders distanced from China, relocating headquarters to Singapore and shutting down services in China to ease U.S. investment concerns.
• Washington reaction: U.S. officials reacted mutedly, viewing the move as compliant with export and investment controls and a model for other Chinese AI firms.
• Beijing surprise: Some Chinese officials disliked the sale, fearing it handed U.S. access to Chinese-developed technology and encouraged similar exits.
• Founders’ background: Xiao Hong and Ji Yichao built Butterfly Effect and earlier apps for international markets before launching Manus in 2024.
• Growth efforts: Manus rebuffed local Chinese investment offers, paused a joint tool with Alibaba, expanded in Singapore, and partnered with Microsoft and Stripe while growing revenue to $125 million run rate.
• Meta integration: Meta plans to operate Manus services, integrate them into Meta AI and social platforms, and leverage WhatsApp and Instagram distribution while Manus talent joins Meta.

---

Meta CEO Mark Zuckerberg wore AI-powered smartglasses at a September conference in Menlo Park, Calif. Nic Coury/AP

By

Raffaele Huang

,

Kate Clark

and

Amrith Ramkumar

Dec. 31, 2025 12:01 am ET

Workers at Butterfly Effect, an artificial-intelligence startup with Chinese roots, gathered in March to count down to the launch of a demo of a new AI-powered tool called Manus.

Released in the shadow of DeepSeek, a China-built AI model that rocked the U.S. market in January owing to its advanced capabilities and low cost, Manus became an overnight hit. The product, which used models from Anthropic and others to produce detailed research reports and perform a host of other tasks, showcased the talents of China’s entrepreneurs and startups in a burgeoning global battle for AI supremacy.

The assumptions that underlie the East-West divide were turned upside down Monday when Meta Platforms META 1.10%increase; green up pointing triangle said it was buying Manus, which is now based in Singapore. The deal, which was valued at $2.5 billion, according to people familiar with the matter, included a $500 million retention pool for the startup’s employees, one of the people said.

Chinese AI chip makers, chatbot developers and robotics startups have recently raised billions of dollars and are preparing for initial public offerings in Shanghai or Hong Kong. But they still face challenges squaring off with U.S. rivals that can access enormous sums to compete globally.

The founders of Manus made strategic decisions to distance it from its Chinese roots, helping position the company for U.S. investment.

Winston Ma, a professor at New York University School of Law and a partner at Dragon Capital, said that if the deal closes smoothly, “It creates a new path for the young AI startups in China.” The question is “whether D.C. will support this or say this is just another way of U.S. capital being invested into a Chinese company,” he said.

Manus is designed to handle more-complex tasks than a typical chatbot. adek berry/AFP/Getty Images

Before being approached by Meta, Manus executives and the startup’s investors weighed staying independent and raising substantially more capital, people familiar with the matter said. Like many successful AI startups, Manus faced a difficult reality: Without a platform partner such as Meta, it would be challenging and costly to reach global scale, and raising additional capital could make the company too expensive for potential suitors.

Meta spokesman Andy Stone said that there would be no continuing Chinese ownership interests in Manus after the transaction and that the startup would discontinue its services and operations in China.

Beijing and Washington react

Meta’s deal surprised some officials in Beijing, some of whom disliked the agreement because they considered Manus an example of China’s AI power, people familiar with the officials’ thinking said. They believed that the sale would give the U.S. access to technology developed by Chinese engineers and encourage other startups to pursue a similar funding path, the people said. Beijing appears to have few tools to influence the deal given Manus’s foothold in Singapore.

In Washington, the reaction was muted, a signal that Manus’s moves to avoid violating U.S. rules that restrict outbound investments in key technologies eased concern about its China ties.

“The indicators on this one seem to be all pointing at least on the surface in the right direction,” said Chris McGuire, who worked on technology-export controls in the Biden administration and is now a senior fellow at the Council on Foreign Relations. He views the deal as evidence that export and investment restrictions work and could squeeze other Chinese AI companies, pushing them to do more deals with U.S. partners.

The response in Washington is a departure from earlier this year, when Trump administration officials including members of the National Security Council worried about a $75 million fundraising round for Manus led by the U.S. venture firm Benchmark, people familiar with the matter said.

Shortly afterward, the Treasury Department began reviewing whether the funding round ran afoul of rules banning investments in critical technologies for countries that pose national-security risks, the people said. The issue generally fell off the radar of many administration officials after Manus moved its headquarters to Singapore, they said. A Treasury spokesman declined to comment. 

Beginnings in China

Manus’s core leaders are two young Chinese entrepreneurs, Xiao Hong and Ji Yichao—also known as “Red” and “Peak.” 

Xiao, born in 1993 in a small city in China’s southern province of Jiangxi, went to Huazhong University of Science and Technology in Wuhan to study software engineering. After graduating, he founded a company in Wuhan to develop software for the Chinese tech company Tencent’s popular messaging and payment app, WeChat. Tencent later invested in his company.

