AfterQuery: YC's Fastest-Ever Unicorn at $3.2B

AfterQuery: YC's Fastest-Ever Unicorn at $3.2B

AI training-data startup AfterQuery has reportedly raised a round at a $3.2 billion valuation, according to a report from Forbes first picked up by TechCrunch. That is a more than 10x jump in less than five months — the startup only announced its $30 million Series A at a $300 million valuation back in April.

The exact amount of capital raised in this new round has not been disclosed, and AfterQuery could not be reached for comment. But if the number holds, YC partner Gustaf Alströmer says it is the fastest any startup has gone from launch to unicorn status in the accelerator's history.

What AfterQuery actually does

Founded by two founders who are 22 and 23 years old today — graduates of Y Combinator's Winter 2025 cohort, just 18 months ago — AfterQuery sits in the same lane as Mercor and Scale: employing knowledge professionals to do model training. Doctors, lawyers, finance people, and other specialists get paid to work alongside AI systems and produce training signal.

But there's a subtle difference in positioning. Mercor and Scale mostly focus on making models answer questions accurately — correct labels, good rankings, benchmark-oriented RLHF. AfterQuery says it trains models and agents on how professionals complete tasks, which the company describes as "encoding the patterns, decisions, and reasoning of the world's best practitioners."

In plain terms: instead of "what is the right answer," the data is "what does a senior engineer actually do to get there." That is process supervision rather than outcome supervision — the kind of trajectory-level data that seems to drive both RLHF and RLVR rewards these days, and the same emphasis on task-execution traces you see in agent training labs like EnvHarness.

The shape of that data, highly simplified, looks something like this:

{
  "task": "Triage this production incident",
  "expert_trajectory": [
    { "step": 1, "action": "read_pagerduty", "note": "check severity first" },
    { "step": 2, "action": "query_slack_archives", "note": "correlate deploy window" },
    { "step": 3, "action": "read_logs", "focus": "error_rate trend last 2h" },
    { "step": 4, "decision": "rollback", "reasoning": "deploy at 21:40 correlates" }
  ],
  "outcome": "resolved",
  "why_it_worked": "reasoning chain logged at every decision point"
}

That trajectory is illustrative, not a leaked dataset — but it captures the idea. The model isn't just learning the final action; it's learning the reasoning steps, the order of operations, what a practitioner checks first, and why.

The numbers behind the run-up

When it launched its Series A in April, AfterQuery said it had already crossed a $100 million annualized revenue run rate and named customers including Nvidia, Legora, and the Korean AI lab Motif Technologies. Its investors back that story with some heavyweight signatures: the Series A was led by Altos Ventures with participation from The Raine Group, Y Combinator, BoxGroup, and Latitude Capital, and the cap table reportedly includes angels from Google DeepMind, OpenAI, Anthropic, Meta Superintelligence Labs, and Microsoft AI.

Read that list again. Every frontier lab has an internal interest in better expert-reasoning datasets, and several of their current and former researchers are personally backing a supplier. That is exactly the kind of signal that makes VCs comfortable writing a big number quickly: the customers are the labs, and the labs' own people are the angels.

$100M run rate in 14 months, from a two-person founding team in a Winter 2025 cohort, is legitimate on its own. The valuation story is about the run rate, but not because of it — nobody pays 32x revenue run rate on current growth alone. The bet is that expert-trajectory data becomes the durable moat for how the next generation of agents learns to work.

What to keep your eyes on

Three things worth watching:

1. "Reportedly" is doing heavy lifting. The round is not confirmed by the company, and the amount of new capital is unknown. A valuation is a negotiation, not a bank balance — the actual check could be anything.
2. The moat question. If the differentiation is process-level expertise, then retention is the moat: do doctors, lawyers, and senior engineers keep contributing, or does quality decay as the pool scales? Data-labeling businesses have made this mistake before — scaling headcount fast tends to dilute signal.
3. The safeguards / quality tension we've seen before. Human review in the loop is a feature, not a bug. When you're training models on how experts behave, a cleaning step that edits the expert traces is what keeps the dataset honest — the same lesson as why AI writing pipelines need a human review step.

There is also a workload question for the broader market. If expert-trajectory data becomes the default training input for agentic systems, companies like AfterQuery stop being "data vendors" and start being the infrastructure layer under the whole agentic AI stack — the same layer where frontier labs now govern agent behavior. That's the bet — and at $3.2B, investors are saying they believe the trajectory (pun intended) holds.

Bottom line

AfterQuery is now, reportedly, the fastest launch-to-unicorn story Y Combinator has ever seen: $300M to $3.2B in roughly five months. The unusual part is that the story isn't growth theater — there's a real revenue run rate, real lab customers, and expert-reasoning data that genuinely differs from classic RLHF labeling. Whether the valuation holds depends on whether that data quality survives scale. For now, from $100M run rate to $3.2B valuation, it's the fastest ride in the accelerator's history.

Sources

• Forbes: AfterQuery Becomes YC's Fastest Unicorn At $3.2 Billion
• TechCrunch: AfterQuery reportedly becomes Y Combinator's fastest-ever unicorn, now valued at $3.2B
• AfterQuery: Human Expertise, Reimagined