Signals Inbox·September 2, 2026·AI Infrastructure

Is AfterQuery really worth $3.2B today?

AfterQuery’s $3.2 billion valuation looks aggressive but defensible today: at a conservative $200 million-plus recurring revenue run rate, the company is already at 16 times revenue or less, and its growth, profitability and push into post-training infrastructure make the jump less absurd than the headline suggests.

Q1Did AfterQuery really jump from $300M to $3.2B this fast?

AfterQuery’s $3.2 billion valuation looks credible, but for now it is a reported investor mark rather than a financing round the company has publicly confirmed.

Forbes reported that two people with direct knowledge of the deal put AfterQuery’s valuation at $3.2 billion, with a lead investor already lined up. TechCrunch subsequently reported the same valuation. AfterQuery declined to comment, and neither the size of the new round nor the lead investor has been disclosed yet.

The comparison with AfterQuery’s previous valuation is still remarkable. The company announced a $30 million Series A at a $300 million valuation in April, led by Altos Ventures with participation from The Raine Group, Y Combinator, BoxGroup and Latitude Capital. The new mark is 10.7 times higher roughly five months later.

There is one detail worth keeping in mind. AfterQuery said the Series A had actually closed “a few months” before the public announcement. So investors who paid the $300 million valuation were looking at an earlier and smaller version of the business than readers saw when the round became public.

Even with that qualification, the speed is unusual. AfterQuery was founded in February 2025 and entered Y Combinator without a finished product. Roughly 18 months later, YC partner Gustaf Alströmer described it as the fastest company in Y Combinator history to reach unicorn status.

Q2How much revenue is AfterQuery actually doing now?

AfterQuery is currently doing at least roughly $200 million in annualized recurring revenue if we take its founder’s latest public statement literally, which already brings the $3.2 billion valuation down to 16 times revenue or less.

The hard number we have is older. AfterQuery said in April that it had surpassed a $100 million revenue run rate. Founder Spencer Mateega then said in July that recurring revenue had reached the “hundreds of millions,” and Forbes repeated that figure when reporting the new valuation.

“Hundreds of millions” is annoyingly imprecise, but its minimum normal interpretation is around $200 million. We should therefore treat $200 million as a conservative floor rather than pretend we know whether AfterQuery currently sits at $230 million, $300 million or $400 million.

That uncertainty creates a huge difference in the valuation. At $200 million, investors are paying 16 times recurring revenue. At $300 million, they are paying 10.7 times. At $400 million, the figure drops to 8 times.

The $3.2 billion headline therefore sounds much wilder before we put revenue underneath it.

AfterQuery valuation at different revenue levels

Q3Is AfterQuery really growing fast enough to justify $3.2B?

AfterQuery is currently growing fast enough to make a double-digit revenue multiple believable, although the company’s own description of its contracts tells us not to extrapolate this pace too far.

Using the most conservative interpretation of the public numbers, AfterQuery went from more than $100 million annualized to at least $200 million in roughly three months. Going from exactly $100 million to $200 million over three months would require about 26% compounded monthly growth. Since both endpoints are approximate, we would treat that as an order-of-magnitude calculation rather than a precise growth rate.

What makes the number more believable is that several competitors are doing something similar. Mercor went from a $1 billion annualized gross revenue run rate in February to more than $2 billion by June. Micro1 moved from about $100 million to $500 million in gross annualized revenue over eight months. Sacra estimates that Handshake’s annualized gross AI-data revenue rose from roughly $245 million to $1.1 billion in a year.

Three separate companies expanding at that speed tells us something useful about AfterQuery. Its growth is exceptional, but the market around it is also expanding exceptionally fast. Frontier AI labs are pouring much larger budgets into expert-created training data, evaluations and reinforcement-learning environments.

There is still a catch. AfterQuery itself describes its revenue model in a finance job posting as “lumpy” and driven by large contracts. A few enormous lab agreements can add tens of millions of dollars to annualized revenue very quickly. They can also disappear much faster than thousands of small SaaS subscriptions.

The growth looks real. We would be much more cautious about assuming the same curve continues for another year.

Q4Is AfterQuery more expensive than Mercor, Scale AI and other AI-data startups?

AfterQuery does not currently look obviously overpriced next to the private AI-data companies investors are funding around it.

