Selling Company Data to micro1: Tiers, Terms and Payment

micro1 pays companies to license their operating records, in three published tiers of $100K-500K, $500K-3M and $3M+, and takes operationally mature companies with 30+ employees that work and document in English, US companies first. Its payment step says "We pay in full within 30 days of acceptance", and its FAQ gives it 30 days to review and approve the data first, so the money arrives up to about 60 days from the transfer. In return it takes a licence that is worldwide, royalty-free, fully paid-up, non-exclusive, irrevocable and perpetual, with the right to sublicense through multiple tiers: in economic terms, close to a sale. A separate track takes companies that have shut down, another licenses the trajectories of a company's own AI agents for $1M to $2M a year, and in the Spirit Airlines bankruptcy micro1 offered $12,500,000, payable entirely in cash for one airline's records.

This guide is for the founder, CFO or counsel of a company that has heard micro1 buys company data, has been contacted by micro1 or someone in its referral programme, or is deciding whether to apply. It covers what micro1 is and why it buys, its published tiers and how they moved, who it takes (its size, language and place rules, and its track for closed companies), what records it wants, the licence it asks for and the clauses that matter after signing, when it pays, its process step by step, its agent-trajectory track, its Spirit Airlines bid, how it compares with Turing, Mercor, Handshake AI and Scale AI, and how to apply or be introduced. Every term is micro1's own published wording, cited to its record; where a line is our reading rather than micro1's words, it says so.

A note on who is writing. AIData.Broker is in micro1's referral programme, through a referral link held by AITraining.jobs, a job board with the same owner as this site: micro1 pays a referral payout for a company that signs up through that link once its data is transferred and approved, and the company pays nothing for it. We are not micro1's agent or representative, and micro1's own review decides whether it takes a company. A holder we introduce to a buyer pays us nothing: the buyer pays us its published referral payout, out of its own money; micro1's is $50,000 up to a $700,000 deal, or 7.5% above; $75,000 for a closed company up to $1 million, or 7.5% above. In our own sale the holder pays us a fee fixed in the mandate, only if a sale closes. Either way the holder itself licenses or sells its data to the buyer, and we do not buy a holder's data, take ownership of it, or license or sell it in our own name.

This guide is research from public sources, not legal, tax or investment advice.

Contents

1. What micro1 is and why it buys company data
2. The published tiers and how they moved
3. Who micro1 takes
4. What micro1 wants
5. The licence you grant
6. When micro1 pays
7. The process, step by step
8. The agent-trajectory track
9. Its Spirit Airlines bid
10. How micro1 compares with Turing, Mercor, Handshake AI and Scale AI
11. How to apply or be introduced
12. Questions people ask
13. What this means for you

What micro1 is and why it buys company data

micro1 provides expert human data, RL environments and evaluations to AI labs and frontier robotics labs. An RL environment is a simulated workplace where an AI model practises tasks and is scored on the result, and a company's real records (its tickets, its procedures, the threads where work was argued out) are what make such a workplace realistic. micro1 licenses a company's operating records and turns them into training environments it sells to AI labs. It is one of the fastest-growing companies in the market: it reported $7 million at the start of 2025 and says it surpassed a $1B annual run rate by October 2026, with about 180 people, and it priced its Series B at a $4B valuation.

The money for buying records is real but smaller and less certain than the headline. On 9 October 2026 micro1 announced a commitment to spend $1 billion over the next 12 months acquiring and licensing enterprise operational data, financed with capital from Citi and Hercules Capital whose size and form neither micro1 nor the press states. The one piece of that financing anyone else has filed is Hercules Capital's own quarterly report, which shows a senior secured term loan to micro1 of $20.75 million outstanding at 30 September 2026, up from $10.875 million at 30 June. In the Spirit Airlines bankruptcy micro1 said it had completed more than 50 data transactions in the preceding 45 days, without naming the companies, prices or customers.

