How Two French Engineers In New York Built The Company That Monitors The Entire Cloud
In a Nutshell
Olivier Pomel and Alexis founded DataDog in 2010 after experiencing the Dev-Ops divide at their previous company, building a unified observability platform that rode the cloud and DevOps wave they initially underestimated. The company succeeded despite YC rejection, painful fundraising, and New York location stigma by staying deeply connected to customer feedback and product decisions rather than delegating to strategy. Now public with 20-25 products and heavy AI investment, they're betting on agents using their platform directly while fundamentally restructuring teams around AI productivity gains that let 2-3 people accomplish what previously took 6 over months.
These notes were generated by AI and may contain inaccuracies.
Olivier Pomel and Alexis shared their experience of being rejected by Y Combinator. DataDog did not get into Y Combinator despite applying and getting an interview. Olivier still has the rejection email from Paul Graham. When Olivier agreed to do a session, he received an invite to a YC dinner in Paris, but was told he wasn't invited anymore because they weren't part of YC, resulting in being rejected twice.
Olivier graduated in 1999 and did an end-of-study internship at IBM Research in upstate New York, which at the time was a legitimate research place. Alexis, his co-founder who was a couple of years ahead in school, had already been to that internship and was still working at IBM. They met there and spent time at IBM together. Olivier thought he would stay in New York for 6 months but is still there 26-27 years later.
During the dotcom boom, startups were everywhere in New York. Olivier started working for startups and worked together with Alexis. After the dotcom bust, things got worse with September 11th. Olivier decided to stay for personal reasons and started working again with Alexis in an educational software startup. At the time, after the bust, nobody thought technology would be worth anything again. The company was successful, growing to around 800 people when they left. Olivier was building the dev team while Alexis was building the ops team.
At the educational software company, they lived through developers hating operations and operations hating developers, with people pointing fingers at each other. There were already observability companies out there, though it wasn't called observability - it was called monitoring. The existing monitoring was very job-specific and very reactive. For network monitoring, there would be a product for the ops team, but developers would never touch it. Developers were essentially blind to anything that happened in production.
They started DataDog in 2010 with the premise of bringing Dev and Ops together into one platform. What they didn't realize at the time was that this premise was completely central to the explosion of the cloud. They also completely underestimated the cloud explosion - they thought Amazon was interesting but when talking to real companies, they all said it was a toy they would never use. It turns out they all did.
The bet was DevOps together, and the first infrastructure they supported was cloud infrastructure because they talked to smaller, more modern companies that were built on the cloud. This became the tailwind of everything as the cloud market exploded and DevOps were brought together.
Their product was called infrastructure monitoring initially - the term observability didn't exist yet. It didn't look like infrastructure monitoring that came before. Initially they had called it something different, like a data platform for DevOps to collaborate, because they thought monitoring was a thing from the 90s that was boring and old. However, people love the idea of a new thing but have no idea what they should pay for it, whereas when you say it's infrastructure monitoring, everybody understands it's a real category they need.
It was not obvious that DataDog would work. They didn't get into Y Combinator. The first round was an angel round that was super painful. They couldn't get any smart investors to pay attention. When pitching VCs in New York, the investors would be interested until they saw the competitor slide and realized they didn't understand any of these companies, then they'd pass. When pitching Bay Area VCs, they only invested in the Bay Area, and it was seen as a form of mental impairment to try to start an infrastructure company not in the Bay Area.
They made a number of things different from what Bay Area companies do. They never wrote down the culture - at the time every company had five values or eight principles. Instead, they said if you need "don't be evil" written on the wall, you probably shouldn't work there. Culture flows from the top, so they exemplify the culture and hire, promote, and fire based on the culture. Every year they ask themselves if it's the year they need to write it down, especially as they need to hire and train thousands of people, but so far the approach still works.
Olivier stays heavily involved in product decisions rather than spending time on strategic slide decks. He hates the word "strategic" and mentally replaces it. Product is very important for an infrastructure and product company. It's important for management to be aware of everything happening in product and read all new developments. Being high up allows editing decisions - saying something is too much or not enough, or stopping something if it's not understandable to customers.
He encourages everyone to keep in touch with what's actually happening on the ground: actual support requests, actual sales transcripts, actual customer feedback, and actual employee comments from surveys. He still reads all the employee survey comments. When sampling this information, you get a good sense of the actual fabric of what's happening rather than beautiful summaries that find their way up.
When he discovers something from reading surveys or customer feedback, he picks up random emails or Slacks and replies with a question saying he doesn't understand what's going on. This causes two things: the person receiving it wonders why he's looking at that, and the positive effect is that the whole management chain starts looking down to understand what's actually happening rather than managing up.
The company now has around 20-25 products. They see a lot of usage from customers around their product - customers build their own extensions and workflows. This gives them ideas of what problems customers have and where they think DataDog belongs. They build a lot of things they see customers doing.
