How much did Braintrust raise in its Series B?

Braintrust raised $80M in its Series B round, announced on February 17, 2026. The round was led by ICONIQ, with General Partner Matt Jacobson joining the team. Prior investors Andreessen Horowitz, Greylock, Elad Gil, and basecase capital all returned for this round. [(Source)](https://www.braintrust.dev/blog/announcing-series-b)

What does Braintrust actually do?

Braintrust is an AI evaluation and observability platform. It helps engineering teams monitor and improve AI products running in production. The platform handles large, complex agent traces — sometimes hundreds of megabytes per interaction — and Braintrust built its own database, Brainstore, specifically to manage that scale.

Who are Braintrust's customers?

Braintrust's customers include Notion, Replit, Cloudflare, Ramp, and Dropbox. These are companies building AI into products that need to work reliably at scale. CEO Ankur Goyal says he talks to customers every day to understand their problems and track their progress.

Why did Ankur Goyal start Braintrust?

Goyal built internal eval tools at both Impira and Figma. Going through that difficult process twice made him think the problem was widespread. He started talking to peers to confirm it, then founded Braintrust to solve it. That focus on real user needs has stayed central to how the company builds its product.

Are vibe checks a valid alternative to formal evals?

Goyal says vibe checks are a form of eval — the version that doesn't scale. A developer judges output in their head, tweaks a prompt, and tries again. It works early on. But once a product hits production, with more users and more contributors to quality, informal checks break down. That is when structured evals and tooling become necessary.

Braintrust Raises $80M Series B to Lead AI Observability

By iCharles News · June 15, 2026

Braintrust closed an $80M Series B led by ICONIQ, with Andreessen Horowitz, Greylock, and others returning. The company builds AI eval and observability tools for teams at Notion, Replit, Cloudflare, Ramp, and Dropbox.

Braintrust Raises $80M Series B to Lead AI Observability

What Did Braintrust Just Raise?

Braintrust raised $80M in a Series B round. The announcement came on February 17, 2026. ICONIQ led the round. Prior investors Andreessen Horowitz, Greylock, Elad Gil, and basecase capital all returned, according to Braintrust's announcement.

What Is Braintrust?

Braintrust is an AI evaluation and observability platform. It helps engineering teams monitor and improve AI products in production.

CEO Ankur Goyal founded the company after building internal eval tools at both Impira and Figma. He saw the same painful process repeat itself twice and figured it was a widespread problem. He started talking to peers to confirm it — and it was.

The platform handles the scale of modern AI systems. Traces from today's agents can reach hundreds of megabytes per interaction. They span long multi-step tool calls and intermediate reasoning steps. Braintrust even built its own database — called Brainstore — to manage that complexity. Standard databases could not keep up.

Who Uses Braintrust?

We cover a lot of AI infrastructure news at iCharles, and Braintrust's customer list stands out. These are not early-stage pilots. Notion, Replit, Cloudflare, Ramp, and Dropbox are all using the platform in production.

Goyal says he talks to customers every day. After the round closed, Braintrust mailed notes to customers to thank them directly.

iPhone Fold May Ship in 2027, Not 2026

GitHub Copilot Switches to Usage-Based Billing June 1

China-Linked Hackers Hit US/Canada Research for 14 Months

xAI Trade Secret Suit Against OpenAI Dismissed

Round Details at a Glance

Item	Detail
Round size	$80M
Round type	Series B
Lead investor	ICONIQ (Matt Jacobson, General Partner)
Returning investors	Andreessen Horowitz, Greylock, Elad Gil, basecase capital
Announced	February 17, 2026
Notable customers	Notion, Replit, Cloudflare, Ramp, Dropbox

What Will Braintrust Do With the $80M?

Braintrust plans to use the funding for three things:

Grow its engineering and go-to-market teams
Expand into new offices
Build new products, to be revealed at its Trace user conference

New product details were set for the Trace conference, scheduled the week after the announcement. Goyal did not name specific office locations.

