Ask ten owners what their job history is worth and most will say nothing, because it has never been worth anything outside the shop. That changed when AI companies started paying businesses for the records they already keep. The question now is not whether service history has a price, but which price, from whom, and how far the advertised figure sits from the check.
What do buyers actually publish?
Four buyers publish terms an HVAC contractor can measure itself against. None of them prices by ticket, by job or by technician. They publish a range per company and a size floor.
| Buyer | Size floor | Published range |
|---|---|---|
| Nyne | At least 20 W-2 full-timers, U.S.; 3 years trading | $100K and up, typically |
| micro1 | 30+ full-time employees | $100K to $500K; $500K to $3M; $3M+ |
| Scale AI | Likes 40+; smaller firms considered | $10K to $1M+ (Scale’s word: illustrative) |
| Labelbox Prism | No size rule | Near $50K for small sets; past $1M for the largest |
Two of these, micro1 and Scale AI, also pay a referral fee to whoever introduces a company that signs. That is how introduction services like ours are paid. It does not change what the buyers publish.
Is there a figure for HVAC firms specifically?
One. In a 2026 story, The Information said data labeling startups were handing HVAC firms roughly $150,000 apiece, and it named micro1 as one of them. It is a reported number about real deals, which makes it more useful than a range on a sales page, but it is still one data point. It tells you the trade is being bought. It does not tell you what your shop will be offered.
Why do the ranges run so wide?
Because the buyers are describing very different deals. A $10K partnership at the bottom of Scale AI’s illustrative range and a $3M+ deal at the top of micro1’s are not the same product sold at different prices. The low end is a narrow slice of records delivered once. The high end is years of history, many record types, ongoing delivery, or data nobody else has.
For a service business, five things decide where you land.
How many years you hold
A model learns more from eight cooling seasons than from two. micro1 sets no minimum but favors longer histories, and Nyne wants three years of trading before it talks at all. Years on your current software matter most, since older systems are harder to export cleanly.
How well your technicians write
This is the one owners underrate. A work order that says “Replaced 45/5 MFD capacitor, verified amp draw, system cooling” teaches something. “Fixed, customer happy” teaches nothing. If your shop has enforced complete notes for years, that discipline has a value now that it did not have before.
Whether records link up
A call that links to a dispatch entry, a diagnosis, an estimate with three options, the option sold, an invoice and a callback 20 days later is a complete story. Buyers want stories. Loose tickets with no equipment record attached are worth less.
How much the buyer wants, and how often
Scale AI says its value scales with cadence, meaning how often you deliver. A one-time export of history is a different deal from a quarterly feed of new jobs.
Exclusivity
Prism’s starting point is an exclusive deal. Exclusivity tends to raise the price and shuts other buyers out of that data for the term. A non-exclusive one leaves you free to license the same set elsewhere, if your agreements allow it.
Why are actual licenses often lower?
A published range is a buyer telling you what its best-fitting deals look like. Scale AI is open that its range is only illustrative. Your deal will be priced on the records you can actually deliver once personal details are stripped, the years the buyer wants, and how much cleanup it takes. Plenty of companies that clear a size floor are offered the bottom of a range, or a narrower scope than they expected.
That is not a reason to stay out. It is a reason to know which buyers you qualify for before anyone names a number, and to have your own lawyer read whatever comes back.
Where should you start?
Count your full-time people, technicians and office together. Count the years on your current platform and the jobs that go through it each year. Then compare those numbers with the floors above. If you want the comparison done for you, our tool shows what your records are worth on each buyer’s published terms, and the homepage explains how to sell HVAC data without handing over anything you should keep.
The number worth knowing first is not the price. It is which buyers will take the call.