Mercor is the most useful benchmark because its numbers are recent. The company crossed $2 billion in annualized gross revenue in June and has discussed raising at a $20 billion valuation. That works out to about 10 times gross revenue.

Micro1 gives us another fresh comparison. Its gross annualized revenue climbed to roughly $500 million recently. Secondary-market transactions tracked by Legion have since implied a valuation around $3.6 billion, or just over 7 times gross revenue.

Scale AI was valued above $29 billion in Meta’s investment. Sacra estimates Scale generated around $2 billion of revenue in 2025, which would put that transaction around 14.5 times revenue. Surge AI generated approximately $1.2 billion in 2024 and Forbes currently estimates the company is worth around $24 billion, although the older revenue period makes its apparent 20 times multiple less useful than the Mercor comparison.

Handshake looks cheaper. Sacra estimates around $1.1 billion of annualized gross revenue against its last disclosed $3.5 billion valuation, although its revenue has grown so much since that financing that the comparison is stale.

We should also be careful with the word “revenue.” Mercor and Handshake report gross customer spend before contractor payouts, while AfterQuery has not publicly explained whether its reported recurring revenue follows the same convention. Until we know that, comparing these multiples to one decimal place would create false precision.

The broad result is clear: investors already value major AI-data suppliers anywhere from the high single digits to the teens on reported top-line revenue. AfterQuery sits inside that range.

AI-data company valuation comparison

Q5Does the stock market make AfterQuery look overvalued?

The public market says AfterQuery is priced much more like a fast-growing software company than a traditional human-data services company.

Look at Appen. It is a public AI-data company with roughly $250 million of trailing revenue and recently traded around 0.8 times sales. Its latest half-year results were actually improving, with revenue up 17%, yet the market still gives the company nothing close to an AI-startup multiple.

TaskUs provides a similar reality check. It generates around $1.2 billion of trailing revenue and currently trades at roughly 0.6 times forward sales. TaskUs is profitable, but revenue growth is modest and much of the company still looks like a people-heavy outsourced-services business.

Software investors will pay far more when margins, growth and product lock-in are stronger. ServiceNow recently traded around 9 times sales. Palantir has traded above 60 times sales while posting unusually high growth and margins.

AfterQuery currently lands somewhere around 8 to 16 times its reported recurring revenue depending on where revenue sits inside the “hundreds of millions” range. That is already above mature software companies such as ServiceNow at the upper end of the range, while remaining far below an extreme case such as Palantir.

We think that comparison frames the valuation well. AfterQuery cannot mature into something resembling Appen or TaskUs and still justify $3.2 billion easily. Investors are already betting that its software, research, proprietary datasets and technical integration with AI labs will make the economics much better than ordinary outsourcing.

Q6Is AfterQuery’s revenue really as valuable as SaaS revenue?

AfterQuery’s profitability is encouraging, but the quality of its reported revenue is currently the biggest missing piece in the entire $3.2 billion valuation.

Forbes reports that AfterQuery is profitable. The company’s own Strategic Finance job posting goes further, describing AfterQuery as a financially complex, profitable and cash-producing business. That is impressive for a company growing this quickly with only $30 million of previously announced institutional funding.

The same job posting also tells us why we should be careful. AfterQuery says it deals with millions of dollars in monthly API, compute and contractor costs. It talks openly about working-capital pressure from AI-lab customers and calls its own model a “lumpy, large-contract revenue model.”

Contractor economics are crucial in this industry. Mercor contractors receive roughly 60% to 70% of the amount billed to customers, according to reporting around its latest revenue numbers. Sacra estimates Handshake produces roughly $1.1 billion in gross annualized revenue but only around $450 million after contractor payments.

AfterQuery has never published the equivalent reconciliation. If it currently bills $300 million and keeps a large percentage after expert costs, 10.7 times revenue looks quite defensible for a profitable hypergrowth company. If most of that $300 million flows straight through to contractors, the multiple on the revenue AfterQuery economically retains could be two or three times higher.

Profitability reduces the concern, since we know there is already some money left after those expenses. We still need the gross margin before we can say whether AfterQuery deserves a software-like valuation with much confidence.

Q7Is AfterQuery still basically a data-labeling company?

