Airtable Sold for 2.7x ARR. That Number Is the Whole Lesson.

Airtable is being acquired by Bending Spoons. Enterprise value of $1.285 billion, all cash, announced August 4. With Airtable's net cash on the balance sheet, the equity value lands around $2.25 billion. In 2021 the company raised at roughly $11.7 billion.

Airtable's ARR is about $480 million as of June 2026, and it is growing over 20% year over year. It serves more than 500,000 organizations, including about 80% of the Fortune 100.

Do the division. The operating business sold for roughly 2.7 times ARR.

That ratio is the entire story, and almost nobody is talking about it correctly. The headlines are running "$11B company sells for $1.3B" as if this were a failure post-mortem. It isn't. Half a billion in recurring revenue growing 20% is a real business, better than most of what got funded in the last five years. The company did not die.

The valuation thesis died. Those are very different things, and the difference is where all the learning is.

The short version

Airtable sold to Bending Spoons at a $1.285 billion enterprise value, roughly 2.7x its ~$480M ARR, down from a 2021 raise at about $11.7 billion. This is not an AI-disruption failure story. Airtable saw the shift early and moved hard: it refounded in 2025, hired an ex-OpenAI CTO, bought an agent team, and shipped a standalone product. The multiple still collapsed, because AI did not take Airtable's revenue, it took the option. The company's moat was effort-based switching cost, and effort-denominated lock-in deflates as agents make effort cheap. The enterprise plumbing is still sticky, which is why the ARR is real, but sticky is a cash-flow property, not an option on becoming the next platform. The lesson for anyone running product between $50M and $500M ARR: shipping AI features is the easy half, changing what your revenue is denominated in is the hard half, and the durable side of your roadmap is the part a model plus a motivated user cannot generate.

What people are getting wrong

The lazy read is that Airtable got disrupted by AI and lost. Sit with the timeline and that read falls apart, because Airtable saw the shift earlier than almost any incumbent SaaS company and moved harder on it than most.

Here is what actually happened, in order.

2021. Airtable raises at about $11.7 billion, part of more than $1.4 billion raised in total. Money was free and the pitch was enormous: the horizontal platform where anyone builds business applications without engineers. A next-generation Microsoft, Google, or Salesforce layer.

2022. The market turns. Airtable cuts roughly 254 people, about 20% of staff, and loses several senior executives.

2023. Another cut, reported at 27%. The company says out loud that it is moving upmarket to large enterprise and toward profitability. This is the correct move and it works, in the sense that the business becomes durable.

2025. Howie Liu refounds the company. In April an AI assistant that acts as a personal app builder. On June 24, a public relaunch of Airtable as "the AI-native app platform." In October, the acquisition of DeepSky, and David Azose, who led ChatGPT's business products at OpenAI, joins as CTO.

January 27, 2026. Superagent ships. First standalone product in thirteen years, priced from $20 to $200 per month, built on the DeepSky team, running frontier models from multiple labs. Liu describes it in interviews as something that could eventually eclipse Airtable itself. That is not a defensive crouch. That is a founder trying to kill his own franchise before someone else does.

August 4, 2026. Bending Spoons, fresh off a Nasdaq listing at an $18 billion valuation and coming off acquisitions of AOL and Eventbrite this year, agrees to buy the whole thing for $1.285 billion.

Read that sequence again. Airtable did the thing every board is telling every CEO to do right now. Refound the company. Hire the ex-OpenAI leader. Buy the agent team. Ship a new product line, not a chatbot bolted onto the old one. Cannibalize yourself on purpose.

And the market still priced it at 2.7x ARR.

That is the uncomfortable part, and it is why I think this deal is the most instructive event in enterprise software this year.

The multiple was never about the revenue

A revenue multiple is a bet on what happens next. When investors paid something like 40x forward revenue in 2021, they were not paying for spreadsheets with a database underneath. They were paying for the option that Airtable becomes the place where a generation of business software gets built. Every base a customer configured was supposed to be a brick in a wall nobody could climb.

The AI era did something specific to that bet. It did not take Airtable's revenue. It took the option.

Here is the mechanism, stated plainly. Airtable's value to a customer was proportional to the effort that customer had already put into it. Someone spent three weeks modeling their content pipeline into linked tables, views, automations, and interfaces. That three weeks was the switching cost. That three weeks was also, quietly, the entire moat.

Now a competent person with an agent describes the same content pipeline in a paragraph and gets a working application. Maybe not as polished. Maybe not as reliable at the edges. But the three weeks became an afternoon, and switching costs denominated in effort deflate at exactly the rate that effort gets cheaper.

