Windsurf: AI Coding Startup’s Rapid Rise
Key Points
- Founded in 2021 as Exofunction by MIT grads Varun Mohan and Douglas Chen, the company initially built GPU‑optimization tools before pivoting to AI coding assistance with Kodium.
- After raising a $243 M Series C at a >$1.2 B valuation in 2024 (backers including Founders Fund and Kleiner Perkins), Kodium rebranded to Windsurf and launched an AI‑native development environment with the Cascade agent to compete with tools like Cursor, pricing its premium tier at $15 per seat versus Cursor’s $20.
- In March 2025 Windsurf earned FedRAMP High certification, becoming the first AI coding tool cleared for U.S. government workloads and giving it a strategic edge in regulated industries.
- By April 2025 the company reported roughly $100 M in ARR (up from $40 M months earlier), claimed over a million developers on the platform, and was notably stronger with enterprise decision‑makers than consumer engineers, positioning it as the #2 player behind Cursor.
- The rapid rise of AI startups was underscored when Bloomberg broke news of OpenAI’s $3 B all‑cash acquisition in April 2025, illustrating the multi‑billion‑dollar M&A landscape that Windsurf now operates within.
Sections
- Windsurf: AI Dev Tool Evolution - The transcript outlines the rapid rise of the startup Windsurf—originally founded as Exofunction and later rebranded from Kodium—a well‑funded AI coding assistance platform that pivoted to an AI‑native development environment, secured a $1.3 billion valuation, undercut competitors on price, and achieved FedRAMP High certification in early 2025.
- OpenAI Secures Windsurf Acquisition - In early May 2025, after a letter of intent and exclusive negotiations, OpenAI finalized a $3 billion deal to fold Windsurf into its ChatGPT developer suite, coinciding with Anthropic’s withdrawal of Claude and mixed reactions from the developer community.
- OpenAI Deal Collapse Triggers Exodus - The transcript outlines how the expiration of OpenAI's exclusivity agreement, compounded by Anthropic's strategic actions, forced users toward competitors and precipitated the collapse of a major enterprise deal.
- Google Deal Sparks Silicon Valley Trust Crisis - The speaker argues that Google’s acquisition, driven by AI salary pressures, rewards founders for abandoning their startups and undermines long‑term trust in the Silicon Valley ecosystem.
- Cognition Labs Steps In - The speaker explains how Cognition Labs is rescuing a Google‑sidelined developer platform (Devon/Windsurf), leveraging its FedRAMP certification for government contracts, reviving ties with Anthropic, and illustrating the extreme talent‑money dynamics driving Silicon Valley's talent wars.
- Model vs. Tool Makers' Power Struggle - The speaker outlines how model creators control roughly two‑thirds of the AI ecosystem’s power while tool makers retain about one‑third, using Windsurf’s acquisition and FedRAMP certification as examples, and predicts that code‑based agents will expand as a clear, rewardable use case.
