Within the sea change now sweeping across the world of AI training and data annotation services, two names are standing tall: the better-known incumbent, Scale AI, and its challenger, Mercor, cofounded and led by Brendan Foody. Foody’s recent public commentary on Scale AI opens up a window into how he sees the competitive battleground-and more importantly, how he believes Mercor is doing things differently.
From Startup to Challenger
Brendan Foody co-founded Mercor in 2023 with Adarsh Hiremath and Surya Midha, positioning the company as a marketplace which supplies top-tier human experts-engineers, doctors, lawyers, analysts-to train foundational AI models.
Per TechCrunch, Mercor was already eyeing a valuation of $10 billion+, with run-rate revenues around $450-$500 million.
Foody has insisted that Mercor is building more than just a staffing business—he calls it “a new category of work” where human judgment, nuance, and domain expertise remain central to advancing AI capabilities.
The Critique of Scale AI
On the “20VC” podcast and other forums, Foody hasn’t minced his words in his criticism of the approach taken by Scale AI. He acknowledged that Scale’s former CEO, Alexandr Wang, was “phenomenal” at distribution and sales. At the same time, however, Foody felt Scale lost sight of two very important dimensions: product-quality and scaling quality.
In his words:
“In some ways, Scale lost the focus on product, on scaling quality.
He connected this to compensation and talent strategy. Foody claims that Mercor pays considerably more to its annotators and contractors – somewhere around US$95/hour on average – compared to what Scale and other rivals reportedly pay, at about US$30/hour.
The implication is that by recruiting higher-tier talent and paying them better, Mercor is targeting a higher-quality output, with Foody implying Scale perhaps compromised on that front in favor of scale.
Why This Matters: Quality, Trust & Execution
Foody’s critique represents much more than mere competitive posturing. It touches upon very real issues in the AI training industry:
Quality of data/annotations: AI models are only as good as human-in-the-loop data. Mistakes, shortcuts, or poorly qualified annotators can degrade model performance, introduce bias, or even expose security vulnerabilities.
Security and governance: Scale AI came under scrutiny for reportedly using insecure tools like public Google Docs for high-profile client work, and after a major investment from Meta Platforms.
Scaling versus sustaining quality: While rapid growth can be alluring, if operations begin to outrun governance or talent standards, the business builds up operational and reputational risk.
Talent moat: Foody believes that recruiting high-quality specialists-e.g., Olympiad medalists, Rhodes Scholars-creates a talent moat for Mercor that will always be difficult to replicate over a long period.
While Foody leads, Mercor focuses on:
Elite talent sourcing & compensation: Pay rates and talent standards are tailored to attract domain experts, not general crowd-workers.
Growth with discipline: They say that they’ve experienced tremendous growth-from $1 M to $500 M in run-rate revenue in 17 months-while they have kept their standards high.
Positioning for next-gen AI training: Beyond “simple labeling”, Mercor focuses on high-complexity tasks like RLHF, expert evaluations, and domain-specific workflows, seeing a future wherein humans will be teaching AI how to handle nuance.
New world of work: The company positions itself as building a “new category of work” for domain experts, not just low-cost annotation.
The Rivalry: What It Signals for Founders & Investors The public jabs between Foody and Scale AI reflect deeper shifts within the AI ecosystem. The business model for training AI is maturing – it’s not just about scale or low cost; it’s about quality, repeatability, reliability. Investors and clients alike are paying more and more attention to sustainability. How will these annotation businesses endure under increased scrutiny regarding data practices, security, and model trust?
According to Foody, chasing scale without the underlying fundamentals could just backfire. The competitive pressure from startups like Mercor actually forces the incumbents to sharpen, which can benefit the overall market in terms of quality and transparency. A Note of Caution While Foody’s critique is pointed, it’s also worth noting that Mercor is younger and smaller than Scale AI and faces its own challenges in scaling, maintaining margins, and delivering long-term value. Moreover, Scale AI’s market presence and client base remain formidable. So Foody’s comments should be seen as one side of a competitive narrative-not a definitive verdict on Scale AI’s trajectory. Conclusion Brendan Foody’s comments about Scale AI aren’t rivalry talk; it’s a belief in a model that focuses on premium talent, high-quality output, and long-term differentiation in the AI training business. For Mercor, the objective seems to be to grow fast but not just big-grow well.