AfterQuery now sells a much broader post-training stack than conventional data labeling, and lately it has been pushing further into owning the raw professional data itself.

Its current product catalogue includes supervised fine-tuning data, RLHF, verifier-based reinforcement learning, tool-calling environments, computer-use environments, code-generation data, custom evaluations and off-the-shelf datasets. AfterQuery also offers forward-deployed work where its teams help companies train and integrate agents around internal workflows.

That mix is strategically important because simple annotation is easy to commoditize. Building the environment in which an agent learns to operate Salesforce, a terminal, a browser or a professional workflow gives AfterQuery a deeper role in the training process.

A newer product called Atrium pushes the company one layer further upstream. AfterQuery is currently offering to buy private codebases, CAD designs and financial models directly from companies, with Jira, Slack, Confluence and Salesforce data listed as future categories. In practical terms, AfterQuery is trying to own unique professional data before competitors can turn the same information into training material.

Its off-the-shelf datasets follow the same logic. A custom project is sold once. A dataset that AfterQuery owns and can license repeatedly can produce much better economics.

The company is now evolving toward a mix of expert marketplace, proprietary-data supplier and post-training lab. The valuation becomes easier to defend as the last two parts grow faster than the first.

Q8What can AfterQuery do that competitors cannot easily copy?

AfterQuery has real technical differentiation today, although its moat is still young and competitors have enough capital to chase it.

The nearly 100,000 verified professionals on AfterQuery’s network sound impressive, but we would not treat that number alone as a moat. Handshake already has access to millions of students and graduates. Mercor and Micro1 recruit many of the same doctors, lawyers, engineers and scientists. Expert supply is competitive.

AfterQuery’s more interesting advantage is the software and research sitting around those experts. Spencer Mateega told Forbes that the company builds custom validation systems designed to keep training tasks inside a useful difficulty range. AfterQuery also runs its own post-training pipelines so it can test whether a dataset actually improves model performance before a customer has to do the work.

The company has published results showing what that looks like. One AfterQuery pipeline moved GPT-OSS-20B from a 3.1% pass rate to 17.0% on Terminal-Bench 2.0. Another experiment using AfterQuery’s Office Agent data improved a Nemotron model by a 20.9% net win-loss margin on GDPval, while a Qwen experiment reached 21.4%.

Nvidia provides outside validation. Its Nemotron 3 Ultra technical work used AfterQuery’s Office Agent Training Dataset, with AfterQuery describing an 11.4-point GDPval improvement in Nvidia’s warm-up ablation. Nvidia named AfterQuery as the data partner behind those tasks.

The company is also broadening its evaluation work. In a recently published project with legal AI company Legora, AfterQuery helped build a benchmark containing 5,161 legal cases across 28 practice areas and more than 11,000 source documents.

That is a more interesting moat than “we have lots of contractors.” AfterQuery is building knowledge about which tasks make models better, how to produce those tasks and how to measure the improvement afterwards. The advantage becomes durable if each project makes that system smarter.

Q9Are AfterQuery’s big AI-lab customers actually sticky?

AfterQuery appears to have extremely valuable customers, but we would assume those customers are relatively easy to lose until the company proves otherwise.

The entire AI-data industry gives us reasons to be cautious. After Meta invested heavily in Scale AI, several competing frontier labs reconsidered their relationships with Scale because they did not want a rival sitting close to a critical data supplier. Mercor also saw Meta pause work after a security breach. Huge contracts can move between vendors when strategic incentives change.

AfterQuery’s own financial language points in the same direction. The company describes large, lumpy contracts, meaningful working-capital requirements and payment tracking across frontier labs, big technology companies and newer AI labs. That sounds very different from collecting predictable $99 monthly subscriptions from thousands of independent customers.

We also know the universe of true frontier-model buyers is small. Companies such as OpenAI, Anthropic, Google, Meta, xAI, Nvidia and a handful of Chinese labs can each spend enormous amounts, but winning one or losing one can visibly change a supplier’s revenue.

AfterQuery does have evidence that the relationships go deeper than one-off labeling jobs. Its work spans training datasets, RL environments, evaluations and post-training research, which gives it more ways to stay involved as a model moves from one capability problem to another.

