Oil and Gas Plants Get a Breakthrough AI Brain With $20M Investment

Oil and gas AI took a significant step forward on July 15 when Applied Computing raised $20 million in a Series A round to build what the company describes as a foundation model for the entire industrial plant. The round was led by KBR, the Houston-based engineering and government services firm, with participation from Databricks, the data and AI platform.

The raise positions Applied Computing against decades of fragmented, rule-based software that has governed refinery and petrochemical operations with little change since the 1970s.

The Software That Has Run Oil and Gas Refineries for Half a Century

Oil and gas AI sits atop one of the most entrenched software stacks in any industry. Industrial plants are managed by distributed control systems, hardware-and-software combinations that translate operator commands into valve positions, pump speeds, and heat exchanger settings.

Above that sits advanced process control, the software layer that handles real-time adjustments by solving thousands of simultaneous equations every few seconds. Both layers were designed for determinism: given a known state, produce a known output.

The problem is that plants are never in a perfectly known state.

Feedstock composition shifts by the hour. Equipment ages and drifts from its calibrated baseline.

Seasonal temperature changes alter thermodynamic behavior across the entire unit. Rule-based systems handle these variations badly, forcing human operators to compensate manually or accept suboptimal throughput and energy consumption.

What Applied Computing Actually Built

Applied Computing’s approach draws from physics-informed machine learning.

Rather than training a model purely on historical sensor data, the company embeds the underlying thermodynamic and fluid-dynamic equations of a plant directly into the model’s architecture. This matters because a purely data-driven model can learn spurious correlations, predicting outcomes that are thermodynamically impossible but statistically plausible within a training set.

A physics-informed model cannot violate conservation laws.

The company calls its output a foundation model for the plant. In large-language-model terms, a foundation model is a large, general-purpose neural network pre-trained on broad data that can then be fine-tuned for specific tasks.

Applied Computing is attempting the same architecture shift for industrial process data. Rather than building a bespoke model for each refinery unit, the company trains a single large model on data from many plants and then adapts it to a new operator’s site with comparatively little additional training.

The TechCrunch report on the raise describes the company as an Imperial College spinout, meaning its core intellectual property originated in academic research on physics-based simulation at Imperial College London.

Tech Funding News notes that the Series A closed at the equivalent of 17.4 million euros, given that Applied Computing is London-headquartered.

Why KBR and Databricks Wrote the Checks

The investor composition is as informative as the raise size. KBR designs and builds refineries, petrochemical plants, and LNG facilities.

It sells engineering services to the same Oil and Gas operators Applied Computing is targeting. A strategic investment from KBR gives Applied Computing immediate credibility with plant operators who treat unfamiliar software vendors with deep skepticism, since a process control failure in a refinery does not produce a bad user experience but can produce a catastrophic incident.

Databricks’ participation connects Applied Computing to a software platform that large enterprises already use for data engineering and model training.

Databricks has spent several years pushing into industrial and scientific AI workloads, and a stake in a physics-informed energy AI company fits that direction.

The Oil and Gas sector represents an unusually large addressable market for industrial AI. The energy consulting firm Wood Mackenzie estimated in 2024 that optimizing process operations across global refineries and petrochemical plants could recover between $50 billion and $70 billion in annual value through energy savings, yield improvement, and reduced unplanned downtime.

Applied Computing is pitching squarely at that figure.

From Imperial College Spinout to Industrial Challenger

Applied Computing joins a cohort of AI startups targeting heavy industry with foundation-model approaches. Sight Machine, Aspen Technology, and Cognite have each built AI layers for industrial operations, though none has framed its product explicitly as a foundation model in the generative AI sense. The foundation-model framing matters commercially because it implies rapid deployment across new sites without the months-long custom modeling engagements that have historically made Oil and Gas AI expensive to scale.

The competitive pressure on incumbents is real.

Aspen Technology, the dominant advanced process control vendor, has been part of Emerson Electric since 2022 and carries a product roadmap anchored to its existing customer base. A startup able to fine-tune a pre-trained foundation model to a new plant in weeks rather than months could undercut that installed base on both speed and cost, even if the incumbent’s models carry decades of operational validation.

The Harder Problem Applied Computing Still Has to Solve in Oil and Gas

Oil and Gas AI faces a trust barrier that consumer or enterprise software does not.

A refinery operator who deploys a model recommendation and experiences an unplanned shutdown faces regulatory scrutiny, potential liability, and production losses that dwarf the cost of the AI subscription. Operators want to see a model run in shadow mode, producing recommendations alongside their existing system without authority to act, for months before they consider giving it control.

That means Applied Computing’s commercialization timeline is longer than a typical SaaS startup’s.

The $20 million Series A buys the company time to accumulate shadow-mode validation data across a handful of reference sites, which it can then use to persuade the next tier of Oil and Gas operators. KBR’s involvement may accelerate that process by inserting the technology into new plant construction contracts, where a greenfield site has no incumbent system to displace.

The raise also arrives as energy majors face pressure to reduce the carbon intensity of their operations without shutting down facilities.

A model that recovers 2 to 3 percentage points of energy efficiency across a refinery produces measurable emissions reduction without capital expenditure on new equipment, which is exactly the kind of outcome that satisfies ESG commitments while preserving throughput margins.