Full Transcript
# Windsurf: AI Coding Startup’s Rapid Rise **Source:** [https://www.youtube.com/watch?v=qgQvci8NzM8](https://www.youtube.com/watch?v=qgQvci8NzM8) **Duration:** 00:21:24 ## Summary - Founded in 2021 as Exofunction by MIT grads Varun Mohan and Douglas Chen, the company initially built GPU‑optimization tools before pivoting to AI coding assistance with Kodium. - After raising a $243 M Series C at a >$1.2 B valuation in 2024 (backers including Founders Fund and Kleiner Perkins), Kodium rebranded to Windsurf and launched an AI‑native development environment with the Cascade agent to compete with tools like Cursor, pricing its premium tier at $15 per seat versus Cursor’s $20. - In March 2025 Windsurf earned FedRAMP High certification, becoming the first AI coding tool cleared for U.S. government workloads and giving it a strategic edge in regulated industries. - By April 2025 the company reported roughly $100 M in ARR (up from $40 M months earlier), claimed over a million developers on the platform, and was notably stronger with enterprise decision‑makers than consumer engineers, positioning it as the #2 player behind Cursor. - The rapid rise of AI startups was underscored when Bloomberg broke news of OpenAI’s $3 B all‑cash acquisition in April 2025, illustrating the multi‑billion‑dollar M&A landscape that Windsurf now operates within. ## Sections - [00:00:00](https://www.youtube.com/watch?v=qgQvci8NzM8&t=0s) **Windsurf: AI Dev Tool Evolution** - The transcript outlines the rapid rise of the startup Windsurf—originally founded as Exofunction and later rebranded from Kodium—a well‑funded AI coding assistance platform that pivoted to an AI‑native development environment, secured a $1.3 billion valuation, undercut competitors on price, and achieved FedRAMP High certification in early 2025. - [00:03:57](https://www.youtube.com/watch?v=qgQvci8NzM8&t=237s) **OpenAI Secures Windsurf Acquisition** - In early May 2025, after a letter of intent and exclusive negotiations, OpenAI finalized a $3 billion deal to fold Windsurf into its ChatGPT developer suite, coinciding with Anthropic’s withdrawal of Claude and mixed reactions from the developer community. - [00:07:27](https://www.youtube.com/watch?v=qgQvci8NzM8&t=447s) **OpenAI Deal Collapse Triggers Exodus** - The transcript outlines how the expiration of OpenAI's exclusivity agreement, compounded by Anthropic's strategic actions, forced users toward competitors and precipitated the collapse of a major enterprise deal. - [00:11:14](https://www.youtube.com/watch?v=qgQvci8NzM8&t=674s) **Google Deal Sparks Silicon Valley Trust Crisis** - The speaker argues that Google’s acquisition, driven by AI salary pressures, rewards founders for abandoning their startups and undermines long‑term trust in the Silicon Valley ecosystem. - [00:14:43](https://www.youtube.com/watch?v=qgQvci8NzM8&t=883s) **Cognition Labs Steps In** - The speaker explains how Cognition Labs is rescuing a Google‑sidelined developer platform (Devon/Windsurf), leveraging its FedRAMP certification for government contracts, reviving ties with Anthropic, and illustrating the extreme talent‑money dynamics driving Silicon Valley's talent wars. - [00:17:57](https://www.youtube.com/watch?v=qgQvci8NzM8&t=1077s) **Model vs. Tool Makers' Power Struggle** - The speaker outlines how model creators control roughly two‑thirds of the AI ecosystem’s power while tool makers retain about one‑third, using Windsurf’s acquisition and FedRAMP certification as examples, and predicts that code‑based agents will expand as a clear, rewardable use case. ## Full Transcript
Wind surf is one of the most insane
startup stories that I have run across
in couple of decades hanging out around
Silicon Valley and tech. We are going to
talk about it and we're going to start
by telling you the story of wind surf so
you can get a sense of how long this
saga has been going on. 2021
company was founded as exofunction by
MIT graduates Varun Mohan and Douglas
Chen. The original focus was GPU
optimization tools. So even then they
were in the AI space. They then pivoted
to AI coding assistance and launched
Kodium. 2024 they raised series C. The
valuation at the time was somewhere
between 1.25 billion and 1.3 billion and
the total raise to date is 243 million
almost a quarter of a billion dollars.
Investors at the time of the series C
include Founders Fund and Kleiner
Perkins.
