Scale AI funding: Every round, investor, and valuation (2026)

Scale AI sells the human-made data that AI models learn from. Its contractors label images, write and grade answers, and test chatbots for the labs that build frontier models. It also sells AI applications and model evaluation tools to companies and governments, including the US military.

Scale has raised about $15.9 billion across eight rounds since 2016, all of it equity. Almost all of that came in one deal in June 2025, when Meta bought a 49% non-voting stake that valued Scale at more than $29 billion.

That deal was not a normal funding round. Scale paid much of Meta's money out to its shareholders, its founder left to run Meta's AI lab, and several of its largest customers pulled back. Let's dive into Scale AI's funding history.

Scale raised about $1.6 billion in its first eight years, almost all of it from venture and growth firms. Meta's investment was close to nine times that amount in a single transaction.

Alexandr Wang was 19 when he left MIT to start Scale with Lucy Guo, a Thiel Fellow who had left Carnegie Mellon. The pair went through Y Combinator's summer 2016 batch, which invested a standard check reported at about $120,000.

The first product was an API for human labor. A customer sent raw data, and Scale's network of contractors sent it back labeled. That made Scale useful to self-driving car programs, which needed millions of camera and lidar frames tagged by hand.

Accel led the $4.5 million Series A in July 2017, and partner Dan Levine joined the board. Index Ventures led an $18 million Series B in August 2018, and Mike Volpi took a board seat. By then Scale was routing work to about 10,000 contractors, according to TechCrunch, and Guo left the company the same year.

Founders Fund led a $100 million Series C in August 2019 that pushed Scale past a $1 billion valuation. Coatue, Spark Capital, and Thrive Capital joined, along with Instagram co-founders Kevin Systrom and Mike Krieger and Quora's Adam D'Angelo.

Tiger Global led the Series D in December 2020. Wang told CNBC that Scale was breaking even after it slowed hiring, and the company announced its first acquisition, Helia, alongside the round.

Four months later, Dragoneer, Greenoaks, and Tiger Global co-led a $325 million Series E. Fortune reported the valuation at $7.3 billion, and total funding passed $600 million. Scale did not raise again for three years.

The pause ended with ChatGPT. Large language models needed a different kind of data: written answers, ranked responses, and expert feedback rather than boxes drawn around cars. Scale moved its contractor network to that work through its Remotasks and Outlier platforms, and it sold the results to OpenAI, Google, and Meta.

Accel returned to lead a $1 billion Series F in May 2024. Part of the round was secondary, which let existing holders sell shares, though Scale did not disclose the split. The investor list showed who depended on the data: Nvidia, Amazon, Meta, Intel Capital, AMD Ventures, Cisco Investments, and ServiceNow Ventures all bought in.

In March 2025, Reuters reported that Scale was exploring a tender offer at a valuation of up to $25 billion. Meta made a larger offer instead. On June 12, 2025, Scale announced a "significant new investment" from Meta, and reports put the price at about $14.3 billion for 49% of the company.

The structure matters for what "funding" means here. Meta bought newly issued shares that carry no votes, and Meta's quarterly filing describes "a non-voting minority" of Scale's equity. Meta did not get a board seat.

Scale then distributed much of the cash to its shareholders and vested equity holders, who kept their shares. Bloomberg reported that Accel alone stood to receive about $2.5 billion, and Forbes reported that most employees received a dividend worth about half of their equity package. Scale has not disclosed how much of the money it kept.

Wang left to lead Meta Superintelligence Labs and took about 10 Scale employees with him, though he stayed on Scale's board. Jason Droege, Scale's chief strategy officer and a former Uber Eats executive, became interim CEO.

Scale's other big customers treated Meta's stake as a conflict. Reuters reported that Google, whose search advertising business funds its Gemini models, planned to cut ties, even though it was Scale's largest customer. OpenAI, whose subscription and API business relies on this kind of training data, had already been winding down work, and Forbes later called it Scale's most high-profile loss.