Why it pays is visible in its margins. TechCrunch reported that data micro1 generates and sells to several customers carries gross margins as high as 80% to 90%, a figure from a person familiar with its finances, not from micro1. A licence that is perpetual and sublicensable is what lets a buyer sell the same material more than once. That is not a criticism; in our reading it is the reason micro1's first offer leaves room, and the reason the licence terms below matter as much as the tier.

The published tiers and how they moved

micro1 publishes the most structured price ladder of any buyer of company records, and it has moved twice in a month. On 9 October 2026 the tiers had no ceilings; by 26 October each had a range and the top one began higher; on 4 November the middle and top tiers rose again. Its payment step now says payments typically range from $100K to over $3M; on 26 October the same step said from $150K to over $2M.

Read the tiers as bands of negotiation, not prices. micro1 says compensation is based on "the quality, uniqueness, and value of the data", and its FAQ names the factors: dataset size and volume, workflow complexity, data quality, domain expertise and the uniqueness of the knowledge in it ( micro1's FAQ). Its company-referral page puts approved data packages at $100K-$3M+ on both its operating and its closed-company tracks. Earlier statements were lower: in July micro1 described partnering with 50 companies, paying $100K to $2M+ each (its own forum post), and a July headline in The Information, over a story about a firm micro1 approached, said data labeling startups were paying HVAC companies $150,000 for data (a reported figure).

Two things follow for a holder. The floor has held at $100K since October, so a qualified partnership starts there; and the ceilings have moved up each time, which in a market this young says more about micro1's appetite than about any one company's records. A higher tier is reached by breadth (several teams and functions) or by rarity, so the case to make is what your records show that no other company's do.

What its estimator shows

micro1's payout estimator asks where a company is headquartered, how many people it has and how many years of records, and returns a range. Read on 11 October 2026, it showed:

The estimator is a formula, not a reading of your data. It runs the same formula as Nyne's (and as Handshake's did until late October): a fixed $140,000, plus $2,333 per employee and $11,667 per year of records, with Canada and Europe weighted at 0.75 of the US and everywhere else at 0.3. Two estimators that agree are one anchor, not two valuations. Other buyers' estimators for the same company differ widely: for a US company of about 100 people and about 10 years, the low ends of the public estimators differ about 21 times. Our guide to how AI data buyers calculate offers takes each formula apart.

Who micro1 takes

micro1's FAQ asks for established companies with 30+ full-time employees, mature business operations and substantial historical operational data, primarily in English, and says it is "currently prioritizing US companies, followed by other Western markets". It also says the programme accepts referrals globally, with demand strongest for companies based in the US, UK and Canada, and that companies from other regions are considered case by case, depending on the needs of active opportunities. Its estimator offers four places: United States, Canada, Europe, Other. There is no fixed minimum of files or terabytes: it weighs historical depth, breadth, quality and relevance, and companies with several years of records across several systems are a stronger fit ( micro1's FAQ).

The language rule is strict. micro1 wants companies that operate in English and keep their documentation, workflows and knowledge bases in English, with representatives able to communicate with the micro1 team in English throughout. A company whose records are mostly in German or Spanish is outside what it says it looks for, and another buyer may value those records more: Turing, for one, takes companies from most regions, though it says English-language records sell best. On tax, its licence terms say each party bears its own taxes and any VAT or sales tax is stated separately on the holder's invoice; a company outside the US should read US withholding tax on data licence payments. Its referral programme excludes anyone who lives where micro1 is unable to operate for legal, regulatory or security reasons, without a country list.

The closed-company track

micro1 runs a second programme for companies that are no longer operating, the Defunct Company Data Partnership. Its company-referral page lists the requirements as 30+ employees, "10+ Countries" and English, without saying whether the countries are where the company operated or where the track is open. Its posting of 8 October 2026 asked for more: 50+ peak employees and substantial Microsoft Teams or Slack data, from a list of countries that ran from the United States, the UK and Canada to India, Australia, Singapore, Ireland, the Netherlands, Germany and France. Approved data packages on that track are priced at the same $100K-$3M+ as the operating track, and micro1 pays more for the introduction of a closed company than of an operating one, which shows how much it wants them.