Ji, born in 1992, grew up in Colorado and Beijing and went to a university in Beijing to study computer science. When he was in high school, he created a browser app for Apple’s iPhone, called Mammoth Browser, which later became popular. In 2012, he started his own startup, Peak Labs, and received funding from venture-capital firms such as ZhenFund and Sequoia China—now called HSG.

Xiao set up Manus’s parent, Butterfly Effect, in 2022 and launched a ChatGPT-powered application for browsers, called Monica. The Butterfly Effect apps targeted markets outside China, mainly North America, Japan and South Korea, Xiao said in a recent podcast. Butterfly Effect had offices in Beijing and Wuhan.

In October 2024, inspired by the San Francisco-based AI coding tool Cursor, Butterfly Effect started developing Manus, which used several American AI models that aren’t available in China. The project’s name, “Manus,” came from the Massachusetts Institute of Technology’s Latin motto, “Mens et Manus,” meaning mind and hand.

Manus in March released a demo of its AI agent, which is designed to handle more-complex tasks than a typical chatbot, such as producing a 100-page research report, generating a slideshow or building a website. At the time, most of its researchers and engineers were based in China.

Some in China called it another “DeepSeek moment.” An invitation code that gave people early access to the tool was resold for more than $1,000 on social-media and e-commerce sites.

Growth spurt

Earlier this year, several local governments in China approached Manus and offered to invest in the startup, but its founders turned them down, according to people familiar with the matter. They were concerned that such connections could cause scrutiny in the West and create challenges for its global business, the people said.

Manus also shelved a plan to join with Alibaba to introduce a Chinese version of the tool that had been announced in March, people familiar with the plan said. 

Around that time, Manus secured the funding that drew scrutiny in the U.S. It soon moved its headquarters to Singapore and laid off some employees in China.

Manus has since expanded its team in Singapore, recruiting new staff and showering the city-state’s subway platforms with its black-and-white logo. In recent months, Manus has announced partnerships with Microsoft and the payment company Stripe.

Earlier in December, Manus said its revenue run rate, a metric used by startups representing a short-term revenue figure that has been annualized, had risen to $125 million from $90 million in August.

Meta began negotiations for an acquisition in mid-December, and Mark Zuckerberg wanted to reach an agreement by the end of the year, people familiar with the matter said. Some existing shareholders of the startup didn’t expect the company to be bought out so quickly, the people said.

Selling to Meta gives Manus access to distribution channels such as WhatsApp and Instagram and a well-funded owner able to help cover computing and infrastructure costs. Meta said it plans to continue to operate and sell Manus’s service and integrate it into its suite of social-media products.

“Manus’s exceptional talent will join Meta’s team to deliver general-purpose agents across our consumer and business products, including in Meta AI,” Meta said in a blog post.

The deal marks a quick exit for Benchmark, as well as the earlier investor HSG, the venture-capital firm led by the star Chinese technology investor Neil Shen. Benchmark turned an investment that drew intense criticism from certain members of the Silicon Valley elite into a decisive win.

Write to Raffaele Huang at raffaele.huang@wsj.com, Kate Clark at kate.clark@wsj.com and Amrith Ramkumar at amrith.ramkumar@wsj.com

The Global AI Race

Coverage of advancements in artificial intelligence, selected by the editors

Get WSJ's AI Newsletter

If U.S.-China AI Rivalry Were Football, the Score Would Be 24-18

OpenAI Updates ChatGPT Amid Battle for Knowledge Workers

Disney to Invest $1 Billion in OpenAI and License Characters

AI Hackers Are Coming Dangerously Close to Beating Humans

Copyright ©2025 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8

• Mafia migration: Cesare Mori’s 1920s Sicilian crackdown pushed mobsters into specific American enclaves, allowing researchers to study sudden criminal inflows.
• Data sources: Analysts combined de-anonymized 1920–1940 censuses with Social Security death records, narcotics files, redlining maps, and other historical and modern indicators.
• Immediate effects: Mori enclaves saw higher incarceration, elevated homicides (especially Chicago), and a roughly 40% greater likelihood of redlining compared to similar Italian neighborhoods.
• Housing trend: Despite discrimination, homeownership in these communities rose starting in the 1930s.
• Long-run outcomes: By the 21st century, descendants of Mori enclaves showed stronger economic performance than peers from other local Italian or Sicilian-origin areas.
• Possible mechanisms: Authors propose illicit-income reinvestment plus informal credit/enforcement arrangements partly substituted for formal institutions.
• Author’s stance: Skepticism remains about the purported long-term benefits despite the observed overperformance.
• Additional readings: Listed works cover peer-review bias, evidence in economic reporting, crime age shifts, prosecutors’ racial views, data center investment forecasting, Opportunity Zone jobs, hate-crime reporting bias units, income-based school choice use, and methodological cautions on group-level studies.