In July 2025, Scale cut about 200 full-time staff, 14% of the total, and ended work with 500 contractors. Droege said the company had grown its generative AI data business too quickly. Revenue for 2025 came in just under $1 billion, about half the target Bloomberg had reported before the deal.

Rivals picked up some of the work. Meta's own researchers turned to Surge AI and Mercor for some projects in 2025, according to TechCrunch, and Snorkel AI reached a $3.5 billion valuation in 2026 selling data to the same labs.

Scale has not raised money since. Google resumed some work within months, according to Forbes, and Droege said most of the customers that left have come back, "but not all of them." Francis deSouza, formerly chief operating officer of Google Cloud, became CEO in August 2026, and Wang is now chairman.

Scale's valuation roughly doubled with every priced round after 2020. The pattern held through the Meta deal, which is the last time anyone set a price.

The Series D lifted the price about 3.5 times in 16 months, the largest jump in Scale's history. Each of the next three marks came in at roughly twice the last one, even with a three-year gap between the Series E and the Series F.

Meta paid about 2.1 times the Series F price a year later. Paying $14.3 billion for 49% implies a value of about $29.2 billion, and Meta booked $13.79 billion of the deal as an equity investment in its June 2025 quarter.

The deal also reset the cap table. Because Meta bought new shares, every existing holder's percentage was roughly halved, a change covered in how Scale AI's ownership is split. Holders were compensated through the dividend rather than by selling shares.

No round has tested that price since. Against 2025 revenue of just under $1 billion, it equals almost 30 times sales, for a business whose largest customers spent part of 2025 walking away.

Scale's investors fall into three groups: the venture firms that led its early and growth rounds, the technology companies that joined the Series F, and Meta, which now owns nearly half the company.

Meta first invested in the Series F, as one of several tech companies buying a small stake. A year later it became the largest shareholder by far, with a stake that carries no votes and no board seat. Scale says Meta gets no access to its business information or to customers' data.

What Meta bought was a leader and a supplier. Wang now runs Meta's AI effort as chief AI officer, and Forbes reported that Meta agreed to pay Scale at least $450 million a year for five years, a figure neither company has confirmed.

The terms mirror Meta's own governance. Meta took economic ownership of Scale without control, while its own shareholders hold economic ownership without control under Mark Zuckerberg's dual-class voting structure.

Accel is Scale's longest-standing institutional backer. It led the Series A in 2017 and the Series F in 2024, and Dan Levine still sits on the board. Bloomberg estimated Accel's share of the Meta dividend at about $2.5 billion.

Index Ventures led the Series B, and Mike Volpi has served on the board since. Volpi told Forbes that an eventual IPO could give investors a "second exit" after the Meta payout.

Founders Fund led the Series C, the round that made Scale a unicorn. Tiger Global led the Series D and co-led the Series E with Dragoneer and Greenoaks, during the years when Tiger was writing large checks across late-stage software.

Wellington Management and Durable Capital entered in the Series E. Coatue, Thrive Capital, and Spark Capital returned across several rounds, and most of the 2021 investors came back for the Series F.

The Series F brought in companies that sell to the same AI labs Scale serves. Nvidia, listed as a returning investor, sells the chips these models train on. Amazon joined as a new investor, and its AWS cloud business rents that computing power to the same labs.

Intel Capital, AMD Ventures, Cisco Investments, and ServiceNow Ventures also joined. None of them disclosed a stake. For the chip and cloud companies, a small position in Scale was a bet on demand for training data, and therefore on demand for the hardware that runs it.

Wang held about 15% of Scale after the Series E, according to Forbes, and Guo held just under 5% before the Meta deal. Meta's new shares diluted both stakes, though both founders also received the dividend.

Early angels included OpenAI co-founder Greg Brockman, former GitHub CEO Nat Friedman, Dropbox co-founder Drew Houston, and Twitch co-founder Justin Kan. Investor Elad Gil joined in the Series F.

Scale spends its money on three things: the people who make its data, products for governments and enterprises, and payouts to its shareholders. The balance between them shifted after the Meta deal.