A closed company has two problems an operating one does not. The first is time: once a company stops paying for Slack, Microsoft 365 or its CRM, the history goes, and buyers say records lose value fast after operations cease (Turing names 30 days for Slack and 60 to 90 days for a CRM). The second is authority: a dissolved company's assets may belong to its shareholders, a liquidator or a trustee, and someone has to be able to sign micro1's agreement and give its warranties. Read who owns a dissolved company's data, the SimpleClosure alternatives guide and the estates page before applying.

What micro1 wants

micro1 wants records that show how work is actually done. Its FAQ names "internal communications, SOPs, project histories, CRM records, support tickets, and other business documentation", and says the most valuable datasets capture real operational expertise, decision-making and repeatable workflows. Its data page lists examples by business function:

The communication layer carries much of the value. micro1's chief revenue officer has said that without Slack, CRM or email records the rest of a corpus is worth less, and other buyers say records that link across systems (a ticket tied to the commit that fixed it) are worth more than standalone documents. What micro1 does not need you to do is clean the data first: its FAQ says companies do not have to anonymise before transferring, because micro1 has its own processes for de-identifying and transforming sensitive information ( micro1's FAQ). The guides to Slack data, CRM data, support tickets and financial records cover each kind in depth.

What you promise about the data narrows what you can send. Unless the order form says otherwise, you warrant that the data holds no HIPAA health information, biometrics, card or bank account numbers or children's data, and, to your knowledge, no content generated or synthesised with AI or similar tools except what the order form discloses. A company whose staff draft with AI assistants should say so in the order form, since a material misrepresentation lets micro1 end the deal and take back what it paid (below).

The licence you grant

micro1 publishes its full Data License Master Terms, last updated on 30 October 2026, which is more than most buyers do. The grant is the broadest of the large programmes: "a worldwide, royalty-free, fully paid-up, nonexclusive, irrevocable, perpetual license, with the right to sublicense through multiple tiers". Perpetual means it never ends; irrevocable means you cannot end it, even for breach (your remedy becomes damages); sublicensable through multiple tiers means micro1 may grant the rights to others, who may grant them on again. Put those together for one payment and the economics are those of a sale, which is how to price it: a perpetual, sublicensable licence is a resale right by another name.

The 30 October version added that micro1 owns everything made by or for it from the data: improvements of the data, datasets, models with their weights, outputs, synthetic data, analyses and insights. You keep the data itself, which is licensed, not sold, and micro1's FAQ says companies retain ownership of their underlying data. Either side may end the agreement for any reason on 30 days' written notice, but section 1, which holds the licence, survives: ending the agreement stops future deliveries, not the licence on what you already delivered.

The clauses that cost holders after signing

Two clauses can take money back after the deal is done. If micro1 decides in good faith that the data manifest was materially inaccurate or misleading, or that you made any other material misrepresentation, nothing is payable and you refund what was paid, even after it accepted the data. The data manifest is your own written description of the data, by sources, types, volumes and specifications, which micro1 approves. The terms contain a narrower clause for quality: a claim for a latent defect or a mismatch with the manifest must come within 60 days of acceptance, and you choose whether to correct, replace or refund the proportionate amount. The manifest clause has no such time limit, so count your volumes net of duplicates and describe only what you can deliver.

The prior-licence warranty is the third clause to read before signing, and it bites before the deal rather than after. You promise you have not previously entered into any transaction granting rights to the same data on the same or substantially similar terms. A company that plans to license its records to several buyers on non-exclusive terms has to narrow this promise first, or sign micro1 last; once another buyer holds a similar licence, the promise is already broken. The guide to exclusive and non-exclusive licences and the licence terms page explain where to stand on each of these terms.

When micro1 pays

micro1 pays on acceptance, not on delivery. Its master terms give it 30 business days after delivery to reject the data in writing, with silence counting as acceptance, and you may invoice only after micro1 has accepted the data in writing. Its FAQ counts it slightly differently: micro1 has 30 days to review and approve the data, then asks for an invoice and pays within 30 days of it. Nothing is owed for any part it validly rejects within the window, and the master terms leave the payment period itself to each order form.