For now, though, we would put a lower valuation multiple on $100 million spread across three giant labs than on the same revenue spread across hundreds of enterprise customers. Customer concentration remains one of the clearest weaknesses in the valuation.

Q10Is the AI training-data market really big enough for AfterQuery to keep growing?

The AI training-data market is already big enough to support AfterQuery’s valuation, and current supplier revenues suggest the real market has grown faster than several published estimates expected.

The published market-size numbers are surprisingly messy. Grand View Research puts the global AI training-dataset market around $3.9 billion in 2026. Fortune Business Insights comes out around $4.4 billion. Mordor Intelligence, using a broader definition, is closer to $11.9 billion.

We trust the company revenues more than the tidy TAM charts. Mercor is already running above $2 billion gross annually. Handshake is around $1.1 billion gross. Micro1 recently reached $500 million. Surge generated about $1.2 billion in 2024. Scale generated around $2 billion in 2025 according to Sacra.

Those figures use different periods and accounting definitions, so adding them together does not produce a clean market-size estimate. They do show that billions of dollars are already changing hands among a small number of suppliers, before counting AfterQuery, Turing, Invisible and many smaller companies.

The work itself is also getting more complicated. AI labs now buy expert reasoning, software-use trajectories, coding environments, browser tasks, professional benchmarks, verifier data and reinforcement-learning environments. The latest AfterQuery products reflect that expansion.

We do expect the current growth rates to slow. Synthetic data will replace some human work, verifiers will automate evaluation in domains with objective answers, and labs will bring pieces of the pipeline in-house.

The likely outcome these days is a shift toward harder human work rather than the disappearance of human data altogether. A generic labeling task becomes automated; a difficult legal workflow, financial model or software-agent environment remains expensive. AfterQuery is already concentrating on that harder end of the market, which gives it a better chance of surviving the transition.

Q11Is AfterQuery’s China business becoming a serious risk?

AfterQuery’s China business is already financially meaningful, and it has become one of the clearest external risks to the company’s current growth story.

Forbes reported that at least $50 million of AfterQuery’s recurring revenue comes from Chinese AI labs. The same investigation estimated that China’s six largest AI labs collectively spend about $500 million a year with American AI-data suppliers.

That puts AfterQuery in an unusually strong position in the market. If current recurring revenue were $200 million, Chinese labs would account for at least one quarter. Even at $300 million, the reported $50 million would still equal roughly 17%.

The opportunity is obvious. AfterQuery can sell expertise and datasets to American and Chinese labs instead of depending entirely on the small U.S. buyer pool. Off-the-shelf datasets are especially attractive because the same underlying asset can be sold repeatedly.

Washington is now paying closer attention to that trade. U.S. lawmakers have questioned whether companies supplying data to Chinese AI developers should simultaneously work with U.S. government and defense customers. Advanced chips are already tightly controlled, while the professional data used to improve models currently crosses borders much more freely.

We would not price an immediate ban into AfterQuery’s valuation. There is currently no broad data-export regime equivalent to the chip controls. Still, $50 million is big enough that policy changes could knock a visible hole in revenue.

The China exposure currently increases AfterQuery’s growth while also making that growth slightly less dependable.

Q12How much revenue would AfterQuery need to make $3.2B look normal?

AfterQuery would need about $320 million of annualized revenue for its $3.2 billion valuation to equal 10 times sales, a level that could already be close to where the company sits today.

At 20 times revenue, the valuation requires only $160 million. The company appears to have passed that threshold already based on its “hundreds of millions” recurring-revenue disclosure. At 15 times, the requirement is roughly $213 million.

Ten times is more demanding because AfterQuery needs $320 million. We cannot confirm that it has crossed that level yet, although it falls comfortably inside the possible range implied by the company’s latest wording.

At a more conservative 5 times revenue, AfterQuery would need $640 million. That would require another substantial leg of growth, but it is hardly unimaginable when Mercor has already moved beyond $2 billion gross annualized revenue and Micro1 recently reached $500 million.

As seen above, the accounting definition remains crucial. These thresholds only tell us how much reported revenue is needed. If contractor payouts consume a large part of that top line, we would want AfterQuery to reach considerably more revenue before accepting the same multiple.

Revenue required at different valuation multiples

Q13What would make AfterQuery worth much more than $3.2B?