---

[

](https://images.unsplash.com/photo-1708247874023-f6d71a45113a?fm=jpg&q=60&w=3000&ixlib=rb-4.1.0&ixid=M3wxMjA3fDB8MHxwaG90by1wYWdlfHx8fGVufDB8fHx8fA%3D%3D)

Courtesy Peter Herrmann/Unsplash.

Welcome to the inter-holiday edition of Concrete Evidence, my last words of 2025. There’s no better time to talk about Mafia violence!

A new paper from Bocconi University’s BAFFI Centre explores what happened in America when Mussolini cracked down on the mob in the mid-1920s. The dictator turned loose the “Iron Prefect,” Cesare Mori, for a military campaign against Mafia strongholds on the island of Sicily. Many gangsters fled to the states to see if, perchance, there might be some criminal opportunities available during our Prohibition era.

These folks largely wound up in enclaves seeded by previous waves of immigration—not just from Italy or even Sicily in general, but from the specific places Mori targeted for enforcement. This abrupt change, concentrated in a very particular set of places, allows the researchers to see what a sudden influx of criminality does to a community. The upshot is that these “Mori enclaves” suffered in the short run—but over time rebounded and even thrived.

The authors painstakingly assembled numerous data sources, aided by the fact that old records don’t have the same privacy protections as new ones. They use full-count, de-anonymized data from the 1920 to 1940 censuses. These can tell them that, for example, 22-year-old future mob boss Joe Magliocco lived at 646 Union St. in Brooklyn in 1920. They also have Social Security death records, Bureau of Narcotics files revealing the names of Mafia members and associated businesses, old redlining maps, current-century data on how neighborhoods are faring, and some other odds and ends like Ellis Island manifests and detailed Chicago homicide records.

What happens when the mob shows up? Unsurprisingly, a lot of things get worse.

Folks from these Mori enclaves were more likely to respond to the 1940 census from prison, relative to those from other Sicilian neighborhoods (who’d had similar incarceration trends previously). Homicide rates went up in Chicago Mori enclaves, relative to other Chicago neighborhoods that’d had similar trends previously. Redlining is infamous for its impact on black neighborhoods, but Italian neighborhoods with lots of mob activity were also frowned upon: “Mori neighborhoods were about 40 percent more likely to be redlined than other areas, even when compared to neighborhoods with similar shares of Italian and Sicilian immigrants.”

However, despite the redlining, homeownership in these places rose starting in the 1930s. In the 21st century, people from Mori enclaves have had better economic outcomes than people born elsewhere in the same cities, even after accounting for neighborhoods’ overall Italian and Sicilian immigration levels.

Could the mob presence have actually strengthened these communities over the long run? The authors suggest the “patterns may reflect the reinvestment of illicit profits into local assets and the emergence of informal credit and enforcement arrangements that partially substituted for formal financial institutions.”

I’m skeptical, though it’s certainly interesting that Mori neighborhoods eventually not only matched, but outperformed, what you might expect from an Italian neighborhood otherwise.

Other Work of Note

• How biased are the people who review academic papers against younger scholars and those from less prestigious universities?

• What types of evidence does the Economic Report of the President rely on?

• The changing age profile of crime.

• Prosecutors who are just starting out, who are Democrats, or who previously worked on the defense side are more likely to endorse “color-conscious” as opposed to race-neutral views.

• Forecasting investment in data centers.

• “Opportunity Zones” seem to move jobs more than create them.

• Police “bias units” seem to increase hate-crime reporting.

• Higher-income parents are more likely to take advantage of school choice when it’s available.

• Cautionary notes about studies that rely on random placement “into groups such as schools or regions to estimate the effects of group-level variables on these individuals.”

See y’all in 2026.

Share

It is New Year’s Eve, and the SS Poseidon glides from New York to Athens on its final voyage. Rich Manhattanites clink glasses and blow paper horns. All is hunky dory. Unfortunately for the ageing liner — and its bowtied passengers — it is shortly to be capsized by a midnight tsunami. We watch as a succession of ensemble characters are picked off by flooding, explosions and theatrical heart attacks until a small band of survivors is finally rescued.

A festive hit in 1972, The Poseidon Adventure was among a clutch of disaster movies that thrilled and chilled Seventies audiences. The optimism of the Sixties had dissipated; Vietnam was going down the toilet; New York was broke, with random crime and visible decline terrorising the residents of Fear City and their horrified compatriots. Airport in 1970, Earthquake and The Towering Inferno in 1974: all mirrored the secret fears of cinemagoers whose faith in authority had waned, who braced for catastrophe in a world gone dark with only their pluck to buoy them.