Counted from the day you hand the data over, that is up to about 60 days from the transfer, twice the 30 days its payment step names, and longer if the acceptance window runs in business days. That is not unreasonable (a buyer cannot judge records it has not seen), but it moves the risk of delay onto you. Fix three things in the order form: the acceptance window and written acceptance criteria tied to the manifest, the number of days from invoice to payment, and a right to cure before data counts as rejected. One cost is on micro1: export and delivery are borne by micro1, through a method you agree. micro1, like Google in the Spirit sale, has borne de-identification costs without deducting them from the price. For every buyer's payment trigger in one place, read when AI data buyers actually pay.

The process, step by step

micro1's data page sets out four stages: discovery and pricing, where it identifies licensable datasets and tailors an offer; contracting; delivery and acceptance, where you transfer through its APIs or another approved method and it reviews the data for completeness; and payment ( micro1's process). Its referral page adds the steps before those. Your company's contact sets up a company profile on micro1's company portal and completes the data manifest; micro1 then reviews the company and guides it through an initial AI interview to judge fit; qualified companies are invited to talk to a Data Partnership Manager about their data and licensing terms, then move through contracting, data transfer and approval ( micro1's referral FAQ).

micro1's referral FAQ says timelines vary with the company's eligibility, the complexity of its data, legal requirements and how quickly it completes each step, and that there is no guaranteed timeframe from application to partnership ( micro1's referral FAQ). Put an NDA in place before you describe anything sensitive: several buyers sign one before their intake questions, and at least one buyer's terms say that what you submit before a confidentiality agreement carries no confidentiality obligation beyond its privacy policy.

The agent-trajectory track

micro1's newest line pays for what a company's own AI agents produce. Its data page: "Every agent task produces a trajectory which is valuable training data for showing how real work gets done." A trajectory is the full record of one agent run: the instruction, each step the agent took, the tools it called, what came back and the result. micro1 says it licenses qualifying trajectories for $1M to $2M a year, with your approval. The track has three parts: optional agent build-outs, where micro1 builds and deploys agents inside your workflows, using your systems and permissions; continuous evaluation of your agents, to show where they fail and why; and the trajectory purchases, recurring from launch ( micro1's data page).

This is a different bargain from licensing an archive. It pays every year rather than once, and it pays for data your business produces as it runs, so it suits a company already using agents in real work (support, finance operations, sales operations). It also puts micro1's people and agents inside your systems, which is a security decision before it is a commercial one. Three questions to settle first: what makes a trajectory "qualifying" (the page does not say); what the trajectories contain, since an agent working a support queue touches customers' messages and accounts, which you may not be free to license; and how the master terms' promise that the data holds no AI-generated content except what the order form discloses is handled, since a trajectory is AI-generated by nature and, in our reading, has to be named in the order form. A related model exists for people rather than agents: The Information reported that micro1 approached a 14-person HVAC firm whose CEO now designs tasks from his own job and ranks how anonymised AI models handle them. The guide to RL environments explains what buyers build from such material, and the guide to AI-generated data how buyers treat it.

Its Spirit Airlines bid

The one public auction of a company's internal records shows how micro1 bids. Spirit Airlines, in chapter 11 bankruptcy, put its de-identified corporate records up for sale: an archive of about 100 million emails and 500,000,000 Teams items. At the auction on 14 August 2026, Google opened at $5 million in cash and three qualified bidders took the opening bid of $5.0M to a winning $10.0M in about two and a half hours, with Mercor the alternate bidder at $7,500,000. micro1 was not among the bidders.

On 3 September, twenty days after Google won, micro1 filed notice of a competing bid of $12,500,000 payable entirely in cash from cash on hand, which it described as a $2,500,000 or 25 percent premium over the successful bid, with no financing contingency. It is still the highest public offer for one company's archive, at $12.5M. It came on different terms: micro1 proposed to de-identify the records in house, with offers to exclude sensitive employment material, keep the data in the US and destroy raw employment records. The court's privacy ombudsman noted that "micro1 proposed to conduct the deidentification themselves and not to use an independent party", and asked for an opportunity to research the proposal further if the court considered it.

