Google, Scale AI's largest customer, plans split after Meta deal
Alphabet's Google, the largest customer of Scale AI, plans to cut ties with Scale after news broke that rival Meta is taking a 49% stake in the AI data-labeling startup, five sources familiar with the matter told Reuters.
Google had planned to pay Scale AI about $200 million this year for the
human-labeled training data
that is crucial for developing technology, including the sophisticated AI models that power Gemini, its ChatGPT competitor, one of the sources said.
The search giant already held conversations with several of Scale AI's rivals this week as it seeks to shift away much of that workload, sources added.
by Taboola
by Taboola
Sponsored Links
Sponsored Links
Promoted Links
Promoted Links
You May Like
最強のヒーローチームを編成する力はありますか?
レイドシャドウレジェンド
今すぐインストール
Undo
Scale's loss of significant business comes as Meta takes a big stake in the company, valuing it at $29 billion. Scale was worth $14 billion before the deal. Scale AI intends to keep its business running while its CEO, Alexandr Wang, along with a few employees, move over to Meta. Since its core business is concentrated around a few customers, it could suffer greatly if it loses key customers like Google.
In a statement, a Scale AI spokesperson said its business, which spans work with major companies and governments, remains strong, as it is committed to protecting customer data. The company declined to comment on specifics with Google.
Live Events
Scale AI raked in $870 million in revenue in 2024, and Google spent some $150 million on Scale AI's services last year, sources said. Other major tech companies that are customers of Scale's, including Microsoft, are also backing away. Elon Musk's xAI is also looking to exit, one of the sources said. OpenAI decided to pull back from Scale several months ago, according to sources familiar with the matter, though it spends far less money than Google. OpenAI's CFO that the company will continue to work with Scale AI, as one of its many data vendors.
Discover the stories of your interest
Blockchain
5 Stories
Cyber-safety
7 Stories
Fintech
9 Stories
E-comm
9 Stories
ML
8 Stories
Edtech
6 Stories
Companies that compete with Meta in developing cutting-edge AI models are concerned that doing business with Scale could expose their research priorities and road map to a rival, five sources said. By contracting with Scale AI, customers often share proprietary data as well as prototype products for which Scale's workers are providing data-labeling services. With Meta now taking a 49% stake, AI companies are concerned that one of their chief rivals could gain knowledge about their business strategy and technical blueprints.
Google, Microsoft and OpenAI declined to comment. xAI did not respond to a request for comment.
Rivals see openings
The bulk of Scale AI's revenue comes from charging generative AI model makers for providing access to a network of human trainers with specialized knowledge - from historians to scientists, some with doctorate degrees. The humans annotate complex datasets that are used to "post-train" AI models, and as AI models have become smarter, the demand for the sophisticated human-provided examples has surged, and one annotation could cost as much as $100.
Scale also does data-labeling for enterprises like self-driving car companies and the US government, which are likely to stay, according to the sources. But its biggest money-maker is in partnering with generative AI model makers, the sources said.
Google had already sought to diversify its data service providers for more than a year, three of the sources said. But Meta's moves this week have led Google to seek to move off Scale AI on all its key contracts, the sources added. Because of the way data-labeling contracts are structured, that process could happen quickly, two sources said.
This will provide an opening for Scale AI's rivals to jump in.
"The Meta-Scale deal marks a turning point," said Jonathan Siddharth, CEO of Turing, a Scale AI competitor. "Leading AI labs are realizing neutrality is no longer optional, it's essential."
Labelbox, another competitor, will "probably generate hundreds of millions of new revenue" by the end of the year from customers fleeing Scale, its CEO, Manu Sharma, told Reuters.
Handshake, a competitor focusing on building a network of PhDs and experts, saw a surge of workload from top AI labs that compete with Meta.
"Our demand has tripled overnight after the news," said Garrett Lord, CEO at Handshake.
Many AI labs now want to hire in-house data-labelers, which allows their data to remain secure, said Brendan Foody, CEO of Mercor, a startup that in addition to competing directly with Scale AI also builds technology around being able to recruit and vet candidates in an automated way, enabling AI labs to scale up their data labeling operations quickly.
Founded in 2016, Scale AI provides vast amounts of labeled data or curated training data, which is crucial for developing sophisticated tools such as OpenAI's ChatGPT.