Late in 2024 after the series C, they
had a major rebrand from Kodium to
Windsurf and they launched an AI native
development environment with the cascade
agent intended to be a competitor to
cursor. Everyone was seeing cursor take
off. This was the point we were seeing
bolt take off. Lovable was starting to
take off. Windsurf was positioned in
that ecosystem as a full stack app
builder. It had a free tier and then it
had a uh I think 15 buck per seat
premium at the time versus cursors 20
bucks. They're clearly undercutting on
price. 2025 is where things get
interesting. In Q1 on March 11th,
Windsurf achieves Fed Ramp High
certification, which if you don't know
what it is, count yourself blessed. It's
the first AI coding tool to be approved
for US government workloads. That's what
FedRamp High means. That gives it
obviously a major leg up in regulated
sectors and it plans to take that all
the way to the bank. Early in April,
Windsurf hits reportedly $100 million in
ARR, far behind Curser, but still
clearly in the number two position. It
is the lift to the Uber played by
Curser. This is up from $40 million just
a few months before. and they claim a
million plus developers using the
platform with 350 plus enterprise
customers. One of the things I will call
out here anecdotally is I hear wind surf
more with enterprise customers than I
hear it with individual engineers. And I
think that's a very interesting uh flip
from cursor because cursor I hear from
engineers all the time and I hear it as
yes they're going to consider it but
it's not that surprising if you consider
the ubiquitous number one. Windsurf has
better name recognition with decision
makers and punches above its weight in
decision-making circles. And I think
that's really interesting and that will
continue to play out as we as we
continue the story here. April 16th,
2025, Bloomberg breaks the OpenAI
acquisition news. The reported price $3
billion allcash deal. It would be
OpenAI's largest acquisition ever. We
are at a stage in AI and private
companies where you can have
multi-billion dollar acquisitions by a
startup of another startup. That is how
big AI is right now. Obviously, if
you're at Windsurf, you begin doing the
math on your equity. You want to say
like, how much do I get? What is my vest
date? Everyone goes and looks at their
RSUs. I've done this sort of song and
dance before. They do the math. They try
and come up with something. I will tell
you as someone who has lived through
these things a couple of times. Do not
do the math yourself. you will end up
with a number that is by definition
incorrect because you will forget the
taxes or you will forget the fees or you
will forget the like delay in the payout
due to some sort of uh clause that they
write in the contract. There are all
kinds of ways to change the math that
are legal. It is just there's no point
doing the math yourself. Just don't
bother.
Early May, the letter of intent is
signed and the exclusive negotiation
window begins.
internal memos at this point leak
showing that there's a plan to fold
Windsurf into the chat GPT dev suite.
This is also when Anthropic
is going to start withdrawing Claude
from Windsurf causing a lot of bad
feelings from the developers using
Windsurf. I was a user of Windsurf at
the time. I was very unhappy that Claude
left. But I get it. If you have
something called chat GPT dev suite
coming, the last thing you want is to
let Claude's output tokens and Claude's
data into that environment because
really what you're doing when you're
coding is you have an AI writing code
and you have an engineer rewarding that
one way or another. Either telling the
code output, hey model, great job on the
code output, good job, thumbs up, and
then you have a reward signal or hey,
this didn't work. Can you try it this
way? And then you have a correction
signal. Both are highly valuable signals
if you are in the AI arms race and
Enthropic did not want those signals
going to chat GPT. So OpenAI reportedly
agreed to definitive terms with Windsurf
around this Chat GPT dev suite idea.
May 6th, 2025, a definitive agreement is
announced which is a step beyond just
the negotiation window and the letter of
intent. A step beyond the uh reported
price negotiated initially. the $3
billion price was confirmed again. Uh
this is a rare case where the price
didn't change and Windsurf employees
rightly are celebrating at this point.
They think they've got it made, but we
aren't done yet. Early June negotiations
hit a major snag. Microsoft raises IP
concerns under 2023's agreement that
OpenAI signed with Microsoft back when
OpenAI had many fewer cards than it
holds today. Microsoft has rights to all
OpenAI tech, including acquisitions, and
it looks like Satcha Nadella played his
hand. The rumor is that OpenAI did not
want to give the IP for Windsurf to
Microsoft. There has been, I don't want
to call it a divorce, that sounds
dramatic. There has been a soft
separation between Microsoft and Open
the Eye. Let's use that term. Very
polite. We're all adults here. And there
has been a little bit of tension and
this is an example of that tension.
Basically, OpenAI has become more
powerful in the relationship since the
agreement was signed. They're flexing
their wings. In this case, they flexed
their wings and said, "You know what?
We're acquiring Windsurf and no
Microsoft. We don't want you to have the
IP." And Satcha basically said, "Talk to
my Microsoft lawyers." Which if you've
ever talked with Microsoft lawyers, you
don't want to do. So, at the end of the
day, Satcha was right. Microsoft had
legal rights. GitHub Copilot directly
competes with Windsorf and he was right
to worry about it and Microsoft refused
to grant an exception to their IP
rights.
