Mercor pursues $20 billion valuation as it acquires Deeptune to control AI training stack

Mercor Seeks to Double Valuation to $20 Billion

Mercor, the AI training marketplace that connects domain experts with major AI labs, is in early discussions to raise funding at a $20 billion valuation, according to Bloomberg sources 1

1

4

4

. The AI training data company has told investors it already holds at least one term sheet at that price, representing a dramatic doubling from its $10 billion valuation achieved just nine months ago during its $350 million Series C round in October 1

1

. While the fundraising talks remain at early stages and terms could change, the speed of this potential jump signals investor appetite for companies that enhance AI models with specialized data.

CEO Brendan Foody announced on X that the company's annualized revenue run rate crossed $2 billion in June, marking a 100% increase in just four months 1

1

4

4

. However, this figure represents gross billings rather than net revenue. Contractors who perform the actual AI model training work take home 60 to 70 percent of total billings, meaning Mercor's actual revenue sits closer to $600 million to $800 million 2

2

. At a $20 billion valuation, this implies a multiple of roughly 25 to 33 times net revenue—aggressive pricing even for a high-growth AI startup.

Deeptune Acquisition Raises Conflict-of-Interest Questions

On the same day as the valuation news broke, Mercor announced the Deeptune acquisition, a startup that builds simulated environments where AI agents practice real-world tasks before deployment 3

3

. Deeptune CEO Tim Lupo describes these as "training gyms"—reinforcement learning spaces that replicate enterprise software like spreadsheets and Salesforce, allowing agents to learn without touching real systems 3

3

. The entire Deeptune team is joining Mercor in New York, though financial terms were not disclosed 1

1

.

The deal carries a notable wrinkle. Brendan Foody wrote a personal angel check into Deeptune's $43 million Series A round led by Andreessen Horowitz in March, just four months before his company acquired the startup 2

2

3

3

. Foody told Fortune the investment was made with acquisition intent: "It was in a lot of ways the main motivation, actually" 3

3

. Public reporting has not confirmed whether Mercor's board or outside investors reviewed this conflict-of-interest before the deal closed 2

2

.

Building Training Environments for AI Agents to Control the Stack

Foody framed the purchase as strategic positioning around where bottlenecks in AI training now exist. "Reinforcement learning has reached the point where a model can learn almost any task that can be clearly defined and scored," he wrote in a blog post. "The constraint has shifted to the environments themselves: the places where models practice the work and get measured on whether they did it well" 3

3

.

Mercor now aims to own the complete AI training stack. Its network of more than five million domain experts—including engineers, lawyers, and doctors—writes tasks and verifies whether agents complete them correctly 2

2

3

3

. Its APEX benchmarks score model performance across real-world workflows. Deeptune supplies the simulated software those tasks run inside, having recreated hundreds of enterprise applications over two years 3

3

. Customers include OpenAI, Anthropic, and Google 2

2

.

Data Breach Casts Shadow Over Growth Story

The valuation discussions arrive less than four months after a significant security incident. In March, attackers exploited malware in the open-source LiteLLM library to access Mercor's systems 3

3

4

4

. The hacking group Lapsus$ later claimed to hold four terabytes of data, including source code, user databases, contractor Social Security numbers, passport scans, interview recordings, and facial biometrics 2

2

3

3

. Contract workers have since filed class-action lawsuits 1

1

2

2

.

Meta, a customer at the time, paused all work with Mercor indefinitely following the data breach 2

2

4

4

. Mercor maintains the impacts were "very limited" based on investigation findings 4

4

. Foody told Fortune that OpenAI and Anthropic remained customers and that "every frontier lab has expanded their relationship with us since the data breach" 3

3

. The revenue doubling in subsequent months suggests either strong customer loyalty or limited alternatives for labs requiring training data at this scale 2

2

.

Market Context and What to Watch

Mercor operates in an increasingly crowded and well-funded space. Scale AI commands a valuation around $29 billion since Meta took a stake, while Surge AI is reportedly raising near $25 billion 2

2

. The three founders—former high-school debate teammates who dropped out of college—became the world's youngest self-made billionaires at 22 2

2

.

Whether the $20 billion figure represents fair value or bubble territory depends on perspectives around gross versus net revenue, the lasting impact of the security incident, and governance questions around founder investments in acquisition targets. Investors will need to assess if the growth trajectory justifies the valuation once contractor payouts, breach consequences, and conflict-of-interest concerns are factored in. The outcome will signal how the market values control over training environments for AI agents as frontier AI models grow more capable and data-hungry.