The short version: Applied Computing, a London startup building a foundation model for oil, gas, and petrochemical plants, just raised a $20 million Series A. KBR, the engineering giant, led the round. Databricks Ventures joined. TechCrunch AI reports the company went from stealth to double-digit millions in annual recurring revenue in under 18 months.
That last number is the one worth staring at. Most foundation model startups burn cash for years before anyone pays. This one found customers fast.
What Applied Computing actually built
The model is called Orbital, and it isn’t an LLM. It’s three things stitched together:
- A time series model that reads sensor data as it streams in
- A physics-based model that knows how heat, pressure, and chemistry behave
- A language model that handles the engineering documents and operator notes
An LLM predicts the next word. Orbital predicts the next state of the plant. It watches thousands of sensors tracking temperature, pressure, velocity, and viscosity, weighs them against physical constraints, and factors in what the operators are doing. Technicians can also run simulations: change one thing here, see what breaks over there.
Why this matters
Here’s the stat that explains the whole company. According to TechCrunch AI, co-founder and CEO Callum Adamson says facilities make operating decisions using less than 8% of the data available to them.
Not because they don’t collect it. They do. The problem is fragmentation. Sensor readings live in one system, engineering docs in another, physics knowledge in someone’s head. “It’s getting those three data sources to talk to each other in real time,” Adamson told TechCrunch. “That’s the real key.”
So a refinery sits on a mountain of telemetry and uses a sliver of it. When something goes wrong, engineers spend days or weeks tracing the cause. Applied Computing claims Orbital compresses that into minutes: flag the anomaly, find the cause, model whether the proposed fix creates a new problem somewhere else.
What stands out here is the framing. This isn’t AI writing reports about the plant. It’s AI modeling the plant.
The competitive picture
Applied Computing is walking into a room that’s already crowded:
- AspenTech sells simulation and AI modeling for upstream, refining, and chemicals
- AVEVA does physics-based process simulation and what-if modeling
- Cognite and Seeq own the data layer and analytics workflows
Adamson’s answer to all of them is blunt. The moat isn’t data or domain knowledge. It’s talent. “It’s an AI problem. It’s not a data problem, and it’s not an energy problem,” he said. “If you’re a tier-one AI researcher, where are you going to work? … I don’t think Shell’s on that list.”
That’s a real argument, and it’s the same one playing out across every vertical AI category right now. Incumbents have the customers. Startups have the researchers. Whoever closes their gap first wins.
The KBR angle
The lead investor being an engineering firm rather than a VC tells you something. KBR has already integrated Orbital into its INSITE 3.0 digital platform and is using it for ammonia production. That gives Applied Computing three things money can’t buy on its own:
- Access to real operational data from working plants
- Industry expertise in the building
- Warm introductions to KBR’s customer base
That data point matters more than it sounds. Adamson notes refinery operational data is basically never public, and simulated data can’t fully reproduce what happens inside a live facility. Every deployment feeds the model something competitors can’t buy.
What’s next
The $20 million goes toward international expansion, research and engineering hires, and new deployments. The company opened a Houston office this week, adding to its London headquarters and Bengaluru operations hub. Middle East expansion is in the works. Adamson says a partnership with a European oil major gets announced in the coming weeks.
Wipro is already a partner. Orbital is running at “large, publicly listed” upstream and downstream companies, though Adamson wouldn’t say how many.
The takeaway
Watch this pattern, not just this company. Foundation models built for one industry, trained on data nobody else can reach, sold to buyers with real budgets and measurable ROI. Energy is the proving ground. Chemicals, mining, and heavy manufacturing have the same shape of problem and the same pile of unused sensor data.
If Applied Computing’s revenue curve holds, expect a wave of copycats aimed at every plant floor on earth.
Full details are at the original source.