I’m bummed by the Airtable exit.

I’ve been an indirect Airtable shareholder since my wife joined as engineer #12, just before the Series B in early 2018.

That’s when I fell in love with the product. It felt like the product had been built for me: I knew Excel well enough to run training sessions in my consulting days, and I understood object-oriented programming from C++ in high school. But in 2018, I was a COO at a fintech startup, and needed something to store and manage relational data: order books, CRMs, transaction history, and investment management. Airtable let me do that in a delightful, ergonomic, intuitive interface (unlike the vomit-worthy interface of MS Access).

For years, I loved building stuff with Airtable: MVPs, full-fledged internal applications to manage customer operations, and a portfolio management app used by several prolific angels. When I started getting into vibecoding in early 2025, I used Airtable to power the apps I built. I evangelized Airtable to dozens of customers and candidates.

For a while, the company was on a roll. They had incredible talent density, were a darling amongst VCs, and saw a meteoric valuation rise (>50x from 2018 to 2021). They looked poised to usher in the era of “personal software.”

When the market corrected, I still believed in their potential: I bought secondary shares at a ~$4B valuation in late 2023 and wrote a vigorous defense of their metrics.

This is why it pains me to see them stagnate, and exit for a dime on the dollar. With that heavy heart… here’s my breakdown of the exit math, where it seems the journey stumbled, what they’re attempting with Hyperagent, and the alternate path I would have been intrigued to see.

The stated acquisition price in the press release is $1.275B. This is the “ex-cash” enterprise value. The total exit price, including cash in the bank, is $2.25B. So the company did not exit for less than capital raised (as some have claimed); they raised $1.3B and created ~$1B of additional value.

The exit value (2.7x ARR) struck some as low, but the closest comparables — Monday.com and Asana — both trade at 2.2x sales with similar growth (10–25%). Note: this does not mean “software trades at 2-3x ARR”; it is critical to compare only after normalizing for gross margin and growth rate.

My homegrown metric — a two-year forward gross-profit multiple — would put this category around 3-4x, implying an ex-cash enterprise value of $1.3-1.5B.

All in all, this is a fair price for a company growing quite slowly, without a clear AI story to justify more aggressive forward assumptions.

Paul Bonnet of 20VC did a great breakdown that seems directionally right to me.

I have slightly different figures from my analysis so I’m sharing those below as well. As usual, data powered by Caplight.

This isn’t the outcome founders, investors, or employees hoped for. But given the two rounds of absolute destruction in enterprise software — one in late 2022, another more gradually over the last 12 months thanks to the onset of agentic software — I think it’s a draw snatched from the jaws of defeat.

The market correction was outside their control; how long they took to respond effectively to AI disruption wasn’t. For years, that looked like a massive missed opportunity.

The 2022–23 correction crushed valuations. Public multiples dropped 60-80%. Companies who had raised at peak valuations in 2020-21 suddenly needed to grow 3-4x just to earn back their last mark.

This sucked. But it was only the beginning. The far harder challenge was waiting around the corner, with the rapid onset of LLMs and what that meant for software.

Airtable entered this moment in a pretty solid position, relatively speaking:

But there were three challenges that I think set them back, perhaps irreversibly.

Airtable was fundamentally a product for no-coders: people with software-shaped problems who could think in systems, but could not — or did not want to — write code. It served as a bridge between spreadsheets and software. And when that gap was wide, Airtable thrived.

Agentic coding obliterated the gap within 2 years.

The Airtable product did this by emulating a “safe”, intuitive, ergonomic relational DB interface for a user — so that non-technical people didn’t have to think about painful things like cascading, row-level security, validation, typing, etc.

But when agents could interface with all that for you, that advantage, and the delicate abstractions that Airtable built (on a far more complex architecture) no longer mattered.

The same users — the technical PMs running product analytics, the sales ops and marketing ops people who were used to thinking in systems that optimized sales growth — could increasingly ask an agent to build directly against Postgres, handle validation and versioning, and generate a custom interface on top.

I saw this first-hand; as I graduated from vibe-coding to more legitimate agentic engineering, Airtable stopped scaling with me: my usage moved to Supabase, Neon, and plain vanilla Postgres. Airtable’s basic interfaces paled beside the frontends I could generate within hours using Codex or Claude Code.

Airtable’s early success came from a small, unusually empowered engineering team that spent years in a cave building an insane product.

As the company scaled, that original DNA appears to have become diluted inside a much larger organization. The company also endured a lot of turnover; key early people got frustrated and left, and big company transplants faced the typical organ rejection in a fast-growing startup.

I can’t know how decisions were made internally, but Airtable seems to have lost some of the concentration and clarity that produced the original product.

Even before AI, the product was too many things to too many people, and the fraction of users who could take advantage of Airtable’s insane breadth and depth was vanishingly small. When those power users left a customer, the solutions they had built often stagnated.