June three and Anthropic formally
restricts Windsurf's claw access. This
is during the OpenAI exclusivity period
and they are probably doing it to sort
of see if they can scotch the deal in
addition to uh in addition to the token
impact that I discussed earlier. Um,
as Jared Kaplan put it, it would be odd
for us to sell Claude to OpenAI. That's
that's exactly sort of the frame that I
made, but it's it's much more sus
succinct.
So, this forces winfur users to migrate
to competitors. Now, you have a deal
that isn't done yet, and you have users
leaving the platform. Enthropic is
quietly happy uh because OpenAI is kind
of getting screwed on this deal. Mid to
late June, things begin to unravel.
OpenAI desperately attempts to
restructure the deal. The wind surfboard
begins to get worried about losing
enterprise and government contracts as
their core product starts to collapse
because Claude is starting to pull out.
Claude is, if you don't know, one of the
best coding agents out there. The deal
is effectively stalled. It's not
publicly dead. And there's a difference
there.
July 11th, that was this Friday, 7 a.m.
Pacific, the OpenAI exclusivity period
officially expires. No renewal is
announced and the deal is confirmed to
be under review, which is code for in
trouble.
By the morning, news breaks that the
OpenAI deal has collapsed. Microsoft's
IP issues are cited as the primary
cause. Although if you look into it and
you look at Anthropic's power move, I
would argue that's also a contributing
factor. If you start to lose the most
powerful coding agent, it's hard to make
the case that you're a very powerful and
$3 billion worth development platform.
This underscores the tension between
model makers and the tool layer right
above them. The information, which is a
tech publication, now reports talks have
ended on tech sharing concerns. By 300
p.m. that day, Friday, July 11th, Google
has announced a $2.4 billion deal. They
got on the phone right away.
They structure a reverse aqua hire with
a licensing deal. Now there I have
wondered whether this is effectively a
way to get around FTC examination of the
deal because it's not technically an
acquisition. It is a very unusual
acquisition format that isn't an
acquisition because it's an aqua hire.
They do not acquire the company. Instead
they hire the CEO Varun Mohan and the
co-founder Douglas Chen and about 40 of
the best engineers R&D employees. They
all get paid bank. They get paid. They
split between them 2.4 billion just for
the talent and a non-exclusive license
to wind surfs tech. Non-exclusive. You
understand? Mostly that is 2.4 billion
for 42 people.
That's a good deal. Like divide out
those numbers. That is a good deal. The
remaining quarter of a thousand
employees, the remaining 250 employees
are not included in this deal. Don't
know about this deal. Aren't
participating in this deal. their equity
is more or less worthless because what
they get is $100 million in remaining
cash in the company uh and they don't
have the option to compete on their tech
exclusively because Google can now
license their tech and they don't have
their best R&D employees. They don't
have their founders and they are left
screwed for lack of a better term. Now I
will say not everyone views it that way.
I thought that founding engineer well
founding engineer engineering uh maven
Gurgly Oros had a really good point when
he said honestly as bad as it may be if
you're in it the outcome for wind surf
across the board is better than most
startup outcomes and he is right most
startup outcomes 90% plus go to the wall
they failed the startup equity is worth
zero the fact that you got paid
and that some of your employees and your
founders got something is still better
than the zero outcome you would expect
on most startups. That may be true, but
I think what Girkley misses is the
trustbreaking nature of this choice.
There has been an implicit contract in
the valley where founders are in it for
the long haul with their companies. If
that starts to break down because the AI
salary pressures are so insane that it
is better to be an employee than to be a
founder at a major R&D firm like Google
for AI.
We have cracks in the Silicon Valley
ecosystem.
It does not sit well with me that a
founder would be rewarded for walking
away from their startup and leaving
their employees in the lurch. So as much
as I agree with Gurgaly that on paper
this is better than the outcome the
company could have had I don't think it
is a good thing that builds trust in the
ecosystem long term. I think it's
negative for the ecosystem. I don't
think Google did a service to the valley
here. And I get maybe why they felt they
had to do it because Google unlike
smaller startups like OpenAI would not
be under FTC scrutiny the way Google is.