The Meta deal will be a boon for Scale AI's investors including Accel and Index Ventures, as well as its current and former employees.
As part of the deal, Scale AI's CEO, Wang, will take a top position leading Meta's AI efforts. Meta is fighting the perception that it may have fallen behind in the AI race after its initial set of Llama 4 large language models released in April fell short of performance expectations.
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Economic Times
19 minutes ago
- Economic Times
Apple Paper questions path to AGI, sparks division in GenAI group
New Delhi: A recent research paper from Apple focusing on the limitations of large reasoning models in artificial intelligence has left the generative AI community divided, sparking significant debate whether the current path taken by AI companies towards artificial general intelligence is the right one to take. The paper, titled The Illusion of Thinking, published earlier this week, demonstrates that even the most sophisticated large reasoning models do not genuinely think or reason in a human-like way. Instead, they excel at pattern recognition and mimicry, generating responses that only appear intelligent, but lack true comprehension or conceptual understanding. The study used controlled puzzle environments, such as the popular Tower of Hanoi puzzle, to systematically test reasoning abilities across varying complexities by large reasoning models such as OpenAI's o3 Mini, DeepSeek's R1, Anthropic's Claude 3.7 Sonnet and Google Gemini Flash. The findings show that while large reasoning and language models may handle simple or moderately complex tasks, they experience total failure when faced with high-complexity problems, which occur despite having sufficient computational resources. Gary Marcus, a cognitive scientist and a known sceptic of the claims surrounding large language models, views Apple's work as providing compelling empirical evidence that today's models primarily repeat patterns learned during training from vast datasets without genuine understanding or true reasoning capabilities. "If you can't use a billion-dollar AI system to solve a problem that Herb Simon (one of the actual godfathers of AI, current hype aside) solved with AI in 1957, and that first semester AI students solve routinely, the chances that models like Claude or o3 are going to reach AGI seem truly remote," Marcus wrote in his blog. Marcus' arguments are also echoed in earlier comments of Meta's chief AI scientist Yann LeCun, who has argued that current AI systems are mainly sophisticated pattern recognition tools rather than true thinkers. The release of Apple's paper ignited a polarised debate across the broader AI community, with many panning the design of the study than its findings.A published critique of the paper by researchers from Anthropic and San-Francisco based Open Philanthropy said the study has issues in the experimental design, that it overlooks output an alternate demonstration, the researchers tested the models on the same problems but allowed them to use code, resulting in high accuracy across all the tested models. The critique around the study's failure to take in the output limits and the limitations in coding by the models have also been highlighted by other AI commentators and researchers including Matthew Berman, a popular AI commentator and researcher."SOTA models failed The Tower of Hanoi puzzle at a complexity threshold of >8 discs when using natural language alone to solve it. However, ask it to write code to solve it, and it flawlessly does up to seemingly unlimited complexity," Berman wrote in a post on X (formerly Twitter).The study highlights Apple's more cautious approach to AI compared to rivals like Google and Samsung, who have aggressively integrated AI into their products. Apple's research explains its hesitancy to fully commit to AI, contrasting with the industry's prevailing narrative of rapid questioned the timing of the release of the study, coinciding with Apple's annual WWDC event where it announces its next software across online forums said the study was more about managing expectations in light of Apple's own struggles with said, practitioners and business users argue that the findings do not change the immediate utility of AI tools for everyday applications.
