Aramco and Yokogawa take industrial AI into process control

Aramco named Yokogawa as a collaborator on industrial process control within the set of technology agreements and memoranda it announced at LEAP 2026 in Riyadh, placing a process automation vendor alongside the compute and model partners in the same programme. Of everything in that list, this is the entry that touches the machinery most directly, and it is worth being precise about what process control is. It is the layer of systems that holds a plant at its setpoints: the controllers that adjust a valve when a pressure drifts, the distributed control system that operators watch from the control room, and the separate safety instrumented systems whose only job is to shut a unit down when a reading crosses a line. It is not analytics. It is the software that moves things. Yokogawa is one of a small group of vendors that supplies this equipment to refineries, chemical plants and gas processing facilities worldwide, and it has been pushing for several years at the question of how much of a plant's operation can be run by software rather than by a human operator following a procedure. That question is the interesting one here, because it is where industrial AI runs into a wall the rest of the field does not have. The wall is certification. A model that predicts a bearing failure can be wrong and the cost is a wasted inspection. A model that adjusts a control loop on a live hydrocarbon process can be wrong in ways that appear in an accident report. Process control systems are engineered, tested and certified to standards that assume deterministic behaviour: given this input, the system does exactly this, every time, and an auditor can prove it. Machine learning is not deterministic in that sense, which is why the honest version of AI in process control usually sits beside the control system rather than inside it — recommending setpoints for an operator to accept, or optimising within limits that a conventional interlock still enforces. That is the distinction to watch as this collaboration takes shape, and neither company has yet said which side of it the work falls on. Advisory optimisation and closed-loop control are different products with different approval paths, and the gap between them is measured in years of evidence. The collaboration fits a pattern across Aramco's week. The company set out hardware collaborations with NVIDIA, Groq and Qualcomm in the same announcement set, has a separate partnership with Avathon on industrial AI , and signed a memorandum with Microsoft covering industrial AI and sovereign cloud . Industrial AI has been a running theme at the show more broadly, with IFS and Microsoft arguing for moving it from experimentation to execution . What has not been published is almost everything that would let anyone size this. No value, no timeline, no named facility, no statement of whether the work covers new installations or an existing Yokogawa estate inside Aramco, and no indication of whether it is an MoU or a firmer commitment. The published description is a single line in a partner list. The test is narrow and physical. Not how many plants are named in a press release, but whether a single Aramco unit ends up running an optimisation loop that a process safety engineer has signed off on, and whether either company says so afterwards. TechScoop is reporting from LEAP 2026 in Riyadh.