Two decades later, at the end of history, CGI inflamed the disaster-movie genre: Independence Day (1996), Armageddon (1998) and Deep Impact (1998) offered sweeping apocalyptic visuals — crumbling landmarks, waves dwarfing Midtown skyscrapers, cows mooing in the carousel of a hurricane. These were fantasies of Western annihilation, fantasies which peered into the White House situation room and caught the bead of sweat on the president’s forehead. At a time of relative prosperity and security — and before 9/11 — the focus was on the spectacle of destruction, how far directors could push newly available special effects. The usual bungling authorities still made an appearance: stock hawkish commanders, dumbfounded local cops — but unlike in many Seventies equivalents the threat was external, rather than a crumbling moral core.

These days the disaster blockbuster has fallen out of fashion. Covid killed the zombie apocalypse by bringing the weary realities of contagion to our doorsteps; climate anxiety shattered the comet careening through space with its promise of slow and sweaty demise. But visions of urban destruction have not disappeared; they’ve simply migrated to the news. The 2020s, so far the decade of disease, decline and demographic transformation, have taught us that the most psychically terrible threat is no longer immediate apocalypse — it’s the insidious, years-long erosion of protections and values taken for granted, whose loss is written on the sidewalks of London, and written on the wall for a New York City preparing to welcome an untested Millennial mayor.

In the real world, social contracts are not blown up but forgotten in the back of some dusty drawer. Personal safety is not, for the most part, violated by a lone attacker — it disappears when we acclimatise to the petty transgressions of normal people, those who press their bodies against us to shuffle in on our subway fares, who strut out of Tesco with bottles clinking in their pockets, who hock spit at minor officials when their toddler antics are reprimanded. Disaster happens, yes, when planes crash; but it also happens when they don’t — when dismal passengers rub their bare feet, puff clouds of sickly vape smoke into the cabin, rage at the stewardesses for no reason at all. Politeness, consideration, conformity: these are the casualties of a real-life disaster movie, a world forever changed by a lost expectation of a sensible and cohesive public. Citywide breakdowns recall riots, looting, mass catastrophe — but social orders can buckle under the weight of less spectacular tragedies, and barrier-bumpers, gum-spitters and participants in the pavement pissoir are all doing their bit to lower trust and heighten quotidian horror.

New York landmarks have been nuked, flooded and blown up countless times on screen. Lady Liberty has been decapitated (Cloverfield), buried in sand (Planet of the Apes) and entombed in ice (The Day After Tomorrow). She embodies a shining city, a way of life worth saving — but today that image reads differently, a dismal cliché, probably because the gleaming metropolis she represents has given way to a grimier reality. In the disaster movie, New York and London are depicted quaintly, sentimentally, flattened into tableaux of cheerful hot-dog vendors and chintzy tearooms, overcoats on Madison Avenue and urchins in the East End. The fantasy of the disaster movie is that, apart from the approaching asteroid or the mile-high tidal wave, nothing of note has ever changed in the endangered city. The person-sized disasters of managerial neglect, low-level lawlessness and economic hardship don’t exist.

Yet in 2026, disaster looks different. As New York prepares to inaugurate a 34-year-old, untested and catastrophically naive mayor on New Year’s Day, its merrymaking residents, much like the Champagne-popping revellers of the fictional Poseidon, don’t know what’s coming their way. Melodramatic, you might say — but consider how fragile is the order of the day-to-day. Even without a total naif in Gracie Mansion, life here rests on a knife edge: street-corner encounters turn violent without provocation, sly parcel-swipers continually invade hallways, and looking askance at the wrong commuter can earn you a slashed cheek or lit match. Mamdani has threatened to make the city’s buses, already mobile circus wagons full of antisocial sideshows, free — they could well go the way of the Central Park pool this summer, which became de facto baths for the “unhoused” and handy toilets for the disturbed. Men who pay for sex, ravaging the city’s desperates and driving women and girls into the hands of pimps, may no longer even be rapped on the knuckles: decriminalised “sex work” and scot-free soliciting could rob prostitutes of already scant protections under the guise of well-meaning toleration and sex positivity. And those beaten by partners might, if Mayor Mamdani’s campaign pitch comes off, be referred to swamped and bureaucratic social workers instead of police officers — shoved through a consequence-lite muddle of paperwork instead of their abusive partners being shoved into a cell.