Scale's main cost is people. Through Outlier and Remotasks, it pays contractors around the world to label data, write answers, and grade model outputs. As labs ask for harder material, more of that work goes to specialists in fields such as coding and math, who command higher rates.

That workforce carries legal risk. A California class action claiming contributors were misclassified as contractors reached a proposed $12.5 million settlement in 2026. The US Department of Labor dropped a separate investigation in May 2025.

Defense is the business Scale has pushed hardest since the Meta deal. Its Donovan platform runs on classified networks, and Scale is the prime contractor on Thunderforge, a Pentagon program that uses AI for military planning, with Anduril and Microsoft as partners.

The contracts are getting larger. Scale won a $99 million Army research deal in August 2025, and a Pentagon agreement for data analysis and decision support rose to a $500 million ceiling in May 2026. A ceiling is the most the Pentagon can spend, not booked revenue.

In April 2026, Scale bought ICG Solutions, a defense company that analyzes streaming data in real time. The deal puts Scale in more direct competition with Palantir's government-focused business model for the same budgets.

The second bet is software. Scale sells a generative AI platform and custom applications to companies such as Mayo Clinic, BP, and Allianz. Droege expects the applications business, which ended 2025 at about $200 million in annualized revenue, to pass data labeling within 18 months.

That push needs people. Scale hired more than 500 staff in 2025 and opened offices in London, Doha, and St. Louis. Forbes put headcount at about 1,300 in May 2026, with 500 more hires planned.

The largest use of Meta's money was outside the company. Scale distributed proceeds to shareholders and vested employees and said the rest would go to research and strategic partnerships. It does not publish its cash balance or burn rate, though it says its data business is now profitable.

Francis deSouza's job is to make Scale less dependent on a few AI labs. The plan leans on the US government, foreign governments, and large enterprises in health care, energy, and insurance. Scale says international public-sector revenue doubled in 2025, and it expects total revenue to top $1 billion in 2026, with applications and international sales roughly doubling.

The product shift is from selling data to selling software and evaluations, and Scale's SEAL research lab also runs public model benchmarks such as Humanity's Last Exam. In data, the main rivals are Surge AI, DataAnnotation's bootstrapped parent, which reportedly earned more than $1 billion in 2024, and Mercor, whose rapid funding climb came from selling expert data to the same labs.

An IPO is the most likely next pricing event. Droege has said Scale will "very likely" go public at some point but is "super early" in planning. The main risks are the trust of rival labs, labor claims, competition for expert contractors, and dependence on Meta, whose reported commitment would equal close to half of Scale's revenue. Much of Meta's money went to shareholders rather than growth, so the next round or listing will show whether the 2025 price holds.

Scale AI has raised about $15.9 billion across eight rounds. That includes about $1.6 billion from a 2016 Y Combinator seed through a $1 billion Series F in 2024, plus Meta's investment of about $14.3 billion in June 2025. Much of the Meta money was paid out to shareholders as a dividend.

Meta is Scale AI's largest investor, with a 49% non-voting stake. Other major backers include Accel, Index Ventures, Founders Fund, Tiger Global, Dragoneer, Greenoaks, Coatue, Thrive Capital, Wellington, Nvidia, Amazon, Intel Capital, AMD Ventures, Cisco Investments, and ServiceNow Ventures.

Scale AI was valued at more than $29 billion in Meta's June 2025 investment. That is up from $13.8 billion in the May 2024 Series F and about $7.3 billion in the 2021 Series E. No round has set a new price since.

No. Scale AI is private and has not filed for an IPO. Jason Droege, then interim CEO, said in May 2026 that the company will very likely go public at some point but is early in planning. Meta's stake is a minority investment, not an acquisition.

Scale AI is owned by Meta, which holds about 49% of the equity but no votes, along with its founders, employees, and venture investors. Co-founder Alexandr Wang is chairman, Francis deSouza is CEO, and Accel, Index Ventures, and Founders Fund are among its longest-standing investors. Exact stakes other than Meta's are not public.