Ex-Palantir Founders Raise $38 Million to Rival Their Old Employer

Arrakis has existed for seven months. In that time the London startup persuaded investors to hand over $38 million and a $140 million valuation, emerging from stealth this week with backing from OpenAI’s and Datadog’s own executives. The company, founded in January, helps industrial firms push AI agents into factories, oil and gas sites, and aerospace hangars.

Two of its four founders spent years at Palantir before starting it. They are now running Palantir’s own forward-deployed engineer model, the one that built their former employer’s reputation and its moat, aimed at an industrial sector Palantir is chasing too.

A $140 Million Valuation in Under Four Months

Accel led a $7.5 million seed round in March. By July, Blossom Capital had returned to lead a $30 million Series A, with Accel following on, according to reporting from Tech.eu, which first covered the round in Europe. Cofounder and chief executive Rafael Quintanilla told Fortune the Series A values Arrakis at $140 million post-money.

The backer list reads like a checklist of enterprise AI credibility. It includes:

• Blossom Capital – led the $30 million Series A
• Accel – led the March seed round and returned for the Series A
• GFC, MainObject and Rerail – additional venture participants in the Series A
• Olivier Pomel – founder and chief executive of Datadog, investing personally
• Olivier Godement – OpenAI’s head of business products, also investing personally
• Junaid Hussein – founder of Cambridge Aerospace, joining as an angel

The money is earmarked for offices in New York and the Middle East, plus platform development. The company plans to triple its roughly 15-person team to around 45 people over the coming months.

Two Co-Founders Came Straight From Palantir

Quintanilla was a vice president at Accel before founding Arrakis. During his time there he spent close to a year crossing the United States, Europe, and the Middle East building the firm’s thesis on defense and industrial resilience.

His three co-founders came from elsewhere. Haroun Beltaifa and Romain Fouilland both previously worked at Palantir. Mikhail Galkov was an engineer at Delivery Hero. The wider team also draws alumni from Revolut, Datadog and ASML, according to the company’s own announcement of the round.

That Palantir lineage is not incidental. It shapes how Arrakis says it operates: model-agnostic software paired with engineers who sit inside a client’s operations rather than selling a shrink-wrapped product off the shelf.

What Is a Forward Deployed Engineer?

A forward deployed engineer is a software engineer embedded directly at a client site, building and rewriting the product against real operational problems in short cycles, instead of shipping a fixed, one-size-fits-all system from a distance. Palantir did not just use this model. It invented it.

According to a Stanford summary of the company’s history, Palantir built its innovation engine by placing engineers with end users and revising the product roughly every two weeks based on direct field feedback, a method the university’s own session materials trace back to the company’s earliest counterterrorism and intelligence work before it spread across the industry, as described on embedding engineers directly with client teams. Y Combinator’s own chief executive, Garry Tan, who was Palantir’s tenth employee, has since told founders they should see themselves as

• Forward deployed engineer – an engineer who lives with the client’s data and workflow instead of building in isolation, closing enterprise deals through hands-on delivery rather than a generic sales pitch

Arrakis says its version of the model lets industrial companies train, deploy and scale AI agents in weeks rather than the months or years legacy vendors typically require.

A Well-Traveled Exit Ramp From Palantir

Arrakis did not invent the Palantir-alumni-to-founder pipeline. Venture research firm Concept VC has spent time tracking dozens of Palantir alumni founders, and found that a striking share of them held forward-deployed engineer or deployment strategist titles before starting their own companies.

The pattern shows up across very different industries.

Fourth Age is the strangest entry on that list. It sells forward-deployed engineering back into Palantir’s own customer base. Arrakis is doing something close to the opposite: pointing the same skillset at the industrial accounts Palantir wants for itself.

Competing Against the Company That Trained Them

Fortune’s exclusive interview with Quintanilla noted that Arrakis is far from alone in chasing manufacturing and industrial clients. Consulting firms Accenture and Boston Consulting Group are both racing into industrial AI, Palantir is doing the same work at far greater scale, and Jeff Bezos-backed Prometheus, now valued in the tens of billions of dollars, is pouring money into automating the engineering of physical products. Quintanilla argues Arrakis is carving out a distinct niche from all of them.

Blossom Capital’s managing partner framed the bet around where AI budgets actually go.

Rafael and the Arrakis team have an exceptional combination of technical depth, operational expertise and commercial ambition. While many companies are focused on AI applications at the edge of the enterprise, Arrakis is tackling some of the most complex operational challenges facing large industrial organisations.

Ophelia Brown said that in the company’s funding announcement. The framing matters because Arrakis arrives inside a much bigger financing wave for enterprise AI agents. Paris-based Archestra.AI recently closed a seed round guarding AI agents’ access to company data, while Dust raised $40 million for multiplayer enterprise AI aimed at getting whole teams, not just individuals, working alongside agents. Investors are clearly betting the deployment layer, not just the model layer, is where enterprise AI money will actually land.

Where the Forward Deployed Model Gets Expensive

Landing conservative European industrial firms is its own fight. Quintanilla called his early traction pattern an open secret: family-controlled businesses. He said they think long term and can push top-down initiatives through in a way that larger, more bureaucratic firms cannot.

That is a real advantage for closing early deals. It is also a narrower starting market than the NYSE-listed enterprises Arrakis says it already counts as customers across energy, logistics and industrial sectors.

The forward-deployed engineer model has a well-known weakness. It is a services-heavy way to sell software, and every new client adds cost through the engineers embedded to make the system work, rather than mostly through code shipped once and resold many times. Whether Arrakis can grow past its first handful of clients without that cost structure eating its margins is the open question hanging over the round.

Enterprise AI agents fail in more ways than pricing alone. Amazon’s own experience with agent tooling showed how governance gaps behind Amazon’s Kiro outages can undercut a deployment even when the underlying model works fine. Arrakis is selling directly into industries such as aerospace, energy and manufacturing, where a failed agent does not just annoy a user. It can shut down a physical operation.

The Market Arrakis Is Racing to Define

The category Arrakis is chasing is growing fast on paper. Research firm MarketsandMarkets values the global AI agents market at $7.84 billion in 2025, growing to $52.62 billion by 2030, driven largely by enterprises adopting autonomous, task-executing systems instead of simple chatbots.

Arrakis’s pitch leans directly on that gap between experimenting with a chatbot and actually deploying an agent that finishes a business task. Quintanilla put the company’s mission in blunter, more political terms when the round closed.

“The West is under growing pressure to reindustrialise, but that renaissance won’t be powered by net new companies alone,” he said. “It requires equipping our industrial champions with the tools to harness their data, navigate the AI transition and compete on a global stage.”

Whether that equipping comes from a seven-month-old startup staffed by the same alumni network that keeps producing Palantir’s competitors, or from Palantir itself, is now a live question with $38 million riding on one answer.