Time of India
30 minutes ago
- Time of India
'Neuralink babies'? Scale AI's Alexandr Wang says he is waiting for Elon Musk's brain chips before having kids
In a statement that straddles science fiction and near-future reality, Scale AI founder Alexandr Wang has revealed he's putting off parenthood—for now. But not because of career demands or personal timing. His reason? He's waiting for Elon Musk 's Neuralink to become mainstream. Yes, Wang wants his future children to be among the first humans enhanced by brain-computer interfaces from birth. During a recent appearance on The Shawn Ryan Show , the 28-year-old tech prodigy shared a vision that feels pulled from the pages of a futuristic novel. 'When we get Neuralink and we get these other technologies, kids who are born with them are gonna learn how to use them in like crazy, crazy ways,' Wang said, explaining that the first seven years of life—when neuroplasticity is at its peak—present the most fertile ground for integrating superintelligence into the human experience. Neuralink, Meet Nature Neuralink, founded by Elon Musk, is currently trialing a brain-chip implant the size of a coin. Though still in early clinical stages, the device has already shown stunning potential: one patient with ALS reportedly edited a video using only his mind. But Neuralink isn't alone. Synchron, backed by heavyweights like Jeff Bezos and Bill Gates, is collaborating with Apple to help patients with disabilities use iPhones through brain signals. Motif Neurotech, another contender, is developing a neurostimulator that treats severe depression and functions like a pacemaker for the brain. by Taboola by Taboola Sponsored Links Sponsored Links Promoted Links Promoted Links You May Like Buy Brass Idols - Handmade Brass Statues for Home & Gifting Luxeartisanship Buy Now Undo Wang, who is also taking on a new role at Meta to lead its superintelligence initiatives, seems to believe these brain-machine hybrids are not just medical miracles—they are the future of human learning, cognition, and possibly even evolution. Born to Compute? His vision hinges on a well-documented trait: the astonishing neuroplasticity of young brains. A 2009 study published in Brain Dev. found that children's brains, particularly in the early years, are primed for adaptation. This plasticity not only helps kids learn languages or recover from injury but, in Wang's vision, could also help them learn how to "think" alongside or even through artificial intelligence. It's a radical idea—one that flips the conventional approach to parenting. Instead of shielding children from screen time or tech overload, Wang imagines a future where babies are born wired for the digital age, quite literally. Ethics, Science, and the Silicon Valley Dream As startling as Wang's perspective may seem, it's emblematic of a growing mindset in tech circles: that human limitations are solvable problems. But while Wang may be planning for AI-enhanced progeny, ethical concerns continue to hover over Neuralink and its competitors—from long-term brain health to consent, privacy, and the ever-blurring boundary between human and machine. Still, in a world racing toward a post-human horizon, Alexandr Wang's statement isn't just provocative—it might be prophetic. The question isn't whether Neuralink babies will happen. It's who dares to go first. And Wang, it seems, is ready to raise the world's first AI-native child—as soon as the software is ready.
Time of India
30 minutes ago
- Time of India
India, US close to finalising 'early tranche' of trade deal
File photo: Union minister Piyush Goyal with US secretary Howard Lutnick (Photo: ANI) NEW DELHI: India and the US are close to working out the final contours of the "early tranche" of the proposed bilateral trade agreement and are now starting work on the possible text for the deal, which is likely to be finalised before July 9, when American president Donald Trump plans to remove the freeze on reciprocal tariffs. Govt is likely to offer concessions for some of the farm products that are of interest to the US, along with concessions on automobiles, in return for getting a 10 per cent duty - Trump's baseline tariff - for labour-intensive products, such as textiles, footwear and, possibly, certain auto parts. With the current US administration lacking Congressional mandate to finalise trade deals, a zero-duty regime will have to wait until things settle down in the US. With a 10 per cent levy on products where India has a comparative advantage, the government is hoping to make Indian exports more competitive, especially if the reciprocal tariffs kick in. In any case, India, with 26 per cent reciprocal tariffs, including the 10 per cent baseline duty, was seen to enjoy a competitive advantage over several countries with which its exports compete. Besides, with sanitary and phytosanitary standards and technical barriers to trade also discussed during the last round of negotiations with American officials between June 4 and 10, the government is hoping that some of the issues faced by exporters shipping goods to the US will also be addressed. Digital trade, customs and trade facilitation and the legal framework were part of the discussions in the Capital earlier this month. Government sources indicated that India is keen for a text to be ready in some form before the deal is made public. This is seen to be crucial as the current US administration has been making announcements without the draft text in place and then claiming victory. Even the Chinese authorities have had to contend with the claims made by Trump on social media after the last dialogue. The announcement with the UK was seen to be more structured. By all accounts, several issues, including the court ruling on reciprocal tariffs for which Trump had invoked emergency provisions, will, however, play a part in finalising the early tranche for which a Fall (Sept/ Oct) date had been fixed. Besides, India is keen that an arrangement be worked out where it retains an advantage over its peers in case the tariffs are declared illegal by American courts. Stay informed with the latest business news, updates on bank holidays and public holidays . AI Masterclass for Students. Upskill Young Ones Today!– Join Now
