Miju Labs

← All companies

micro1

5x'd in eight months to a $500M gross run rate, and is the only vendor claiming 80–90% gross margin on anything — by reselling the same dataset more than once.

medium confidence4 minupdated 2026-08-29ai labs · data · margin
Vertical
Expert data for frontier labs
Founded
2021 (some sources 2022)
Headquarters
Palo Alto
Raised
~$38–42M (pre-seed $3.3M Aug 2023; $35M Series A Sept 2025)
Last valuation
$500M (September 2025); a higher round was reportedly being raised in August 2026, terms not established
Revenue
$500M GROSS run rate (August 2026), up from $100M ARR in December 2025. Net reported at $150–200M.
Status
Active; fastest-growing of the small caps
Who runs it · 4 people in the index
Ali AnsariFounder & CEO→JJerome JoszChief Financial Officer (also VP Government per The Org)→WWilliam AlmeidaChief Revenue Officer→Andrew MaasVP of AI (AI Engineering)→

Latest (Sep 2026): Grew from ~$7M revenue at the start of 2025 to $100M ARR (Dec 2025), ~$300M (Apr 2026) and a $500M gross run rate by August 2026, with net put at $150-200M (TechCrunch, 20 Aug 2026). It was raising a new round above its Sept 2025 $500M Series A valuation in August 2026; Inc. (11 Aug 2026) cites a valuation 'over $2.5B', unconfirmed elsewhere. Ansari publicly says micro1 does not sell data to Chinese model makers, and claims 80-90% gross margins on resold off-the-shelf/synthetic datasets; it is also building robotics pre-training data.

What micro1 is saying
micro1 reposted this
Ali Ansari
ceo at micro1
•
the hardware embodiment of frontier models like Claude and GPT is the most urgent AI safety problem in front of us today. we simulated two very simple use cases using claude both in simulation and using robot arms. in one, claude spilled toxic liquids in a lab. in another, the force it used to place an animal toy into a basket was strong enough that it could have physically harmed sensitive material—or anything else in its path. these are simple experiments, using models out of the box today. as researchers increasingly give frontier models arms, legs, and access to the physical world, we need to urgently build and assess guardrails around what these systems can and cannot do. models escaping sandboxes or compromising enterprise security infrastructure are serious concerns. however, hardware embodiments introduce something fundamentally different: an AI system can make a mistake in the physical world, and the consequences may not be reversible. this is not a future safety problem. the capabilities exist today. we’ve released a report detailing this & solutions we propose: lnkd.in/gTpyk7zW
56051 comments • 136 reposts
Ali Ansari@aliansarinik·
the hardware embodiment of frontier models like Claude and GPT is the most urgent AI safety problem in front of us today. we simulated two very simple use cases using claude both in simulation and using robot arms. in one, claude spilled toxic liquids in a lab. in another, the force it used to place an animal toy into a basket was strong enough that it could have physically harmed sensitive material—or anything else in its path. these are simple experiments, using models out of the box today. as researchers increasingly give frontier models arms, legs, and access to the physical world, we need to urgently build and assess guardrails around what these systems can and cannot do. models escaping sandboxes or compromising enterprise security infrastructure are serious concerns. however, hardware embodiments introduce something fundamentally different: an AI system can make a mistake in the physical world, and the consequences may not be reversible. this is not a future safety problem. the capabilities exist today. we’ve released a report detailing this & solutions we propose. link in comments below.
417620189KView on X ↗
micro1@micro1_ai·
Meet Ashley Ramsay, customer service operations expert at micro1. Her husband has served in the Air Force for 13 years, with four relocations along the way. Her background is in hospitality, but she’s found her calling training AI. Ashley is part of a growing number of military spouses and veterans finding real, paid opportunities training AI. micro1 is proud to support military families with flexible opportunities to apply their expertise and help shape the future of AI. Link to watch the full interview in the comments.
1013652.9KView on X ↗
All posts in the feed →

micro1 is the small-cap that grew fastest in this vertical and the only one that has articulated a credible route out of the labour-arbitrage margin trap.

CEO Ali Ansari raised a $3.3M pre-seed in August 2023 and a $35M Series A at $500M led by 01 Advisors in September 2025, with Adam Bain and Joshua Browder joining the board (Sacra; TechCrunch). Revenue went from $100M ARR in December 2025 (TechCrunch) to a $500M gross run rate by August 2026 — 5x in eight months (Dataconomy; Sacra).

The 5x compares two figures on possibly different bases

The December 2025 figure is reported as "$100M ARR" with no basis stated; the August 2026 figure is explicitly a $500M gross run rate. If the ARR figure was net, the growth is not 5x. No source resolves it, and the atlas carries the 5x because both sources do — not because the bases have been checked. See GMV is not revenue.

The number that matters, and the number that does not add up

micro1 claims 80–90% gross margin on "off-the-shelf" datasets resold to multiple clients (Dataconomy).

That is the single most commercially interesting claim in the vertical. Every other business here sells an hour once at a 27–40% margin. A dataset built once and licensed repeatedly has software economics, and it is the same structural escape that Mercor is buying its way toward with RL environments and that Mechanize is selling directly. What is missing is the split: nobody has disclosed what share of micro1's $500M comes from off-the-shelf datasets versus bespoke expert hours, and the 80–90% figure means very little without it.

