Luminary Cloud secures $72M to take simulations from days to seconds with Physics AI

Updated 09:00 EDT / September 15, 2025

Luminary Cloud raises $72M to speed up AI-driven physical product design

Luminary Cloud has secured a $72 million Series B to push "Physics AI" into mainstream product development. The company plans to expand research and scale go-to-market for manufacturers that need faster, more flexible simulation and design cycles.

The startup arrived 18 months ago with a claim: run engineering simulations in the cloud up to 100x faster than desktop tools. Computer-aided engineering has been slow to move off workstations due to performance, control, and security concerns. Luminary's bet is that speed, scale, and collaboration will tip the balance.

Why product teams should care

Physics AI uses physics-informed models to predict how real products perform-cars, aircraft, electronics-in near real-time. Instead of waiting hours or days, engineers can iterate designs almost instantly, explore more options, and feed results directly into control systems for faster testing loops.

If the accuracy holds, this compresses design cycles and broadens what you can test within the same budget. More design alternatives, tighter feedback with manufacturing, and earlier risk discovery become standard practice, not stretch goals.

What Luminary is shipping

• Shift Models: Pre-trained, industry-specific models. Shift-SUV (with Honda) for automotive and SHIFT-Wing (with Otto Aviation) for aerospace.
• Model Context Protocol Server: Lets models call external functions to accelerate development and connect to tools and data sources.
• AI-assisted notebook: Helps engineers prototype workflows, run studies, and process results without heavy scripting.

The strategy move: a production line for simulation models

Luminary is positioning its platform as a repeatable pipeline to build, validate, and deploy simulation models. As timelines compress, the ability to run accurate simulations in seconds-not days-shifts from advantage to baseline requirement.

Action plan for product development leaders

• Start with a pilot part or subsystem: Pick a known bottleneck (thermal, aero, crash, NVH). Define success metrics: accuracy vs. high-fidelity baseline, time-to-result, and design alternatives per week.
• Build a validation harness: Compare Physics AI outputs against trusted solvers and lab data. Track error bounds by scenario and set gates for deployment into control loops.
• Map the integration path: Identify where this slots into your CAD/PLM/CAE stack and test handoffs to test benches, HIL rigs, or digital twins.
• Quantify ROI: Measure engineer-hours saved, compute cost per study, and time shaved from design freezes and DV/PV milestones.
• Plan for governance: Set model versioning, data lineage, and access controls. Clarify IP boundaries with suppliers and tool vendors.

What to watch next

Accuracy on out-of-distribution scenarios, model generalization across platforms, and how well these models update with new test data will determine real-world value. Funding was led by N47 with participation from Sutter Hill Ventures and NVentures, the venture arm of Nvidia.

For context on the approach behind Physics AI, see this overview of physics-informed machine learning from Nature Machine Intelligence. Learn more about NVentures here.

Skill up your org

If you're building AI fluency across engineering and product teams, explore role-based learning paths at Complete AI Training.