Sometimes they have more strategic projects where nobody's asking them to do something but they think the world is going that way. These tend to have potentially bigger impact because they can be brand new things, but you tend to be wrong more often because you just made it up.
With AI in particular, things are changing so fast that you can't wait to see demand materialize from the customer base - you need to get ahead of it, which means getting wrong more often and needing to be at ease with that.
They've been public since 2019. Their lockup expired the day of COVID lockdowns, which coincided with a big market crash. One of their biggest investors got scared and dumped everything that day. The stock went down around 65%, from being worth $10 billion to around $4 billion. The crash lasted two weeks, then the market went up like crazy in 2021-2022, then started crashing again.
Being public hasn't fundamentally changed how they lead the company. As a private company, you spend a lot of time talking to investors because you need to raise funding again, and every investor should believe you're the future. As a public company, investor relations is very choreographed - a couple of days every quarter on earnings calls, spending a week preparing the right message.
The main issue created by being public is if you do something unpopular with the market, you might lose 20-30-40% of stock value. While it doesn't impact the company in the short term, it creates a compensation issue because people are paid in RSUs. The actual risk isn't company survival but losing key performers because their equity is worth less. People join at different points in time, so their compensation is at different levels, requiring more detailed comp management.
The AI revolution has caused rapid changes in market perception. Early on it wasn't obvious whether DataDog would be a winner or loser. Now they're on the winner side. They started in the "don't know" side from investors.
There's a ton of disruption, and for public investors especially, it's really hard to understand what things look like three, four, five years from now. Public investors have less time and look at 20-30 different names, coming in and out a couple of times a year, whereas VCs are in 7-8 companies and stick around for 5-10 years.
The approach is to tell the right story and produce numbers that agree with that story consistently. For the company side, working with AI means taking more shots and being wrong more often. The way development has changed three times in the last year - how you write code or don't write code. It's impossible to predict a year ahead, so you have to be very reactive and try a lot of things all the time.
They're investing heavily so the product can be used by agents, as they see a lot of that with customers. They're building a lot of the smarts directly into the product - if you use Claude Code to work on ops issues, it's like trying to send a postcard by buying a business class ticket. There's a lot that should be able to be done directly fused on the data plane inside observability rather than outside of it. They're building so the system can do everything on its own.
AI companies growing super fast also need observability. They have the top 10 AI companies in the world as customers, plus startups building on AI and big enterprises. The top 2-3 AI labs use their product in weird ways because they have infinite compute and inference and models nobody else has yet, giving a glimpse of what's coming, though it's not necessarily representative since the rest of the world doesn't have infinite compute.
Datadog is not token maxing internally. Co-founder Alexis has told the engineering team that within two quarters they will not be writing code anymore. The reality is that the team will still write and read code, but the balance inverts from mostly writing with occasional automation to mostly automating with occasional writing.
The shift changes everything about team leadership and hiring. Smaller teams can accomplish significantly more, allowing a team of two or three to explore ideas instead of a team of eight. However, the destination remains unclear—no one knows what processes will look like a year from now, or how many PMs, designers, security people, and engineers will be needed to build products.
The company has already seen major ROI. In December, as models improved and during the holiday period when people had time for hack projects, excellent developers rebuilt entire systems alone in a couple of days—work that previously required a team of six over six months. This convinced leadership to pivot the organization despite the pain involved.
Alexis and the speaker have worked together for nearly 20 years. Before founding the company, they worked together for almost 10 years without starting a business, which provided valuable experience in establishing working boundaries. They maintain regular communication through standing lunches without specific agendas, allowing space to discuss whatever arises. This relationship provides essential support since founders have few people they can truly confide in.
It is now possible to start a company anywhere in Europe and secure funding, unlike when the founders started Datadog. However, pursuing the US market as soon as there is product-market fit remains absolutely critical. The whole company does not need to relocate to the US—typically one founder moves while the rest of the team can remain in Europe.
The founders initially moved to San Francisco after starting in Europe in 2012 but kept the team in Europe. They later built a Paris office primarily for talent acquisition, as French employees in the US either couldn't renew visas or wanted to return to France for family reasons. The company hired many French employees through internship programs.
The advice the speaker would give their 15-year-old self is to always move faster, especially regarding hiring and firing. While firing is commonly cited as needing speed, hiring faster is equally important. When seeking roles like head of sales, founders should hire promising candidates immediately rather than waiting for the perfect profile. This allows running more cycles and learning through actual experience, since it's difficult to know if someone will work out before trying them.
The speaker made the mistake of being paralyzed trying to make perfect hires. They also note that the first time firing someone feels horrible, with fears that everyone will hate you, but typically the rest of the team asks what took so long.
Keep Y Combinator in your library
Save the videos and channels worth coming back to, and find them again in one place.
