Thomas Kiessling, named CEO of PassiveLogic on July 8, 2026, is putting the Salt Lake City company's commercial story behind a new product launch: Level 3 Autonomy, a system PassiveLogic says can run buildings and industrial infrastructure through a physics-based world model rather than fixed control scripts.
The timing matters because Kiessling is arriving with exactly the background PassiveLogic needs to sell that thesis beyond a technical audience. PassiveLogic says he was previously chief technology officer of Siemens Smart Infrastructure, helped shape Siemens' industrial AI strategy and led innovation programs including Building X, Siemens' cloud-based digital building platform. Earlier, according to PassiveLogic, he co-founded AMPLY Power, an EV fleet charging-as-a-service company acquired by BP.
PassiveLogic is calling Level 3 Autonomy the "world's first system for generalized autonomous control of buildings and infrastructural robotics." That is the company's claim. The concrete product pitch is easier to assess: PassiveLogic says L3 uses its physics-based World Model to coordinate building systems, anticipate demand and adjust controls in advance to balance comfort, energy use, operations and risk. PassiveLogic says it is targeting Level 4 Autonomy later in 2026, enabling learning at the edge and extending prediction horizons.
The launch turns PassiveLogic's long-running framing into a sharper commercial wedge. On its homepage, PassiveLogic describes buildings as "the largest operating system on Earth" and says its stack spans a world model, reasoning agents and autonomous control systems. Its Autonomy Platform pairs with Hive, the autonomous edge layer that operationalizes the world model in the building. PassiveLogic also highlights tools such as Lens for capturing a site's equipment and spaces into that model.
The bet: buildings as robots
Kiessling's first job is to make a category sound operational, not abstract. PassiveLogic is not selling another dashboard for facility managers. It is arguing that a building, a data center or an industrial site should be treated as a robot: full of sensors, actuators, dependencies, safety constraints and decisions that play out over time.
That framing is useful because conventional building automation has been built around programmed sequences. PassiveLogic says most building automation systems remain at Level 0, where control is hard-coded and the reasons behind actions are opaque. In the company's version of Level 3, control is generated from a model that knows what equipment is, why it exists, how it behaves, how actions create outcomes and how people and assets interact with the space.
That is also why PassiveLogic stresses traceability. In a building, an AI control decision cannot be treated like a chatbot answer. If a system is going to adjust ventilation, thermal comfort, equipment schedules or energy consumption across a live property, customers will need to understand intent, causality and safety boundaries. PassiveLogic says its approach is a "white box" because the control path is dynamic and trainable while remaining determinative and traceable.
NVIDIA is named, but the architecture is not
PassiveLogic says Level 3 is powered by NVIDIA (@nvidia). The announcement does not specify which NVIDIA chips, cloud services, software libraries or reference architecture are involved. That omission matters because "powered by NVIDIA" can mean several different things in an AI infrastructure story, from investor alignment to GPU usage to a deeper technical partnership.
The investor list gives NVIDIA another point of contact. PassiveLogic says its backers include NVentures, Prologis Ventures, Addition, Johnson Controls, noa, Brookfield Growth, G2 Venture Partners and Keyframe Capital. PassiveLogic did not announce a new financing round with the product launch, and the release does not disclose current revenue, valuation, total capital raised, customer count or deployment count.
That restraint is important. The strongest version of PassiveLogic's story is a product and distribution story, not a funding story. The customers it wants are commercial building owners, industrial operators, data center managers and infrastructure teams that do not want one-off engineering projects each time they automate a facility. PassiveLogic's pitch is that a generalized world model can reduce the custom work that has historically kept advanced building controls confined to a smaller set of high-end assets.
What PassiveLogic still has to prove
The most useful missing details are the deployment details. PassiveLogic says it has moved autonomy "into deployment in live environments," quoting Kiessling, but the July 8 release does not name a customer, site, portfolio, data center, industrial facility or pilot partner for Level 3 Autonomy. It also does not say whether Level 3 is generally available, in limited release or being rolled out through selected deployments.
The product logic is clear: if a building's model captures equipment, physical constraints and human usage, the controller can make better decisions than a fixed sequence written once and tuned manually. The business test is whether PassiveLogic can make that setup fast enough and reliable enough for operators who are paid to avoid surprises. PassiveLogic is trying to replace custom engineering with a repeatable autonomy stack. That requires trust from facilities teams, owners, insurers and procurement buyers who treat building operations as a risk function.
Kiessling gives PassiveLogic a more credible route into that market. A former Siemens Smart Infrastructure CTO can speak the language of enterprise infrastructure buyers, grid flexibility, controls, energy savings and building operations. His AMPLY background also gives him experience selling physical infrastructure software and services into a market where deployment economics matter as much as software capability.
PassiveLogic's Level 3 launch is the clearest statement yet of how the company wants to compete: own the model of the building, own the control layer and make autonomy a measurable operating function. The next proof point will be less about the phrase "Physical AI" and more about whether PassiveLogic can point to named portfolios where energy use, comfort and maintenance outcomes improved without constant manual tuning.