The higher price has not displaced the lower one. Google's agreement pays its $10M only on delivery of de-identified data certified by an independent agent, and the estate had already shown, in the auction itself, that it would choose a safer structure over a higher number: Mercor offered $5.2M with third-party de-identification or $7.0M with its own tools, and the debtors took the independently scrubbed bids. The hearing to approve the sale has been adjourned again and again over employee data, most recently to 11:00 a.m. on 17 November 2026, and the notice of 4 November stated no change to the sale at issue.

Three lessons for any company dealing with micro1. It pays for scale and moves fast, with cash. It prefers to de-identify in house, which is efficient for it and a term you can negotiate: who de-identifies, where and to what standard is worth money, as the Spirit bids show. And competition moved that archive's price to 2.5x the opening bid, which no single programme's first offer would have revealed. The deals page records every bid, and the guide to selling customer data in bankruptcy covers the court process.

How micro1 compares with Turing, Mercor, Handshake AI and Scale AI

micro1 is one of five large programmes a company of its size is likely to meet. Four pay a price; one, Turing's Project Lazarus, now pays a share of each sale of a record it licenses to labs. The differences that matter are less in the headline range than in the licence, the payment trigger and who removes personal data.

Against Mercor, micro1 publishes far more: its full master terms against Mercor's no published licence term, though Mercor pays sooner and on delivery rather than acceptance. Against Handshake AI, micro1's use is broader: Handshake binds itself to AI model training and evaluation only, while micro1's grant covers any use and sublicensing through tiers. Against Scale AI, the difference is where personal data is removed: Scale anonymises inside your own systems before anything is extracted, micro1 after the copy reaches it. Against Turing, the difference is the model itself: micro1 pays once, at a price you know before you sign; Turing pays only when labs buy, at a split you learn from its agreement, and for its first 90 days no one else may sell the record. Turing has a guide of its own, and the comparison of every programme covers a dozen smaller ones beside these five.

Track records are thin everywhere. micro1's court filing is the strongest public evidence of volume, and it is self-reported. Our board records that no holder has confirmed a Project Lazarus deal, none a sale of company records to Handshake and none one to Mercor. Ask any of them for a reference you can call before you send a sample.

Worked example: an 80-person accounting firm

Example. A US accounting and tax firm with 80 staff and twelve years of records in Microsoft 365, Teams, a practice-management system, QuickBooks Online and a CRM is contacted by someone in micro1's referral programme. On published rules it fits micro1 ( 30+ employees, English, US), Handshake AI ( 20+ full-time, 3+ years), Scale AI ( 40+ employees preferred), Turing (no size rule) and Mercor's listing for an accounting and tax firm, at $25k to $865k.

Its first question is what it may license at all. Its client tax files are not its to share: a tax preparer that discloses information furnished to it for preparing a return, or uses it for any other purpose, "shall be guilty of a misdemeanor", unless an exception or the client's specific consent applies. So client returns, and the email and Teams threads about them, come out before anything else is decided. What remains (its procedures, review checklists, internal training, engagement workflows and the de-identified history of how work moved through the firm) is what micro1's finance examples describe as financial processes, SOPs, reconciliations, approvals and reporting workflows, and it must also hold no bank account numbers or similar under micro1's warranty.

Its second question is the number. micro1's estimator, read in October, showed $285,896 to $604,463 for 50 employees and 10 years and $379,290 to $888,088 for 100 employees and 10 years; the firm sits between them, a range that straddles micro1's $100K-500K qualified tier and the one above it, which is for records spanning several teams and business functions. Those are indications from a formula, not offers. Its third question is order: if it signs micro1's master terms first, it warrants no earlier licence on similar terms, and a later licence to Mercor or Handshake AI would need that promise narrowed; if it signs elsewhere first, it must narrow micro1's warranty before signing micro1. It may still choose micro1, with the payment days, deletion days and refund limit written into the order form. It will know why.