These are the real disasters headed for the American city: a peek across the Atlantic is all New Yorkers need to envision the immediate possibilities of progressive mismanagement. Sadiq Khan’s London, a city swamped in antisocial miasma, shows us what happens when incompetent policing collides with vanishing public values, when going easy on the violent and depraved becomes cruelty to the innocent. Sexual offences on the Tube up 10% in a year; violent crime up 40% in a decade; security tags on bottles of Persil. Of course, New York has its own charms — subway cars becoming sleeper trains for fent addicts with hands down their pants; guns. With Mamdani installed, the complaints of the law-abider look soon, as in London, to be drowned out by the appeals of our betters to see bad behaviour as evidence of victimhood. As Fraser Nelson has pointed out, Khan has overseen a TfL whose staff are told not to intervene when riders vault or shove through barriers, creating a stark division — also visible in New York — between those who gleefully bump and those who sourly tap their way in. The Met have a similar institutional limpness when it comes to shoplifting; as Nelson puts it: “You can walk into a shop, steal anything you like, walk away — and the chances are you won’t be caught.”

Why do we love disaster movies? The blockbusters of the Nineties impart a strange comfort: they impress on us the value of what is under attack, how lucky we are to live unmolested by comets and quakes, how charming and functional regular life is. But how true is that today, when urban life is the way it is? While violent crime in New York is down since the Nineties, petty offences — just like in London — have jumped since Covid owing to lax enforcement, creating the perception of depressing lawlessness. Maybe that’s why the big-budget disaster film has fallen out of fashion; the city no longer feels like the symbol of national or Western order, and the stakes of mass destruction seem lower when the target is itself a crucible of chaos. It is telling that the only recent cinematic success in this genre, Twisters (2024), jeopardised not a noble metropolis but a ramshackle rodeo town and the main character’s humble homestead. Today the atmosphere of the city more closely recalls the Seventies, when moral decline and distrust in authorities created the first true classics of the disaster genre. If the on-screen apocalypse has a third wind, it will be in this mould. As for reality, will 2026 be a disaster? Very possibly; the real question is whether those in London or New York, so accustomed to demoralising, ubiquitous delinquency, will even notice.

---

• Death and Recognition: Lou Gerstner died at age 83 after being celebrated for rescuing IBM from decline.
• First Outsider CEO: Gerstner was the first non-IBM executive to lead the company, taking the helm in 1993.
• Culture and Mission Shift: He reoriented IBM’s culture and mission toward business services instead of competing with Microsoft on operating systems.
• Workforce Reduction: As part of the turnaround, he laid off roughly 10% of employees, challenging IBM’s lifetime-tenure tradition.
• Financial Turnaround: Under Gerstner’s nine-year leadership, IBM’s share price rose from about $13 to roughly $80, supporting many jobs.
• Post-IBM Career: After IBM, he worked in private equity and chaired the Carlyle Group from 2003 to 2008.
• Philanthropic Efforts: Gerstner Philanthropies invested over $300 million in biomedical research, education, and humanitarian causes.
• Policy Note: The White House considered acquiring a 10% stake in an unnamed chip maker after Trump’s confrontation with Intel CEO Lip-Bu Tan, raising questions about MAGA corporate statism.

---

---

By

The Editorial Board

Dec. 29, 2025 5:52 pm ET

29

---

BPC > Only use to renew if text is incomplete or updated: | archive.ph

BPC > Full article text fetched from (no need to report issue for external site): | archive.today | archive.md

Louis V. Gerstner Jr. in 1993. Stringer ./Reuters

Starting a business is hard, but harder still is reviving a once-great company that has fallen from its glory days. You have to change the culture in addition to finding a new business mission. That was the great achievement of Louis V. Gerstner Jr., who revived IBM in the 1990s and died Saturday at age 83.

Lou Gerstner was the first outsider to take the helm at IBM, the storied firm that once looked so dominant in computer mainframes that the Justice Department spent foolish years trying to break it up. But like many former giants, IBM lost its leadership to upstarts as the personal computer and software eras surpassed mainframes.

Taking over at IBM in 1993 after stints at RJR Nabisco and American Express, Gerstner pivoted the company from hardware to focus on business services. He stopped trying to compete with Microsoft’s Windows operating system and instead offered customers expertise in how to integrate corporate data and networks. He laid off some 10% of the company’s employees, which was a shock to the IBM tradition of lifetime tenure but underscored the necessity of change for company survival.

The strategic shift worked, and IBM revenues boomed. During his nine years as CEO, IBM’s market value rose from about $13 to $80 or so a share. The turnaround saved tens of thousands of jobs and provided comfortable livelihoods and retirements for countless families.

Gerstner worked in private equity after he left IBM, serving as chairman of the Carlyle Group from 2003-2008. He was a notable philanthropist with a special passion for education, as he showed in op-eds promoting reform in our pages over the years. Gerstner Philanthropies has invested more than $300 million in biomedical research, education and humanitarian causes.

Business management has fallen out of media and popular favor as entrepreneurs became the glory boys of American capitalism. But CEOs with vision and courage are crucial to making companies succeed. This is the reason that shareholders and boards are willing to pay good managers well. Lou Gerstner was one of the best, to the great benefit of hundreds of thousands of Americans.

