Arrakis raises $38M for industrial AI agents
Arrakis is targeting the difficult last mile of enterprise AI: deploying agents inside complex industrial operations rather than office workflows.
Most AI-agent startups begin with digital office work. Arrakis is taking the harder route into factories, logistics networks, energy systems and other environments where software decisions affect physical operations.
What happened
Arrakis emerged from stealth with approximately $38 million raised in a matter of months. The total includes a $30 million Series A led by Blossom Capital and an earlier $7.5 million seed round led by Accel.
The company was founded by former Accel investor Rafael Quintanilla alongside executives and engineers with experience at Palantir and Delivery Hero. Arrakis describes its platform as an AI deployment layer for industrial enterprises. It helps customers identify valuable workflows, connect agents with operational data and existing systems, and move projects from prototypes into production.
Its target sectors include aerospace, energy, manufacturing, construction, logistics and telecommunications. The new capital will fund product development and expansion into the US and Middle East.
Why it matters
Industrial companies often have valuable data but fragmented software, strict safety requirements and processes that cannot tolerate frequent errors. A general chatbot is not enough. Agents must understand specialised workflows, interact with legacy systems and operate within clear permissions.
That creates a potentially valuable market for companies that can handle integration and deployment—not just provide a model. Arrakis is effectively betting that the difficult implementation work around AI will become a defensible product category.
The bigger picture
The enterprise-AI market is moving from experimentation toward accountability. Customers increasingly want evidence that an agent can reduce downtime, improve throughput or accelerate a specific operational process.
Arrakis’ opportunity is large, but the model may be service-intensive. The company will need to show that lessons from one deployment can be turned into repeatable software rather than requiring a bespoke consulting project for every customer.