Applied Computing raises $20M Series A to unify fragmented oil and gas plant data with physics-aware foundation model
London startup's Orbital model combines time-series, physics, and language AI to compress multi-week facility investigations into minutes, attracting KBR and energy majors as early customers.
Last verified:
Orbital’s industrial reasoning gap
Applied Computing, a London-based startup founded in 2023, raised $20 million in Series A funding led by engineering conglomerate KBR, with Databricks Ventures participating. The funding accelerates deployment of Orbital, the startup’s foundation model designed to synthesize fragmented industrial sensor data into real-time facility predictions for oil, gas, refining, and petrochemical plants. According to TechCrunch, Applied Computing has grown from stealth to double-digit millions in annual recurring revenue within 18 months—an unusual velocity for an industrial AI startup with no disclosed customer count.
The core problem Orbital addresses is acute: industrial facilities generate thousands of sensor streams measuring temperature, pressure, velocity, and chemical viscosity, yet operators make decisions using less than 8% of available data, according to Applied Computing CEO Callum Adamson. The bottleneck is not data collection—equipment already gathers the measurements—but integration. Operators struggle to unify real-time sensor readings, static engineering documentation, and domain-specific physics and chemistry knowledge fast enough for actionable analysis.
How Orbital differs from conventional LLM architecture
Orbital departs from large language model design by fusing three distinct model types: a time-series component that analyzes sensor trends over time, a physics-based model that encodes chemical and thermodynamic constraints, and a language model that interprets equipment metadata and operational commands. This hybrid approach lets Orbital flag facility anomalies, identify root causes, and simulate downstream effects of proposed fixes—all within minutes. Adamson told TechCrunch that Orbital compresses investigations that once consumed days or weeks into seconds, reducing energy consumption and stabilizing output.
The model also enables technicians to run “what-if” simulations across interconnected systems, addressing a critical pain point: a maintenance action in one section of a refinery or chemical plant can cascade into failures elsewhere if not carefully modeled. By predicting these interactions, Orbital helps operators avoid expensive unforced downtime.
Market adoption and competitive landscape
Applied Computing has embedded Orbital into the workflows of multiple large, publicly listed energy operators, though the startup declined to disclose exact customer names or count. According to TechCrunch, partnerships include Indian energy conglomerate Wipro and KBR, which has integrated Orbital into its INSITE 3.0 digital platform, specifically for ammonia production modeling. Adamson also signaled imminent announcements with a major U.S. upstream operator and a European oil major.
The startup enters a crowded market dominated by entrenched suppliers. AspenTech and AVEVA offer established simulation and AI-powered modeling software for upstream, downstream, and chemical operations—platforms with deep integrations and customer lock-in. Applied Computing’s thesis is that a purpose-built, physics-aware foundation model can compress decision-making cycles and reduce operational friction in ways legacy software cannot.
Why This Matters
Industrial AI adoption has historically faltered on the integration problem: energy operators possess the raw data but lack tools to synthesize it into timely decisions. If Orbital delivers on its promise to collapse weeks-long investigations into minutes, the model could reshape capital allocation in industrial operations—shifting dollars from reactive maintenance and energy waste toward predictive optimization. For KBR and its customers, the integration signals confidence that foundation models can reason about constrained physical systems, not just language.
The success of Applied Computing will also test whether hybrid architectures (time-series + physics + language) can outcompete pure LLM approaches in domains where physics constraints are non-negotiable. If Orbital’s early deployments sustain their claimed speed and accuracy under independent scrutiny, the model could become a template for foundation models in other physics-constrained industries: power grids, semiconductor fabs, pharmaceutical manufacturing, and aerospace. Conversely, if the startup struggles to expand beyond energy majors with the resources to deploy new software infrastructure, the tight coupling between facility geometry, sensor placement, and Orbital’s training data may prove a moat that limits scalability.
Frequently Asked Questions
How is Orbital different from a large language model?
Orbital combines a time-series model (for sensor trends), a physics-based model (for chemistry and thermodynamics), and a language model (for understanding equipment and operations), whereas LLMs predict the next word. This hybrid approach allows Orbital to reason about facility state and predict cascading effects across interconnected systems.
What data sources does Orbital integrate?
According to Applied Computing CEO Callum Adamson, the model unifies sensor readings (temperature, pressure, velocity, viscosity), engineering documentation, and physics/chemistry knowledge—three data streams that operators previously struggled to reconcile in real time.
Who are Applied Computing's customers and partners?
The startup is deployed at large publicly listed upstream, refining, and petrochemical operators (names undisclosed), with confirmed integrations at Indian energy firm Wipro and engineering giant KBR (via its INSITE 3.0 platform for ammonia production). Adamson also cited work with a major U.S. upstream operator and an upcoming European oil major partnership announcement.