Gap in the record

The net-revenue reporting is internally inconsistent in its own source. The article states that micro1 retains 60–70% of gross — which on $500M would be $300–350M — and then gives net run rate as $150–200M. Those cannot both be true. The $150–200M net on $500M gross implies a 30–40% take, which matches every other take rate in the sector, and that is the reading to trust. The 60–70% retention claim should be discarded. Independently tagged [WEAK] on this point in the research notes and flagged as arithmetically inconsistent in a second compilation.

The reason this matters beyond micro1: a "$500M run rate" headline is 2.5–3.3x the number that actually accrues to the company. See GMV is not revenue and What a rake can actually be.

Supply and demand

micro1 accepts roughly the top 1% of applicants — PhDs and senior engineers (Sacra). Named customers are OpenAI and Anthropic, with a push into the Fortune 1000. That push is the strategically interesting part: enterprise buyers are more fragmented than labs, which is the only structural cure for the One customer is a binary event problem that defines this vertical. It is also, on current evidence, a small share of the book.

Ansari states micro1 does not sell to Chinese model makers (Dataconomy) — a deliberate contrast with the reporting around Surge AI and Mercor, and a hedge against an export-control regime for training data that does not exist yet but has been publicly argued for by Alexandr Wang.

Valuation

The $500M mark from September 2025 is now stale to the point of being misleading: at a $500M gross run rate it implies roughly 1x gross, or 2.5–3.3x net — by a wide margin the cheapest headline in the private cohort, and only because nobody has re-priced it. A further round at a materially higher valuation was reported to be in progress in August 2026.

Gap in the record

Terms of the 2026 round could not be established — no size, lead or post-money is on the record.

The read

micro1 is the clearest test in the atlas of whether the dataset-resale model is real. If the 80–90% margin holds across a growing share of revenue, this is the one company in the vertical that could eventually justify a software multiple. If it stays a rounding error against bespoke expert hours, micro1 is a smaller Mercor at the same 30–40% take, and What the public market pays for labour applies to it identically.

Linked from
specialist

Contra Labs

Not a startup — a business line of a six-year-old freelance marketplace, launched five months after a $740K cheque. The template everyone wants to copy, and the parts of it that do not survive inspection.

vertical

Expert data for frontier labs

Labs buy throughput of credentialled human labour — annotation, preference data, reasoning traces, RL environments — and pay nine figures for it at a staffing margin.

page

What the model actually is

Buyers with money but not capability, sellers who are fragmented, and a middleman who organises them and keeps a big cut. Four conditions have to hold at once — and calling the result a marketplace is how it gets mis-priced by a factor of ten.

vertical

Contingency recruiting marketplaces

Paraform has raised $65M and paid $50M to recruiters. Back the split out and the whole lifetime business is roughly $71M of placements and $21M of net revenue — against a perm-placement pool an order of magnitude smaller than the TAM slide says.

topic

GMV is not revenue

The same company is worth 13x or 38x depending on which number you divide by. Nearly every headline in this sector quotes the one that flatters — here is how to tell in ten seconds.

playbook

Pricing and the contract

The buyer's alternative is nearly always to hire someone, and increasingly an offshore someone at half the cost. Price against that — then decide gross or net deliberately, because the same facts that book you gross make you an employer.

taste

Sizing the taste market

$50–300M of externally-purchased spend, central $100–150M, built from a $6–10B expert-data market with a $3.8B verified floor times a 1–3% creative share. The uncomfortable half: published practice pays $0–16/hour for aesthetic judgement, and a field-defining benchmark cost $13,433.55.

page

The capital register

Every company the sweep found, with what it raised, what it was marked at, what it earns and whether that revenue is gross or net. Plus the exits, the failures and the absence of a public bear case.

page

Evidence register

Not a bibliography — a graded list of the claims the atlas leans on, what each one holds up, where it came from and how much weight it will take.

taste

The generalists in creative

Mercor already sells the exact product — narrated senior design reasoning — at $150–250/hr, from Pentagram and Wolff Olins pedigree, to an unnamed frontier client. Contra's 'up to $100/hr' is the middle of the band, not the top of it.

page

What we could not establish

The questions that would most change a decision, the thirty-three contradictions the research left unresolved, and the structural holes where no evidence exists anywhere — not just where we failed to find it.

vertical

Expert networks

$200 to the expert, $800–900 to the client, held for four decades. The oldest version of this model has the most durable rake in the atlas — and may not sell to AI labs at all.

company

Mechanize

$9.1M of seed capital, ~35 people, no disclosed revenue, and reported Google talks at $1.5B+ — the licence-and-hire template for selling environments instead of hours.

company

Prolific

The only company in the vertical that publishes its take rate — 42.8% on top of participant pay — and the best public anchor for what this spread actually is.

company

Paraform

The archetype of the recruiter-side marketplace: pays independent recruiters ~70% of a placement fee and keeps the rest. $65M raised, $50M paid out, and no independent evidence of any of it.

company

GLG

The largest expert network, a 70–80% take held for decades, and a market share that halved anyway. Its 2021 S-1 is the highest-value document in the atlas — quoted in one research file and recorded as unopened in another.