The White House considers taking a 10% stake in the chip maker, after Donald Trump meets with Intel CEO Lip-Bu Tan, shortly after he demanded Tan's immediate resignation. Is this another example of MAGA corporate statism?

Copyright ©2025 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8

Already a subscriber? Sign In

---

Videos

• Deal Overview: Nvidia is paying $20 billion for a nonexclusive Groq inference-chip license, with about $13 billion already wired and the remainder to arrive later, including stock packages for employees joining Nvidia.
• Strategic Implication: The licensing agreement gives Nvidia access to Groq’s specialized inference technology and blocks a GPU alternative from becoming a serious competitor.
• Relationship Timing: Groq founder Jonathan Ross and Nvidia CEO Jensen Huang grew closer over recent months, including meetings at the U.S.–Saudi Investment Forum before the Christmas Eve announcement.
• Founder Gains: Ross owns roughly 10% of Groq, a stake worth $2 billion based on the deal value, and received an Nvidia stock package worth several hundred million dollars.
• Investors Positioned: Outside investors such as Disruptive, Chamath Palihapitiya’s Social Capital, BlackRock, and 1789 Capital stand to benefit, with the company planning distributions to investors and employees.
• Revenue Projections: Groq projected $1.4 billion in revenue for 2026, up from $500 million this year and $90 million in 2024, driven largely by sovereign customers like Saudi Arabia’s Aramco.
• Remaining Assets: At least six bids are expected for Groq’s remaining assets, including GroqCloud, with bids predicted to exceed $1 billion; CFO Simon Edwards has been named CEO.
• Contextual Implication: The deal underscores escalating competition for inference talent and technology amid broader tensions in the AI and global trade landscape.

---

Groq founder Jonathan Ross and Nvidia CEO Jensen Huang, in tie, became closer in recent months. David Paul Morris/Bloomberg; Tom Williams/CQ Roll Call/ZUMA Press

Nvidia’s NVDA -1.21%decrease; red down pointing triangle licensing deal with startup Groq is a sign of growing competition for top talent and technology on the next front of the AI war: inference.

The $20 billion licensing agreement gives Nvidia access to Groq’s specialized inference-chip technology used to run artificial-intelligence applications, and prevents a GPU alternative from becoming a serious competitor.

The deal represents a remarkable turn for the nearly decade-old chip startup. The company at one point nearly ran out of cash and asked staff to trade salaries for equity, its founder, Jonathan Ross, said on the 20VC podcast earlier this year.

It is unclear when the two companies initiated deal talks. Ross and Nvidia Chief Executive Jensen Huang grew closer over the past two months, and spent time together in November at the U.S.–Saudi Investment Forum in Washington, D.C., according to people familiar with the matter, before announcing the deal on Christmas Eve.

Here is what we know about the deal:

What Nvidia is paying Groq

Nvidia is paying $20 billion for a nonexclusive licensing agreement with Groq, according to people familiar with the matter. Of that, about $13 billion has already been wired to the startup, with the remainder set to be sent in the coming months, the people said. The overall deal size includes stock packages for the employees joining Nvidia, one of the people said.

How much Groq’s founder stands to make

Groq founder Ross, who previously worked at Google and helped create its TPU chips, has received an Nvidia stock package worth at least several hundred million dollars, the people said. Ross owns about 10% of Groq—a stake worth $2 billion based on the headline value of the Nvidia deal. Other investors that stand to see significant gains include venture firm Disruptive, the company’s largest outside investor, which led its most recent financing at a $6.9 billion valuation, as well as Chamath Palihapitiya’s Social Capital, BlackRock and Donald Trump Jr.’s 1789 Capital.

More specifics on the deal

There are no performance milestones tied to Nvidia’s licensing payments. The company, which raised billions in venture-capital funding, will make distributions to investors and employees soon.

How much revenue Groq was expected to generate

Groq had been projecting revenue of $1.4 billion in 2026, up from about $500 million this year and about $90 million in 2024, people familiar with its financials said. Its revenue growth was driven largely by sovereign customers, including Saudi Arabia’s Aramco.

What’s next for the remaining portion of Groq

There are expected to be at least six bids for Groq’s remaining assets, including its AI-inference platform, GroqCloud, people familiar with the matter said. Bids for that portion of the company are expected to top $1 billion, one of those people said. Groq Chief Financial Officer Simon Edwards has been named CEO.