The expanding product surface (”Blocks” → “Extensions,” then “Interfaces,” then “Automations”) became too technical for non-technical users, yet not flexible enough for technical users. Agentic coding made the problem more acute: Airtable was now not flexible enough for non-technical users, either, because those users could increasingly generate “good enough” custom software themselves.

Airtable tried to break out of this trap. Airtable AI could autofill records, while Omni (the agentic interface on Airtable) promised to design your bases and interfaces. Neither worked well (I tried, for hours and hours). They did not hold a candle to what you could achieve with 10x less time, 10x more flexibility, and 10x more scalability using coding agents.

I wrote back in March about the looming chasm that legacy software companies needed to cross.

Airtable was one of the companies I named as trying to make the leap, but perhaps not being aggressive enough. They had not “burned the boats” like Notion or Intercom. Instead, like Figma and Box, their install base was massive but their architecture and product surface was fundamentally designed around a prior era. Despite valiant attempts, they still appeared to be sticking agentic interfaces on top of existing products rather than ruthlessly killing old products, charting a different vision, and driving relentlessly toward it.

And… I get it. To do that, you have to accept the possible destruction of hundreds of millions — or billions — in ARR. That would make you a pretty bad fiduciary. But the willingness to reinvent yourself is the only moat in an environment like today — where we all have a “this changes everything” moment roughly once a month.

Which brings us to this exit.

A few months ago, Airtable launched Hyperagent as “an Airtable product.” It was slick, seemed independent from the core product, and appeared to be gaining traction.

In today’s announcement, Bending Spoons agreed to buy the legacy Airtable business; Hyperagent had been transferred into a separate company before the deal. The Information reports that existing Airtable investors will also receive stakes in the new entity.

Until now, Airtable’s response to AI looked incremental: new agentic interfaces grafted onto a product designed for the prior era. Hyperagent is not that. The reorganization creates a hard separation between the profitable but stranded past, and the exciting agentic future.

Instead of burning the boats, Airtable sold the fleet and kept the speedboat.

Reading between the lines, it looks like Howie decided that a fresh product needs distance from the weight and inertia of a legacy product and large organization. Both could not thrive under the same roof. They handed the legacy business to a firm built to operate mature software, freeing Hyperagent from the architecture, organization, and fiduciary burden of protecting hundreds of millions in existing ARR.

It’s late in the game — but it is decisive and bold. It looks like the founders didn’t so much give up, as accept an unglamorous outcome for the legacy asset — so they can start again with a clean sheet. Existing investors get some liquidity from Airtable, while retaining a call option on the future of Hyperagent.

I think there was one other radical path under the same umbrella — one I’d have been very curious to see:

On top of that substrate, agents could have generated just-in-time interfaces and workflows from natural language. A user could ask for a custom operations application; Airtable would generate the frontend and underlying code, while every write still passed through governed data contracts, validation, approvals, audit trails, version history, and rollback.

Hyperagent could have been a terrific collaborator — an agent coworker for whatever use case — on top of that environment.

In effect, Airtable could have combined Replit’s generative experience with Retool-grade enterprise primitives (but, you know, nicer looking)… all built for non-technical operators. Software generation has commoditized; but a software factory with built-in governance could have perhaps enjoyed a compelling moat — allowing agents to generate and operate software without corrupting the underlying data or bypassing organizational controls. Airtable already owned much of the raw material.

Would Airtable have had the speed or organizational will to complete that reinvention? It’s risky; a pivot of that magnitude may have ended up strangling all of it — the original legacy product, Hyperagent, and the new product. All things considered, the sale + Hyperagent spinout is a clean break, and gives each the room to breathe.

Airtable made a series of opinionated decisions about how people could access and build on its architecture. They made perfect sense before LLMs; but in the years since, they had painted the product and strategy into a corner.

Airtable did not move fast enough to reinvent that business, leaving it in a tricky spot. Too big and valuable a business to scuttle. An architecture that was a liability and not an asset for the world we live in today. A team that needed to be big to support the existing product, which limits its ability to move fast. A product, brand, and organization with a lot of history (and hence baggage), which is a burden that grows heavier with each passing month. And founders who are tired of this burden after 12+ years in the seat... all along wondering about the opportunity cost — this is the best time in history to be building something greenfield.

Unlike the Airtable founders, I haven’t built a software unicorn. But I’ve navigated the utter chaos in SaaS for eight years: half a dozen pivots, as many fundraising rounds, three layoffs, raising too much and regretting it, and stepping away from companies and projects that once felt like my baby. This year, I’ve been wrestling with many of the same questions, albeit with less at stake — so I have deep empathy, albeit on a much smaller scale.

It’s very, very easy for an armchair observer to wax philosophical. But it’s worth respecting just how hard this journey is: building a company to half a billion dollars in ARR is no joke. At the end of the day, they built an incredible product and routinely delighted millions of users like me. They created a billion dollars in enterprise value beyond the capital raised, and found the business a logical landing spot.

At some point they have to set the weight down and move on. This transaction appears to be just that. And now they are untethered to do what every founder wishes they could, today: build for the future.

Let those who have done better cast the first stone.