AfterQuery could grow well beyond a $3.2 billion valuation if its current hypergrowth continues while more of the business shifts toward reusable data, software and post-training infrastructure.

The revenue hurdle is surprisingly manageable. Once AfterQuery reaches $500 million of economically meaningful annual revenue, today’s valuation falls to 6.4 times sales. At $600 million, it is 5.3 times. Continued growth therefore fixes a large part of the valuation problem on its own.

The quality of that growth will decide how high investors can go. Atrium, off-the-shelf datasets and reusable RL environments could improve margins because AfterQuery can monetize the same underlying data more than once. Its research operation can deepen customer relationships by helping labs identify failures, create training data and measure the result inside one workflow.

We would also want to see technical validation keep coming from outside AfterQuery. The Nvidia work is a good example because a major model builder publicly documented using AfterQuery data. Legora’s new legal benchmark shows the company extending the same approach into a specific professional vertical.

A broader customer base would complete the bull case. If AfterQuery can keep several competing frontier labs, expand enterprise AI work and reduce its dependence on any one geography, the revenue becomes much easier to value like infrastructure.

Under that scenario, $3.2 billion could look modest quite quickly.

Q14What could make AfterQuery’s $3.2B valuation fall apart?

AfterQuery’s $3.2 billion valuation would become hard to defend very quickly if revenue growth slows before the company proves strong margins and durable customer relationships.

Gross-versus-net economics come first. A company billing hundreds of millions while passing most of the money to contractors deserves a very different multiple from a software company keeping 80% of each incremental dollar.

Customer concentration comes next. AfterQuery sells into a small group of labs capable of moving huge budgets between suppliers. A single customer loss could erase months of growth.

Competition is equally intense. Mercor, Handshake, Micro1, Surge AI, Scale AI, Turing and other providers are chasing similar budgets and expert pools. AfterQuery’s research and validation systems give it an interesting edge today, but none of those competitors lacks money or technical talent.

Automation could squeeze the market over time. Better synthetic data and automated verifiers will remove some work currently performed by humans. AfterQuery needs to keep moving toward professional tasks where human judgment remains difficult to synthesize.

Then there is China. The reported $50 million or more of recurring revenue from Chinese labs gives AfterQuery useful diversification today, but political restrictions could turn that exposure into a revenue shock.

We would become much more bearish if two of these problems appeared together. Slower growth alone is manageable. Lower margins alone are manageable. Slower growth combined with weak margins and contract losses would make a double-digit revenue multiple very difficult to sustain.

Q15So is AfterQuery really worth $3.2B today?

Yes, we think AfterQuery can plausibly be worth $3.2 billion today, although the valuation is already pricing in a better business than the financial disclosure has proved so far.

The easiest way to get misled is to stare at the jump from $300 million to $3.2 billion. The more useful comparison is with AfterQuery’s current scale and with what investors are paying for Mercor, Scale AI, Surge AI and Micro1. On that basis, the valuation sits inside the range already established for fast-growing AI-data companies.

The speed of growth carries the most weight for us. AfterQuery moved from a nine-figure run rate into the “hundreds of millions” unusually quickly, while competitors experienced their own massive revenue jumps. That tells us the company is participating in a real procurement boom rather than manufacturing growth in isolation.

The product has also become more interesting. AfterQuery now spans expert data, RL environments, evaluations, post-training research, reusable datasets and direct acquisition of private professional data. Its published model-training results and outside work with companies such as Nvidia and Legora give us evidence that the technical layer has value.

We are less convinced by the durability of the revenue. AfterQuery still has not disclosed gross margin, contractor pass-through, exact current revenue or customer concentration. Its own finance materials describe large, lumpy contracts and millions in monthly variable costs. Those details stop us from calling the company obviously undervalued.

As pointed out above, private peers already trade around similar top-line multiples, so $3.2 billion does not require AfterQuery to become an unprecedented business. It requires the company to keep its revenue somewhere in the current trajectory, preserve profitability and prove that a growing share of what customers buy comes from proprietary data and technical infrastructure rather than replaceable human labor.

For now, we would call the valuation aggressive but justified by the available evidence. The revenue quality still has to catch up with the revenue growth, but AfterQuery has already grown enough that $3.2 billion no longer looks like a valuation built mainly on hype.