Arrakis, a London and Paris company building AI agents for industrial operations, has come out of stealth with a $30 million Series A led by Blossom Capital. Existing investor Accel participated alongside GFC, MainObject and Rerail, Fortune reported, bringing total funding to about $38 million including a $7.5 million seed that Accel led. Fortune puts the post-money valuation at $140 million.
The company was founded in January 2026 — seven months before the announcement. Its founders are CEO Rafael Quintanilla, a former Accel investor, with Haroun Beltaifa, Romain Fouilland and Mikhail Galkov; Tech.eu reports the founding team includes Palantir and Delivery Hero alumni. Angel investors include Datadog CEO Olivier Pomel, OpenAI head of business products Olivier Godement, and Cambridge Aerospace founder Junaid Hussein.
The Thesis
Quintanilla’s argument, as he put it to Fortune, is a direct challenge to where AI money has gone so far: “Most AI investment to date has targeted the 30% of workers behind a desk. The real ROI lies in the 70% running industrial operations.”
For this publication that sentence needs no translation. Construction is the canonical 70% industry — enormous headcount, thin margins, work that happens in physical space, and a software layer that has historically served the trailer better than the field. If the thesis is right anywhere, it should be right here.
It also lands against a gap this publication has covered directly. McKinsey’s estimate that AI could automate 39% of construction’s nonphysical work sits awkwardly beside single-digit daily adoption. Arrakis is making the same observation from the investor’s side of the table and treating it as an opportunity rather than a paradox.
The Construction Claim, Examined
Here is where care is required, because the sourcing on this point is genuinely mixed.
Arrakis’s own website lists six target sectors, and “Engineering and Construction” is one of them, alongside aerospace and defence, chemicals, energy and commodities, shipping, and telecommunications. Tech.eu’s write-up also names construction among the sectors served.
Fortune’s account — the most detailed reporting on the round — does not mention construction at all. It names aerospace, energy, logistics and manufacturing, says the company had five customers at the time of writing, and describes one of them as a New York-listed shipping company that cannot be named because of NDAs. Tech.eu similarly refers to NYSE-listed customers in energy, logistics and industrial sectors without naming any.
So the accurate statement is this: engineering and construction is a sector Arrakis says it targets, and there is no public evidence yet of a construction customer. That is not a criticism of a seven-month-old company. It is the ordinary condition of a startup that has raised on a horizontal industrial thesis and is landing its first logos where the sales cycle is shortest. But it is the difference between a contech company and an industrial AI company with construction on the roadmap, and the two deserve different levels of attention from contractors.
The performance figures circulating with this round need the same treatment. Quintanilla told Fortune the company’s approach delivers a two-to-four-fold quality improvement while cutting token costs by roughly 70%, and Fortune reports that the shipping customer moved from monthly to daily cash-flow visibility. Those are founder-supplied claims about unnamed customers. Some sharper accuracy figures have circulated on social media; they do not appear in either Fortune’s or Tech.eu’s reporting, and we have not repeated them here.
Why It Still Matters to This Industry
Two reasons, both structural.
The first is that the competition for construction’s AI budget is no longer confined to construction software. Arrakis is middleware for industrial workflows, sold on outcomes — its site describes value-based pricing where clients “only pay for the value our Agents deliver.” A company like that does not need to understand construction the way Procore does. It needs to understand procurement, scheduling and document flow well enough to run them, and those problems rhyme across heavy industries. The threat to vertical contech is not that a horizontal player builds a better takeoff tool; it is that a horizontal player makes the surrounding workflow layer generic.
The second is the pricing model. Outcome-based pricing is unusual in construction software, where seat licences dominate, and it is a natural fit for an industry that is instinctively suspicious of software ROI claims. If agents can be sold against measured results rather than user counts, that is a meaningful change in how this industry buys.
Arrakis is also the second company we have covered in a month whose thesis treats construction as one line on a broader infrastructure bet. 1001 raised $30 million for “sovereign AI” across ports, energy, airports and construction megaprojects, and the shape is near-identical: build the horizontal operations layer, list construction among the target domains, and find out which vertical bites first. Contractors evaluating either should ask the same question — is this a product for me, or a product that might one day be pointed at me?
What to Watch
The single most informative thing Arrakis could do for this industry is name a construction customer. Until it does, the E&C listing on its website is a statement of ambition, and contractors should read it that way.
Beyond that, watch whether the horizontal thesis survives contact with deployment. The reason vertical software exists is that the last 20% of any industrial workflow is idiosyncratic, and construction’s version of that 20% — the subcontractor structure, the change order, the fact that every project is a new company that dissolves at the end — has defeated a long line of generic enterprise tools. Arrakis has raised enough money and hired from enough serious places to get a real shot at it. Seven months in, that is all that can honestly be said.