I want to be careful here, because this is the part people overstate. Generated apps are not yet enterprise-grade for the workloads Airtable's Fortune 100 customers actually run. Permissions, audit, uptime, data governance, and the boring plumbing of integrations are still hard, and Airtable is still good at them. That is precisely why the company still has $480 million in ARR and 20% growth rather than a collapse. The installed base is sticky.

But sticky is a cash-flow property. Sticky is not an option on becoming the next platform layer. Buyers pay 2.7x for sticky. They paid 40x for the option, and the option is gone.

Why Bending Spoons is the right buyer, and what that tells you

Bending Spoons has a known playbook: acquire established software franchises with strong brands and real cash generation, run them with a fraction of the staff, optimize monetization, hold the cash flows. Evernote, WeTransfer, Vimeo, Meetup, komoot, the remains of Hopin, AOL, Eventbrite. They are excellent at it. They just went public at $18 billion on the strength of being excellent at it.

When a company like that is the highest bidder, the market is telling you something exact. It is saying the present value of this business is greater than the present value of its future story. That is the definition of a cash-flow asset rather than a growth asset.

If you run product at a company between $50 million and $500 million in ARR, this is the sentence to sit with. Not "will we get acquired." The question is: is the most rational owner of my business someone who will fund the next bet, or someone who will harvest the current one? That answer is being recalculated across the entire 2015 to 2021 SaaS cohort right now, and it is being recalculated against AI.

Eight things I am taking from this

This is the part I care about, because market commentary is cheap and I have to make decisions on my own product this week.

1. Ask what part of your product an AI cannot generate

Take your product surface and split it in two. On one side, everything a capable model plus a motivated user could reproduce in a weekend. On the other side, everything that requires something the model does not have.

The model does not have your customers' proprietary data. It does not have the accumulated decisions and context of their organization. It does not have their permission graph, their compliance posture, their audit trail, or their vendor relationships. It does not have their integrations that took eighteen months of partner negotiation. It does not have trust, and trust is earned in production over time, not generated.

Airtable's builder surface, the thing that made it famous, sat mostly on the first side of that line. The enterprise plumbing sat on the second. The market priced the second side and discounted the first, which is roughly the inverse of how the 2021 pitch was constructed.

Do this exercise on your own roadmap and be honest about the result. If most of your next two quarters is on the generatable side, you are building depreciating inventory.

2. Effort-based lock-in is a depreciating asset

I have watched a version of this before. At Adobe, a large part of the value of the creative tools was the years users had spent learning them. Learning curves as moats work beautifully until the cost of learning drops. At Salesforce, the moat was never really the UI, it was the data model, the org's configuration of it, and the ecosystem wired into it.

Ask what unit your lock-in is denominated in. If the answer is hours of user labor, assume that unit is deflating around 30% a year for the foreseeable future and reprice your strategy accordingly. If the answer is data gravity, regulatory approval, or the fact that you are the system of record that four other systems read from, you have something more durable. Taste, judgment, and accumulated context hold their value. Manual configuration effort does not.

3. Being early on AI is not the same as being valuable

This one stings, and it is the reason I wrote this post.

Airtable was early. Refounded in mid-2025. Ex-OpenAI CTO by the fall. Agent acquisition. Standalone product by January. If speed of AI adoption were the variable that determined outcomes, Airtable should have re-rated upward.

It did not, because the pivot was a product pivot on top of a business that still monetized like 2019. Seats. Per-user tiers. An installed base that bought a collaborative database and now gets pitched an agent platform. You can change what you ship far faster than you can change who your customers are, what they budget for, and what your revenue is denominated in.

The lesson is not "pivot slower." It is that shipping AI features is the easy half. The hard half is changing the shape of the business: what you charge for, who signs the check, and which budget line it comes out of. If you are adding agents while keeping per-seat pricing on a workspace product, you are doing the easy half and calling it a transformation.

4. Running two S-curves off one balance sheet is brutal

Superagent was, in my read, a defensible strategic call. Build the new thing outside the old thing so the old thing's constraints do not strangle it. Liu even said publicly it might eclipse the core product.

But it launched into a company of roughly 700 people, two large layoffs into its history, with a core business that needed continued enterprise growth to justify anything above cash-flow value. One balance sheet, one leadership team, one recruiting brand, two very different products with different buyers and price points, $20 a month on one side and enterprise contracts on the other.

If you are attempting this kind of dual transformation, separate three things explicitly or do not attempt it: funding, metrics, and the team's attention. Shared funding means every bad quarter in the core steals oxygen from the new bet. Shared metrics means you will judge a zero-to-one product by a scale-up product's dashboards and kill it in month seven. Shared attention means your best people spend Monday on the new thing and the rest of the week firefighting the old one.

5. Cash is a cushion, not a forcing function

Airtable had raised $1.4 billion and reportedly still held around half of it while generating cash. That is an enviable position and it bought years of runway to figure things out.