Write to Kate Clark at kate.clark@wsj.com

What’s Next for Nvidia

From global trade tensions to AI

Nvidia Licenses Groq’s AI Technology

S&P 500 Hits Record, Nvidia Rising

Trump’s OK of AI Chip Sales in China Called 'Dangerous'

Nvidia Tops List of Best-Managed Companies

• Meta Purchase: Meta Platforms is acquiring Singapore-based AI startup Manus for over $2 billion.
• Manus Capabilities: Manus develops AI agents that produce research reports and build websites using models from Anthropic and Alibaba.
• Integration Plans: Manus services will continue operating and be integrated into Meta’s social-media products while scaling to more businesses.
• Leadership Alignment: Manus CEO Xiao Hong will report to Meta COO Javier Olivan after the acquisition.
• AI Competition: The acquisition positions Meta in the AI agent segment alongside rivals like Google, Microsoft, and OpenAI.
• Manus Growth: Manus has millions of users, crossed $100 million in annual recurring revenue, and raised $75 million in a recent funding round.
• Funding Background: Benchmark led Manus’s funding round, boosting valuation to $500 million, with other investors including HSG, ZhenFund, and Tencent.
• Global Operations: Manus, founded by Butterfly Effect in 2022, moved its headquarters to Singapore after developing outside China with predominantly Chinese researchers.

---

By

Angel Au-Yeung

,

Raffaele Huang

and

Kate Clark

Updated Dec. 29, 2025 11:31 pm ET

---

Meta’s Chief Executive Mark Zuckerberg went on a recruiting blitz this year to build an AI dream team. Associated Press

Meta Platforms has agreed to acquire AI startup Manus, a Singapore-based company with Chinese founders that conducts deep research and performs other tasks for paying users.

Meta is closing the deal at more than $2 billion, according to people familiar with the acquisition. Manus was seeking a fresh round of fundraising with a valuation of $2 billion when Meta approached the startup, some of the people said.

Manus’s co-founder and chief executive, Xiao Hong, who often goes by the nickname “Red,” will report to Javier Olivan, chief operating officer of Meta, some of the people said.

The acquisition is among the highest-profile examples of a major U.S. tech company buying an artificial-intelligence product developed in Asia’s AI and startup ecosystem.

Manus gained a wide following after previewing an AI agent in March that was capable of producing detailed research reports and building custom websites, using AI models developed by companies such as Anthropic and China’s Alibaba. That demo followed the release of DeepSeek, a made-in-China AI model that rocked Silicon Valley due to its advanced capabilities, coupled with claims by its developer that it was developed with far less computing power than American rivals.

The deal is a move in a new direction for Meta, which is investing aggressively in AI to compete with Google, Microsoft and OpenAI. The deal would help the social-media giant cement its position in the product segment of AI agents, an increasingly intense battlefield of AI companies that make tools to conduct complex tasks with minimal human input. Microsoft has operated a popular AI assistant, Copilot.

Meta says it plans to continue to operate and sell Manus’s service and integrate it into its suite of social-media products. Meta has previously touted so-called open source models that are largely free to access, modify or distribute.

“We plan to scale this service to many more businesses,” Meta said in its announcement about the deal. 

Meta’s existing AI offerings are widely available free in services including Instagram and WhatsApp, and the company has also fully incorporated AI into its advertising in ways that have fattened its bottom line, according to analysts.

After Meta faced unexpected challenges earlier this year while preparing to roll out a new model, Chief Executive Mark Zuckerberg went on a recruiting blitz to build an AI dream team, offering top executives and researchers multimillion-dollar paydays.

The company acquired a 49% stake in startup Scale AI that valued Scale at $29 billion, and Scale founder Alexandr Wang joined the social-media giant as its chief AI officer.

Manus has garnered millions of users since its spring launch, including some who pay subscriptions to use its models for analysis, coding and other tasks.

In April, Manus raised $75 million in a fundraising round led by venture firm Benchmark. As part of the deal, Benchmark’s general partner Chetan Puttagunta joined the company’s board. After the investment, Manus’s valuation was boosted to $500 million, people familiar with the matter said. It also counts HSG, ZhenFund and Tencent as major investors.

In December, the startup announced it had crossed $100 million in annual recurring revenue, eight months after it launched. Around the same time, Meta started negotiating with the company for the acquisition, some of the people said.

“Joining Meta allows us to build on a stronger, more sustainable foundation without changing how Manus works or how decisions are made,” Xiao said in an announcement Monday. Xiao is one of two Chinese co-founders of the company.

The parent company behind Manus, Butterfly Effect, was founded in 2022 and had offices in Beijing and Wuhan. In October 2024, the company started developing Manus. Although most of its researchers and engineers were based in China, Manus was launched outside the country as it used many American AI models that aren’t available in China.

After securing investment from Benchmark, the company officially moved its headquarters to Singapore. U.S. lawmakers have criticized Benchmark for backing an AI company with ties to China. Manus has also shelved its plan to develop a version for the Chinese market, people familiar with the matter said.

Manus has around 100 employees, mainly in Singapore.

