Let's get the number on the table first: $40 billion. That's the valuation Cognition, the maker of the AI coding agent Devin, is now discussing with investors for its next funding round, according to reporting from Bloomberg picked up on August 12 (TechCrunch).

Here's the part that makes people blink. In May 2026, Cognition raised roughly $1 billion at a $26 billion valuation (PYMNTS). So the company is looking to add about $14 billion in paper value in under 90 days — a jump of roughly 54% between one round and the next, with the ink barely dry.

What's driving it isn't a demo or a keynote. It's a revenue line that has, by the company's own account, roughly doubled since the spring.

The thesis: Cognition is being repriced not on hype but on how fast enterprises are actually paying for AI that writes code — and that speed is now setting the pace for the whole sector.

🧠 Why This Matters

Valuations in private AI have gotten numb-inducingly large, so it's worth being precise about what changed here. When Cognition closed its May round, its annualized revenue run rate was about $492 million (PYMNTS). Reporting tied to the new talks puts that figure nearing $1 billion — roughly double, in about three months (TechCrunch).

That matters because "AI coding assistant" used to mean autocomplete you could turn off. Cognition sells something it pitches as an agent that does engineering work — legacy-code modernization, platform migrations, the grinding maintenance that companies have deferred for years. And the customers writing the checks are not startups experimenting: they include Goldman Sachs, Citi, Mercedes-Benz, and NASA, plus U.S. Army and Navy work (PYMNTS). When names like that show up on an ARR chart doubling every quarter, investors stop arguing about the multiple.

📊 Deep Dive

Cognition's rise reads less like a slow climb and more like a staircase taken three steps at a time. The company was founded in 2023 and launched Devin as, in its telling, "the first AI software engineer." Then it started buying and compounding.

The trajectory in plain figures:

  • Devin ARR, September 2024: about $1 million (Cognition)
  • Devin ARR, June 2025: about $73 million (Cognition)
  • Valuation, September 2025: $10.2 billion, after a $400M-plus round led by Founders Fund, weeks after buying rival Windsurf (CNBC)
  • Valuation, May 2026: $26 billion on roughly $1B raised, ARR about $492 million (PYMNTS)
  • Valuation, August 2026 (in talks): $40 billion, ARR nearing $1 billion (TechCrunch)

The Windsurf deal is the hinge. Cognition acquired the coding-tool company in mid-2025, and CEO Scott Wu has been blunt about the effect on the top line.

"Our acquisition of Windsurf more than doubled our ARR." — Scott Wu, CEO, Cognition (Cognition)

Wu frames the product not as a headcount replacement but as a shift in what engineers spend their day on.

"We envision a world of software abundance where engineers become architects." — Scott Wu, CEO, Cognition (Cognition)

It's a tidy pitch, and the enterprise adoption gives it teeth. But a $40 billion price on a company nearing $1 billion in revenue still implies a multiple of roughly 40 times run-rate sales — the kind of number that only works if the growth keeps compounding.

⚠️ The Catch

Start with what's confirmed and what isn't. The $40 billion figure comes from reporting on early talks, not a closed round — terms can move, and the size of the new raise (reported as more than $1 billion) is still soft (Bloomberg). "In discussions at" is not "raised at."

Then there's the crowd. AI coding is now one of the most contested corners of software, and Cognition is being repriced in the same weeks rivals are getting swept up by giants. The competitive heat is real, the switching costs for enterprise customers are lower than anyone likes to admit, and a 40x revenue multiple leaves no room for a growth stumble. Doubling ARR from $492 million was hard; doubling again from $1 billion is harder, and every megacap with a cloud is now selling a coding agent of its own.

🎯 What Happens Next

Watch for three things. First, whether the round actually closes at $40 billion or lands lower — the gap between the talked-about number and the signed one will tell you how much conviction investors have at this multiple. Second, whether the ARR that "nears $1 billion" crosses it and keeps climbing, because at this valuation the growth rate is the story. Third, how much of that revenue is durable enterprise seat expansion versus one-time migration projects that don't repeat.

If Cognition closes near $40 billion, expect the print to reset the ask for every other AI coding startup on the fundraising trail this fall.

🧩 Bigger Picture

Zoom out and Cognition is a clean read on where AI money is actually going. The frothy 2023–24 story was foundation models; the 2026 story is applications that turn those models into revenue — and coding is the first place where "the AI did real work and a Fortune 500 paid for it" is measurably true at scale. Wu himself argues the ground has shifted, saying "agents are already doing real work alongside individual developers" (Cognition).

That's why a repricing this steep isn't dismissed out of hand. Investors aren't betting on a chatbot that might work someday; they're extrapolating a revenue curve that has, so far, kept bending upward. The risk is the same as the reward: the entire $40 billion case rests on that curve not breaking.

For now, Cognition has done the one thing that makes a valuation stick — it grew into the last one before anyone finished arguing about it. The question isn't whether $40 billion is a lot. It's whether, in another 90 days, it'll look cheap or reckless.

Sources