It also removed the deadline. There is a version of this company that, with less cash, would have been forced to make the hard monetization and focus calls in 2023 rather than 2025. I am not claiming that version wins. I am saying the balance sheet let the strategic question stay open long enough for the market to answer it instead.

Set your own deadlines. If you have runway past the point where the market will still fund your story, the runway is not protecting you, it is just letting the story go stale in private.

6. Growth rate is not durability, and buyers know the difference

Twenty percent growth on $480 million is good. It did not command a growth multiple. That gap tells you the buyer was looking past the headline growth number to the composition underneath.

The questions that determine which multiple you get are not "how fast did we grow." They are: how much of that growth is expansion inside existing accounts versus new logos, what is net revenue retention in the segment you actually want, how concentrated is it, how much of the base is on legacy pricing you cannot raise, and what does the cohort curve look like three years in.

If you cannot answer those five questions about your own product from memory, you do not know what your business is worth, and someone else will tell you.

7. Horizontal was a 2010s position, not a permanent one

Horizontal platforms won the last decade because being generic was expensive. Building a flexible, general-purpose surface that adapts to any workflow required real engineering, and once you had it, being general was an advantage.

Generic is now cheap. When a model can produce a passable general-purpose tool for any given workflow on demand, generality stops being a position and becomes table stakes. What becomes valuable instead is depth: the specific data, the specific compliance regime, the specific integration into how a specific industry actually operates.

I would rather own the workflow of record for a defined domain than the workspace for everyone.

8. Know which exit you are building toward, and say it out loud

There is nothing dishonorable about building a durable cash-flow business that Bending Spoons or someone like them eventually buys. It employs people, it serves 500,000 organizations, and it returns real money. It is a good outcome measured against most startup outcomes.

It is a terrible outcome measured against a $11.7 billion valuation and $1.4 billion of invested capital, because that capital was priced against a different promise. The pain in this story is not the business. It is the gap between the business and the promise, and that gap gets paid for by employees holding options struck at the promise.

Decide early which one you are building. Then price, hire, and raise consistently with that answer, because the mismatch is what hurts people.

Where my argument is weakest

I would not trust this piece if I did not steelman the other side, so here it is.

You can explain most of this deal without mentioning AI at all. The 2021 valuation was a zero-interest-rate artifact and was never real; the correction started in 2022, well before agents were commercially interesting. Airtable's shares were reportedly trading around $4 billion on secondary markets earlier this year, so most of the markdown had already happened before this deal set a price. And Bending Spoons pays cash-flow prices by design. That is their model. The clearing price says as much about who showed up to bid as it does about the future of no-code.

All fair. Here is why I still hold my read.

The 2022 and 2023 corrections were the multiple compression that hit every SaaS company. What is specific about 2026 is that Airtable did everything the AI playbook asks for, spent real money and real credibility on it, and the recovery did not come. A story can survive a rate cycle. What it cannot survive is a change in what the underlying product is worth, and the value of "a surface where humans manually configure business applications" is what changed.

If Airtable had refounded in 2023 instead of 2025, would this have ended differently? Honestly, probably somewhat. Two years is a lot. That is an argument for moving faster, not an argument that the pressure is not real.

What this changes for me, concretely

I build Heidi, an agentic productivity platform, so I have to apply this to myself or the post is just spectating.

The temptation in my category is to compete on builder surface. Nicer canvas, better agent composer, prettier workflow graph. That is the generatable half. Anyone with a good model and six months can ship a decent version of it, and the deflation curve on that surface is steep.

So the bets I am funding are on the other side of the line. Memory that accumulates a company's actual decisions and context over years, which cannot be generated because it is not in the model, it is in the organization. Trust infrastructure for long-running agents, meaning permissions, reversibility, guard rails around actions, and receipts for what an agent did and why. Being the system other AI tools read from, rather than another place people have to go.

None of that is as demoable as a slick builder. All of it is on the side of the line that held its value in Airtable's price.

I am also taking the pricing lesson seriously. If the work that used to take a person three weeks now takes an agent an afternoon, then charging per person for the seat where the work used to happen is charging for the wrong thing, and it puts you on the wrong side of your own product's success. That is the whole case for moving off per-seat pricing before the market makes you.

One thing to try this week

Open your roadmap. Take the next six things you plan to ship and sort them into two columns: things a capable user with a frontier model could approximate on their own in a weekend, and things they could not.

Count the engineering weeks in each column.

If the first column is winning, you have found your version of the Airtable problem, and you have found it while you still have time to do something about it. That is a better week than most.

Sources: Bending Spoons investor newsroom (August 4, 2026), Reuters, TechCrunch, Inc., The Information, Airtable newsroom.