Airtable Sold at 2.7x ARR. Here Is What Seat-Based Pricing Is Worth Now

Key takeaways

• Airtable's core platform sold for $1.285 billion on roughly $480 million of ARR, about 2.7x, despite growing more than 20% a year and serving 80% of the Fortune 100.
• The December 2021 peak valuation was $11.7 billion, so the sale landed at roughly one tenth of the peak.
• Companies that repriced around outcomes went the other direction: Intercom's Fin charges $0.99 per resolved conversation, HubSpot charges $0.50.
• Outcome pricing sets a hard revenue ceiling per unit of work. Your inference cost sets the floor. The distance between them is your gross margin.
• Before you copy anyone's per-resolution price, measure your own token cost per resolution.

What actually happened?

On August 4, 2026, Airtable agreed to sell its core platform to Bending Spoons for $1.285 billion in cash. Roughly $480 million in ARR. More than 20% annual growth. Cash in the bank. No forced process. The multiple works out to about 2.7 times revenue.

Monetizely's write-up of the deal frames it as the clearing price of what a Jefferies trader nicknamed the SaaSPocalypse, the market-wide repricing of seat-based software. The agent business, Hyperagent, was carved out before signing and stayed with founder Howie Liu. Bending Spoons bought the seats.

That carve-out is the whole story in one line. The part that bills humans went to a buyer who runs products for cash. The part that does the work stayed with the founder.

Why did a growing company sell at a harvest multiple?

Because the multiple is not about growth any more. It is about whether the pricing metric survives agents.

Per-seat licensing rests on one assumption: revenue grows with customer headcount. If an agent resolves the ticket, the customer needs fewer support seats. If it manages the pipeline, fewer CRM seats. The better your AI gets, the fewer units you bill. That is a business model eating itself, and buyers now price it accordingly.

The contrast in the same market is stark. Per Monetizely, AI-native platforms are fetching 25 to 30 times revenue, private SaaS deals close at 4 to 5 times, and Airtable printed 2.7. Same year, same economy, three different answers to one question: does your revenue survive automation?

What did the companies that repriced actually do?

They changed what the invoice is for.

Intercom rebuilt around Fin, an AI agent that resolves support conversations, and priced it at $0.99 per resolution. The customer pays when the issue is solved and does not pay when it is not. Seats became an access fee. Intercom knowingly gave up existing revenue to force the migration, and by May 2026 renamed itself Fin.

HubSpot ran the staged version: AI bundled into tiers in 2024, a credits currency in 2025, then outcome pricing in April 2026 at $0.50 per resolved conversation and $1 per qualified lead.

Note what HubSpot's number does. It undercuts Fin by half. A price war for AI-delivered outcomes started before most companies had picked a metric.

Here is the part nobody is modelling

This is where the pricing conversation usually stops, and it is exactly where it should start.

A per-outcome price is a ceiling. When you say $0.50 per resolved conversation, you have fixed your revenue per unit of work and handed your competitor a number to undercut. Your cost per unit of work, meanwhile, is not fixed at all. It is a token bill that moves with prompt length, retries, tool calls, context size, and whichever model you routed to that week.

So outcome pricing squeezes from both ends. The ceiling falls as competitors price against you. The floor moves every time your agent gets chattier. Gross margin per resolution is the number that decides whether the model works, and most teams have never calculated it once.

An illustrative example: say a resolution takes 25,000 input tokens and 1,500 output tokens after retries. At $3 per million input and $15 per million output, that is about $0.098, roughly 20% of a $0.50 price. Comfortable. Now say chain-of-thought and tool calls triple the token count, and you route to a premium model at 4x the rate. The same resolution eats most of the price. Nothing changed in your pricing page. Your margin quietly went to zero.

That is the calculation to run before you announce a per-outcome price, not after.

What should you do about your own pricing metric?

Three questions, in order.

First, what unit of work does your customer actually value? A resolved ticket, a qualified lead, a completed document. Name it before you price it.

Second, what does one of those units cost you to produce at your worst-case token path, not your happy path? Measure the p90, not the average.

Third, at what price does that unit still clear your target margin after a competitor cuts theirs in half? If there is no such price, the metric is wrong.

Seat pricing did not die because it was unfashionable. It died because the number it counted stopped growing. Any metric can meet the same end if you do not know what it costs you to deliver.

The takeaway: the market has now priced an unfinished pricing transformation at 2.7x ARR, and the companies that made the move kept their base and their multiple.

If you are sketching an outcome price, run your token cost per unit of work first with the Calcaas LLM cost calculator.