Why Did ICONIQ Lead This Round?

Matt Jacobson, General Partner at ICONIQ, pointed to customer focus as the deciding factor.

"At ICONIQ, we have seen that a defining trait among the generational companies is deep, authentic customer obsession. We believe Ankur and the Braintrust team embody this mindset and have been building their product from the start to serve the evolving needs of their customers." — Matt Jacobson, ICONIQ

Jacobson said the validation from leading AI teams using Braintrust confirmed the depth of that commitment.

Are Vibe Checks Enough — Or Do You Need Real Evals?

Goyal does not dismiss vibe checks. On the Product Growth podcast, he explained that a vibe check is actually a form of eval. It is the "do things that don't scale" version.

When a developer uses their AI product and judges the result in their head, that is a vibe check eval. They tweak the prompt, try a different model, and go again. It works at small scale.

The problem starts in production. More users arrive. More engineers and subject matter experts contribute to quality at once. The informal mental scoring breaks down fast. Teams need software, process, and tooling to keep pace. That is the gap Braintrust fills.

This challenge shows up across the AI stack. For example, GitHub Copilot usage-based billing is reshaping how developers think about AI cost at scale. And decisions about where AI runs — cloud versus local — are also getting more complex, as the Perplexity CEO hybrid local discussion shows.

Why Is Traditional Observability No Longer Enough?

Agents are no longer demos. They are embedded in engineering workflows. Their traces are long-running, multi-step, and massive. Traditional observability tools were not built for hundreds of megabytes per interaction.

Braintrust built Brainstore because existing tools could not handle the load. The problem is not just scale — it is also speed of change. AI systems shift constantly, faster than humans can inspect directly. Teams need clarity and confidence on every update.

Other companies are solving the same infrastructure gap in different ways. Salesforce acquired Fin to bring AI customer service infrastructure in-house rather than patch it together from outside tools.

What Comes Next?

The next confirmed milestone is the Trace user conference. Braintrust planned to announce new products there, the week after February 17, 2026. That is the next concrete event the company has committed to publicly.

Frequently asked questions

How much did Braintrust raise in its Series B?: Braintrust raised $80M in its Series B round, announced on February 17, 2026. The round was led by ICONIQ, with General Partner Matt Jacobson joining the team. Prior investors Andreessen Horowitz, Greylock, Elad Gil, and basecase capital all returned for this round. [(Source)](https://www.braintrust.dev/blog/announcing-series-b)
What does Braintrust actually do?: Braintrust is an AI evaluation and observability platform. It helps engineering teams monitor and improve AI products running in production. The platform handles large, complex agent traces — sometimes hundreds of megabytes per interaction — and Braintrust built its own database, Brainstore, specifically to manage that scale.
Who are Braintrust's customers?: Braintrust's customers include Notion, Replit, Cloudflare, Ramp, and Dropbox. These are companies building AI into products that need to work reliably at scale. CEO Ankur Goyal says he talks to customers every day to understand their problems and track their progress.
Why did Ankur Goyal start Braintrust?: Goyal built internal eval tools at both Impira and Figma. Going through that difficult process twice made him think the problem was widespread. He started talking to peers to confirm it, then founded Braintrust to solve it. That focus on real user needs has stayed central to how the company builds its product.
Are vibe checks a valid alternative to formal evals?: Goyal says vibe checks are a form of eval — the version that doesn't scale. A developer judges output in their head, tweaks a prompt, and tries again. It works early on. But once a product hits production, with more users and more contributors to quality, informal checks break down. That is when structured evals and tooling become necessary.

Sources

according to Braintrust's announcement braintrust.dev
he explained aakashg.com

Keep reading

Bezos Launches Prometheus AI at $41B Valuation

Nadella Warns Microsoft Staff: Stop Tokenmaxxing

Anthropic Fable 5 Ban: US Blocks Foreign Access

Salesforce Acquires Fin (Intercom) for $3.6B

0 Comments

Not a member yet? Join the community