How to apply or be introduced

There are two ways in, and micro1's own review decides in both. Apply directly: micro1's data partnerships page has a get-in-touch path, and its company portal registers the company and collects the data manifest ( micro1, data partnerships). Apply from someone who can sign for the company and speak for its data, in English.

Be introduced: get a free valuation of your data and press Introduce me on it. We introduce your company to every referral partner whose published rules take your data. micro1 counts a company only if it registers and completes its own application through the referral link, so for micro1 your page shows its sign-up through our referral link the moment you have pressed, and you complete micro1's application there yourself; the others are introduced by us. From there you deal with micro1 directly, under its terms, and you see the most any partner may pay us before we introduce you. The introduction does not change micro1's published terms, and the company pays us nothing for it. Where our own sale earns more for a holder based in the United States, the United Kingdom and Canada, the valuation says so, and every fitting partner is introduced as well.

Whichever way you go, prepare before the AI interview: a list of your systems with rough volumes and years, a decision about what stays out (client confidential material, HR, legal advice, anything under a confidentiality duty to someone else), and the name of the person who can sign. The readiness check, the datasheet tool and the guide to preparing data for sale cover each of them.

Questions people ask

Can I sell my company's data to micro1? Yes, if your company fits its rules: 30+ employees, English-language records, US companies first, then other Western markets. Strictly, you license it: micro1 takes a perpetual, irrevocable licence while you keep ownership of the underlying data.

How much does micro1 pay for company data? Its tiers are $100K-500K, $500K-3M and $3M+, raised twice since October. Its estimator gives a range from headcount, years and place, such as $285,896 to $604,463 for a US company of 50 people and 10 years, but the offer comes after it reads your data.

How long does micro1 take to pay? It pays within 30 days of acceptance, and it has 30 business days after delivery to accept or reject. From transfer to money is up to about 60 days; fix the days in the order form.

Does micro1 buy data from closed companies? Yes, through its Defunct Company Data Partnership, listed with 30+ employees, "10+ Countries" and English as requirements, at the same $100K-$3M+ for approved data packages. Someone must still have authority to sign for the closed company.

Is micro1 legitimate? It is a large, well-funded company, valued at $4B, and it publishes its full licence terms. Its claim of more than 50 data transactions in 45 days is its own, with no company or price named. Read the master terms with counsel, and ask for a reference.

Do I have to anonymise my data before sending it to micro1? No; its FAQ says micro1 de-identifies and transforms sensitive information itself, removing personal information "when relevant". You still warrant the data holds no health, biometric, card, bank account or children's data unless the order form says otherwise, so screen for those first.

Can micro1 resell my data? Not as received, as a stand-alone data product, and not before it has been anonymised. But it may sublicense it through multiple tiers, and it owns the datasets and synthetic data it makes from it.

What did micro1 bid for Spirit Airlines' data? It filed a competing bid of $12,500,000 in cash after Google won the auction at $10M, proposing to de-identify the records itself. The sale to Google is still awaiting approval, with the hearing set for 17 November 2026.

What this means for you

micro1 is a serious buyer with the clearest published terms of the large programmes, and those terms are written for micro1. Before you sign:

1. Price it as a sale. A perpetual, irrevocable, sublicensable licence for one payment is close to selling the records; compare the offer with what a sale is worth, not with a licence that pays again.
2. Write the days into the order form. The acceptance window, the days from invoice to payment, the deletion schedule and any exclusivity live in the order form, not the master terms.
3. Make the manifest exact. The refund right turns on it, with no time limit, so count net of duplicates and describe only what you can deliver.
4. Decide the order. The prior-licence warranty rewards licensing to micro1 first; if you want several buyers, narrow it before you sign anyone.
5. Put every offer side by side. The buyer match shows which programmes your company fits, and the comparison of every programme their terms together.
6. Know the number before the offer. A free valuation prices your records against public prices, and from it you can be introduced to micro1 and every other referral partner that takes your data, at no cost to you.

For everything recorded about the company, see micro1's page; for the market around it, the buyer board, what AI buyers check before buying data and how to sell company data to AI labs.

This guide is research from public sources, not legal, tax or investment advice.