The Horizon of Expertise

First came the race for computing power.

We’re spending $400 billion on GPU infrastructure build out this year, and that number will rise (opens in new tab) to $500 billion next.

Then came the race for talent. The $250 million pay packages (opens in new tab), the scramble for acquihires, the NBA-ification of tech talent markets.

Each race has followed the latest constraint in the largest industrial build out in history. Because AI is no mere technology cycle. AI is railroads, electrification, broadband.

Consider the fact: Gartner (opens in new tab) estimates that the global enterprise SaaS market is worth nearly $1.5 trillion. But that’s only 10% of overall corporate spending. The pull of AI isn’t just subsuming that 10% line item; it’s doubling, tripling, potentially quadrupling it. AI is an accelerant to the labor that eats up nearly 80% of all corporate spending, making labor costs more efficient and more valuable. The stakes and opportunity unlocked by that fact are enormous; in the next decade, we may see the first ten trillion dollar company. AI investment is scaled to the value it can produce. As Larry Page has apparently (opens in new tab) told his internal teams at Google, "I am willing to go bankrupt rather than lose this race."

But there’s a new race beginning to emerge, and the realization of the value from all of that GPU and researcher investment depends on its outcome. We’re learning what differentiates economically useful AI and just another slop machine, and it’s not computing power or your research bench. It’s the quality of your data.

The New AI Race

Early foundational models scraped all of their data for free from the internet. This has made them powerful, but not particularly good at delivering economic value. The difference between where AI is and where it needs to be is the difference between deciding how to launch a new product based on what a keyword-stuffed marketing blogpost says, and based on guidance from an early PM who has scaled a few unicorns.

Our economy runs on the expertise that lives within people’s heads. It’s why knowledge matters. It’s why people are paid more as they grow in their careers. And AI will need to demonstrate that same expertise in order to meaningfully deliver economic value.The so-called “browser wars” (opens in new tab) are an entrant to this new race. The bet is that by seeing the same emails you see and how you work across tools, AI-enabled browsers will be able to capture the data they need to provide real, economically valuable intelligence.

But expertise is more than the sum of our clicks and tabs. This is why Mercor (opens in new tab)’s approach is currently winning the data and reinforcement learning race. At its core, Mercor is a company that believes in human expertise. Mercor works directly with real experts at the top of their fields to learn about how they solve problems: the context that guides their decisions, the taste that informs what they make, and the workflows that they navigate. It is rapidly becoming the place where AI learns to do real, economically valuable work.

The Horizon of Expertise

But the opportunity Mercor unlocks is bigger than just realizing the possibility of AI’s potential economic gains in the existing workforce. It unlocks a new horizon for what’s possible in human progress.

Between 1945 and 2010, the volume of “disruptive breakthroughs (opens in new tab)” in science and technology declined dramatically. It’s a trend that runs contrary to spending on innovation, overall educational achievement rates, and the growing amount of knowledge.

This is because breakthroughs demand more expertise now than they did before. Epiphany doesn’t just drop out of trees anymore; most observable breakthroughs were long ago added to the body of human knowledge. As our collective knowledge grows, so too does the time any domain expert must spend consuming that knowledge. To get to a point where we’re moving beyond what’s currently known or understood in any given field now takes decades of learning. The floor for breakthrough has begun to exceed a single human lifetime’s capacity for expertise.

Mercor allows us to reach the horizon of expertise much more rapidly, and to look beyond, where the possibility for progress lies.

Our Series C Investment in Mercor

Our conviction in Mercor took root long before there was consensus on the expertise opportunity. Because even more important than the size of the opportunity is the founding team behind it. Brendan is a visionary gifted with the natural magnetism of generational leaders. Adarsh is technically brilliant, a heat-seeking missile for the most important problems. And Surya’s a deep thinker at home in complexity, always 10 years ahead of everyone else. But they’re more than the sum of their qualities. Maybe it’s because they’ve known each other since childhood and grew up on their high school debate team together. But when we first met the team, we knew we’d found something explosive.

When we led Mercor’s Series B earlier this year, it was clear they were building something far bigger than anyone anticipated: a new category of work for knowledge workers. Just months later, that conviction has only deepened. Mercor has now raised a $350 million Series C at a $10 billion valuation, a 5x step-up from February, with Felicis proudly investing alongside Benchmark, General Catalyst, and new investors including Robinhood Ventures.

Mercor’s momentum speaks for itself. The company has averaged more than $1.5 million paid per day to its network of 30,000 experts. Beyond redefining how AI models are trained, Mercor is creating an entirely new marketplace for work, with experts earning an average of $85 per hour.

They don’t necessarily need the capital. Unlike a lot of rapidly accelerating AI companies, Mercor is profitable, and has never churned an enterprise customer. But the road ahead is long, and when the opportunity is this big, it’s still early. We believe that Mercor is 1% of the way to the new horizon of expertise– and we’re thrilled to partner with them on the rest of the journey ahead.