Applied Computing has just closed a $20 million funding round led by engineering giant KBR, with participation from Databricks Ventures. This is more than a financial transaction: it signals how artificial intelligence is delving deeper into industrial sectors, moving away from generic demos toward models purpose-built for real operating environments.

The London-based startup, with offices in Bengaluru and Houston, develops Orbital, a platform that combines physics-informed AI with models for chemical engineering, time-series forecasting and language. The stated goal is to improve efficiency, cut emissions and make upstream, downstream and petrochemical operations more reliable. The underlying principle is simple yet disruptive: a generic Large Language Model doesn’t understand the fluid dynamics of a pipeline or the reactions inside a catalytic cracker. What’s needed are models trained on specific domains, capable of blending physical laws with operational data.

The deal with KBR is not merely an equity investment. The two companies have signed a multi-year agreement to co-develop exclusive AI products. This locks in an existing commercial relationship and erects a significant barrier to entry. KBR, which works with governments and large enterprises across energy and defense, offers direct access to critical infrastructure where reliability and safety matter more than raw compute power.

For anyone tracking the evolution of foundation models, the move confirms a structural trend: value no longer resides in the generalist model, but in the ability to adapt it to a vertical domain using proprietary data. BloombergGPT for finance paved the way; here it applies to a sector where errors can mean environmental disasters or million-dollar plant shutdowns. It is no surprise that Applied Computing chose to bake physics into its AI, reducing hallucination and improving forecast robustness.

For those evaluating deployment scenarios, the most interesting aspect revolves around data control. Energy utilities handle sensitive operational information and are often subject to strict regulations. Although Orbital is currently a largely cloud-based platform, the natural direction for the sector is hybrid or on-premise, where models run close to the data to ensure low latency and sovereignty. It is no coincidence that the company is strengthening its presence in Houston, the heart of the American oil & gas industry, where technology decisions also involve Total Cost of Ownership and infrastructure resilience.

The entry of Databricks Ventures adds another piece: the startup will be able to leverage the data lakehouse ecosystem to manage the industrial data volumes that feed the models. For energy operators, this means unifying process, maintenance and IoT data into a single pipeline, accelerating time to value.

Ultimately, Applied Computing’s round shows that the market is rewarding those who build specialized AI, not those chasing yet another variant of a general-purpose LLM. And for an industry like energy, where the balance between innovation, safety and data control is delicate, the vertical AI path looks less like a gamble and more like a competitive necessity.