Corporate strategic money is moving into industrial AI: CATL anchored DeepCtrls' latest raise after installing the startup's control software at more than ten of its plants, and Aramco Ventures came in alongside. It is the third round in about two months for a company whose physics-based engine optimizes cooling and power in data centers.
A battery giant and an oil major have put money into the same industrial software startup, and the sequence behind the checks explains why. CATL led DeepCtrls' Series B+ round, worth hundreds of millions of yuan, after running the company's control software across more than 10 of its own facilities. Aramco Ventures took part as a strategic backer. The round closed in September 2026 and is the third raise DeepCtrls has completed in roughly two months.
The sequence matters more than the size. CATL first ran DeepCtrls' PhyAI engine as a pilot at a single air-conditioning facility, then expanded it across more than ten company sites before writing the lead check. "The commercial relationship came before the investment," the company said in its funding announcement, describing a model in which industrial customers become strategic backers once a system proves itself inside a live plant.
DeepCtrls builds what it calls a control layer for Physical AI — software that combines physics-based models of equipment with machine-learning algorithms, then issues automated operating decisions rather than recommendations. Its PhyAI engine predicts how cooling systems, industrial equipment and power infrastructure will behave under changing loads, then adjusts setpoints within safe operating limits. Traditional industrial control relies on predefined rules that switch equipment on or off; DeepCtrls replaces that with continuous optimization across temperature, workload and electricity consumption.
The round drew Taiping Innovation Investment, GF Xinde Investment Management and Fosun Capital as new participants, while existing backers Source Code Capital and Forebright Capital increased their commitments. DeepCtrls did not disclose the exact deal size, the valuation, or the ownership stakes involved. The company previously closed a Series B in July 2026 worth hundreds of millions of yuan, led by a strategic investment from solar manufacturer JinkoSolar, with SDIC Innovation and China Merchants Bank International participating.
DeepCtrls' customer list is the clearest evidence of where the money is going. The company says it serves hundreds of enterprise customers, naming Tencent, ByteDance, Nvidia, TSMC, LG and Thailand's PTT among them, with deployments across Asia, the Middle East, Europe and North America. Those names span two distinct markets: factories and data centers.
The data center side is the one growing fastest. AI training and inference clusters run continuously and must balance temperature, computing workload, electricity demand and cooling performance in real time — a control problem that scales with rack density. DeepCtrls is developing liquid cooling control and software that coordinates computing workloads against available power, targeting a market where electricity and cooling represent a rising share of operating expenses. For an operator running thousands of racks, a 1 percentage point gain in power usage effectiveness translates directly into lower opex, which is why the control layer attracts capital even when the underlying hardware does not change.
Aramco Ventures' participation points in a different direction. The fund invests globally in AI, analytics, automation, robotics and energy efficiency where the technology is strategically relevant to energy and industrial operations, and its involvement connects DeepCtrls to Middle East energy, industrial and infrastructure markets. For Aramco, the appeal is the same control problem at a larger scale: refineries, gas processing and power generation all run on equipment that consumes energy continuously and responds poorly to static setpoints.
The competitive field is crowded but fragmented. Emerson, Honeywell and Siemens dominate traditional industrial control and building management, while AI-native entrants including DeepCtrls, BrainBox AI and Phaidra target the optimization layer above those systems. DeepCtrls' differentiator is the physics constraint: because its models encode equipment behavior and operating rules, the company argues they transfer across facilities rather than requiring a bespoke build for every site — the difference between a project business and a product business.
The investor mix here is the story. Venture funds supplied the earlier rounds; the current one is anchored by two industrial giants that can deploy the technology inside their own operations. That structure gives DeepCtrls something a pure financial round cannot — reference sites, operating data and a distribution channel into energy and manufacturing, the two sectors where AI adoption has lagged software and services.
For CATL, the investment extends a diversification push beyond battery manufacturing into AI-driven automation, and gives it a stake in software it already uses to cut energy costs at its own facilities. For Aramco Ventures, the position adds a physical-world AI asset to a portfolio that already spans analytics and automation. Neither company disclosed the size of its commitment, and neither has stated whether the technology will be deployed across additional sites beyond those already running it.
The near-term test is whether DeepCtrls converts its pilot base into standardized products. The company said its July round would fund Physical AI algorithm development, standardized product expansion and international growth across Southeast Asia, the Middle East, Europe and the Americas — the same priorities the Series B+ extends. Three rounds in two months gives it capital to move on that plan; whether hundreds of enterprise customers become thousands depends on how quickly the control layer can be installed without custom engineering at each site.
This article is for informational purposes only and does not constitute investment advice.