And so Google had to tread very
carefully. And that may be part of why
the Aqua Hire deal happened.
But we aren't done yet.
By the evening,
Logan at Google, Logan Kilpatrick,
tweets a welcome and everything is
official. Controversy starts to explode.
Details of the structure leak on July
12th, Saturday. The 2.4 billion leaks.
Vested employees got cash payouts.
Employees with less than 12 months
tenure get nothing, which is actually
very typical in startups. And the
remaining company has no core team,
which I discussed. In the afternoon on
Saturday, it's announced that Jeff Wang
will be the interim CEO. The company
states their SAS focus will continue and
emphasizes the Google license is not
exclusive.
Social media explodes.
Uh, lot of commentary on employees
counting out sort of their rewards they
were going to get and then getting
nothing. Uh, channel partners begin to
pull support for for Windsurf. #
Windinssurf betrayal starts to trend. By
July 13th, Sunday, more internal details
leak. Apparently, there was a rambling
CEO Zoom speech announcing the deal.
Employee anger begins to peak.
By the afternoon, industry analysts are
weighing in. Predictions of bankruptcy
within 12 months. Remaining company
called a zombie. Enterprise customers
advised to migrate. This is all very
bad. If you are still at Windsurf,
right, for the 250 employees, this has
been a weekend from hell. And I I cannot
emphasize this enough. I would not want
to be you guys. I I am hoping that you
are in a better spot now. And we're
going to get to why. July 14th, Monday,
Cognition Labs, the makers of Devon,
announced an acquisition of the
remaining parts of Windsurf. This
includes their IP, their product, their
trademark, their brand, their training
data, and all remaining 250 employees.
The key difference here is there is 100%
employee financial participation, full
vesting, acceleration, and wave cliffs.
In other words, Cognition Labs saw what
Google did and said, "We have an
opportunity to step in as the good guy
here. We can be the good guy. We will
acquire Wind Surf. We'll bring everybody
aboard. It doesn't matter whether you
vested or not. If you've worked less
than a year or not, which by the way,
when you're in a startup and it's hiring
like on a power law like this, most of
your employees by definition are less
than a year in because you're hiring so
fast. So, I would expect for most of
these employees, waving the cliff was a
big deal.
Essentially, Cognition Labs makes them
whole where Google screwed them over.
And it's a nice move for them because
they're basically saying we have an
agentic developer in Devon. Now we have
tools for a development environment in
Windsurf. In theory, if they can get
themselves sorted out, they still have
that Fed ramp high certification which
could get Devon into government work.
There's some upside here for Cognition.
It wasn't just that they were being
charitable. Now, of course, the pricing
isn't disclosed. I can't imagine it was
$3 billion, not least because Cognition
doesn't have the money that OpenAI had.
By afternoon,
more conflicting reports start to
surface.
Uh, Windinsurf reposts the Cognition
announcement. The show is going on.
Cognition emphasizes that we're friends
with Anthropic again, meaning that they
think they're going to be able to get
Claude back into Windsurf to revive it.
We will have to see how this all goes.
At the end of the day,
I want to draw a couple of lessons. One,
this is underscoring the
skewed nature of the talent wars and the
power law for talent wars in Silicon
Valley in a way I have never seen
before. If you can pay a founder who is
already a multi-millionaire so much
money that founder finds it worthwhile
to abandon their fiduciary
responsibilities to their company,
potentially expose themselves to a
lawsuit because there there's some board
level responsibility here to uh work at
the best interest of the company and
it's arguably not the best interest of
the company to walk away from the
company and join Google for millions of
dollars. So they're exposing themselves
to that and on top of that they are
frankly breaking a code of honor in the
valley that has been true for decades.