Write to Angel Au-Yeung at angel.au-yeung@wsj.com, Raffaele Huang at raffaele.huang@wsj.com and Kate Clark at kate.clark@wsj.com

The Global AI Race

Coverage of advancements in artificial intelligence, selected by the editors

Get WSJ's AI Newsletter

If U.S.-China AI Rivalry Were Football, the Score Would Be 24-18

OpenAI Updates ChatGPT Amid Battle for Knowledge Workers

Disney to Invest $1 Billion in OpenAI and License Characters

AI Hackers Are Coming Dangerously Close to Beating Humans

Copyright ©2025 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8

Appeared in the December 30, 2025, print edition as 'Meta Buys Singapore AI Startup Manus'.

---

What to Read Next

[

Chinese Technology Firms Aim to Raise Over $1 Billion via IPOs

](https://www.wsj.com/business/chinese-technology-firms-aim-to-raise-over-1-billion-via-ipos-15073fbc?mod=WTRN_pos1)

[

Hong Kong’s initial-public-offering market is set to start the new year with a bang, as three Chinese companies line up to list in the city early January.

](https://www.wsj.com/business/chinese-technology-firms-aim-to-raise-over-1-billion-via-ipos-15073fbc?mod=WTRN_pos1)

Continue To Article

---

[

Tech, Media & Telecom Roundup: Market Talk

](https://www.wsj.com/articles/tech-media-telecom-roundup-market-talk-fcce67d0?mod=WTRN_pos2)

[

Find insight on Doximity , Nvidia, CES and more in the latest Market Talks covering Technology, Media and Telecom.

](https://www.wsj.com/articles/tech-media-telecom-roundup-market-talk-fcce67d0?mod=WTRN_pos2)

Continue To Article

---

[

Octopus Energy to Spin Off AI Arm Kraken at $8.65 Billion Valuation

](https://www.wsj.com/business/energy-oil/octopus-energy-to-spin-off-ai-arm-kraken-at-8-65-billion-valuation-dfd3068f?mod=WTRN_pos4)

[

The British renewable-energy startup raised $1 billion in its first stand-alone funding round for the artificial-intelligence platform.

](https://www.wsj.com/business/energy-oil/octopus-energy-to-spin-off-ai-arm-kraken-at-8-65-billion-valuation-dfd3068f?mod=WTRN_pos4)

Continue To Article

---

[

Tech, Media & Telecom Roundup: Market Talk

](https://www.wsj.com/tech/tech-media-telecom-roundup-market-talk-f3a0c0b9?mod=WTRN_pos5)

[

Find insight on Nvidia, Doximity, China’s gaming market and more in the latest Market Talks covering Technology, Media and Telecom.

](https://www.wsj.com/tech/tech-media-telecom-roundup-market-talk-f3a0c0b9?mod=WTRN_pos5)

Continue To Article

---

[

Five Things to Know About Nvidia’s $20 Billion Licensing Deal

](https://www.wsj.com/business/deals/nvidia-20-billion-groq-deal-9f8d3a5b?mod=WTRN_pos6)

[

The chipmaker’s AI-inference agreement with startup Groq indicates growing competition for top talent and technology.

](https://www.wsj.com/business/deals/nvidia-20-billion-groq-deal-9f8d3a5b?mod=WTRN_pos6)

Continue To Article

---

[

If U.S.-China AI Rivalry Were Football, the Score Would Be 24-18

](https://www.wsj.com/tech/ai/the-ai-scorecard-how-the-u-s-built-a-leadand-could-lose-it-to-china-7939603a?mod=WTRN_pos7)

[

If the U.S.-China artificial-intelligence rivalry were a football game, America would be leading at halftime—but it just made a risky trade.

](https://www.wsj.com/tech/ai/the-ai-scorecard-how-the-u-s-built-a-leadand-could-lose-it-to-china-7939603a?mod=WTRN_pos7)

Continue To Article

---

[

This trader says AI IPOs and a falling dollar could drive the S&P 500 to 8,200 next year

](https://www.marketwatch.com/story/this-trader-says-ai-ipos-and-a-falling-dollar-could-drive-the-s-p-500-to-8-200-next-year-90c14fe7?mod=WTRN_pos8)

[

Samantha LaDuc has nailed some big calls on stocks in recent years. She sees another S&P 500 high coming in 2026, but an unhappy Main Street.

](https://www.marketwatch.com/story/this-trader-says-ai-ipos-and-a-falling-dollar-could-drive-the-s-p-500-to-8-200-next-year-90c14fe7?mod=WTRN_pos8)

Continue To Article

---

[

Scotland Reigned as the U.K.’s Best-Performing Luxury Housing Market of 2025

](https://www.mansionglobal.com/articles/scotland-reigned-as-the-u-k-s-best-performing-luxury-housing-market-of-2025-dec8c8aa?mod=WTRN_pos9)

Prices in Edinburgh ticked up, while high-end home values declined almost everywhere else, including prime London 

Continue To Article

---

---

Videos