That's still worth it to them because
Google's paying them so much. And that
underscores how massive these pay
packages are. Over like a weekend ago,
Zuck got the I think the head of AI at
Apple. I don't know why you would go to
Apple for that, but he got the head of
AI at Apple for $200 million, which is
more than Tim Cook at Apple makes, I
think, depending on how you count the
equity investing. Anyway, it's a ton of
money.
When you make that much, you break the
usual incentives at Valley. If you can
be a $200 million employee, why would
you ever found?
And I think that we are starting to see
some of that dislocation. The second
thing that I want to call out is that
there is massive dislocation in the tool
layer and that we should view coding
agents, the coding environment as the as
the tip of the iceberg for the
disruption that model makers can pose.
Because at the end of the day, if you
are a model maker and you believe good
models will get better over time, you
should be able to eat more and more of
workflows over time just by delivering
good models. And you can do something
like, I don't know, clawed code, which
is frankly just a terminal, and compete
with cursor,
you're going to eat some of these tools
for lunch over time, and you kind of
don't care. And so we are seeing this
tectonic struggle between the toolmakers
that are trying to build on top of this
modelmaker tech and the model makers who
are really incentivized to grab as much
territory as they can.
Modelm makers hold the cards here as you
saw with Claude cutting off access to
Windsurf but they don't hold all the
cards. And so one of the things that's
super interesting about this dynamic is
that it's sort of a twothirds one-third
situation where you might say that
twothirds of the power sits with the
model makers, but the tool layer has a
third of the power. At the end of the
day, Windsurf still got acquired.
They're still going to be a viable
company. The Fed Ramp High certification
did not go to Claude Code. It went to
Windsurf. And so tool makers are still
going to get their day in the sun.
They're still going to get their
dollars, but the way they negotiate the
relationship with model makers is going
to continue to be fraught because the
model makers have a lot of incentives to
make sure that their models continue to
get better and the toolmakers are going
to have to negotiate those incentives
while still serving customers. And
that's sort of the story that tension is
the story of Windsurf over the last
couple months. The third thing I want to
call out is that we are going to see
more and more agents in code. And again,
that is a cuttingedge tip of the spear
for where we will see them other places.
The reason code is leading the way here
is because code is easy to reward. Code
runs or it doesn't. It's easy for model
makers to figure out whether their code
worked or it didn't. And that's true for
development environments. It's true if
you launch your own agent. Cursor
recently launched their own model that
codes. Uh and it's true if you have an
agent that runs on code. All of the
stuff we're seeing in the development
environment now, this whole conversation
around wind surf is an early indicator
of where the rest of the industry is
going because it's an easier problem for
AI to solve.
When you think about the next 12 to 15
to 18 months,
Devon acquiring windsurf means more work
on agents in code. It means the agentic
agenda for lack of a better term. Not
that it's a secret cabal, but the
agentic goal of like getting AI agents
with tools and policy and guidance to
actually do important coding work that
is moving forward. Devon is, I would
expect, going to be fully integrated
into Windinsurf and we're going to see
more of this hybridized uh motion where
you have both a development environment
and an agent management environment for
engineers in the same place. Cursor
heavily going that way as well. That is
going to lead to new surfaces, new
experiences. The mobile environment has
become a much bigger deal for developers
than it ever was. It used to be you
could just build for the desktop because
always developers would be on the
desktop. Not anymore. If you're running
agents, you can be on your mobile.
There's a lot of really weird
implications and that's very typical of
what happens when you have an underlying
technology shift like we are all living
through right now. Fundamentally, the
story of Windsurf is the story of the
evolution of LLMs from being just sort
of autocompletes for coders to being
genuine coding agents to being genuine
uh assistants in code. And what happens
when valuations start to shake loose at
that point? That's why OpenAI went for
Windsurf in the first place. That's why
they got the Fed Ramp High
certification. That's why Devon went for
them. And that is also why that tech is
why Google needed them because Google
doesn't have a development environment
that is popular with developers and if
they're going to get Gemini to win long
term they need it. So that's the story
of Windsurf. I am sure there will be
more to the story. It's been one of the
most eventful of the last couple of
decades I would say and I'm curious what
do you think you would do in Windsurf
shoes?