PhysicsX: AI Beats Humans in Jet Engine Component Design
British startup PhysicsX has made a significant breakthrough in PhysicsX AI jet engine design, demonstrating that generative artificial intelligence can outperform top human engineers when designing critical aerospace components.
The company, which describes itself as “the physics AI company for industrials,” recently raised $300 million at a $2.4 billion valuation. In a newly published blog post, PhysicsX researchers detailed how they adapted 3D AI foundation models using physics-guided diffusion loops to bypass traditional CAD constraints, ultimately generating optimized engineering geometries automatically. The team demonstrated this approach by creating a jet engine bracket 18.5% lighter than the best existing human-designed alternatives.
Why Traditional CAD Design Has Limits
Most engineering workflows today rely on parametric CAD software, which effectively traps design possibilities within predefined, rigid parameters. PhysicsX engineers Daniel Owen-Lloyd and Sanmitra Ghosh noted in their blog that existing 3D foundation models, such as TRELLIS.2, typically produce visual assets rather than functional, physically viable engineering components.
To bridge that gap, PhysicsX applies Low-Rank Adaptation, a fine-tuning technique that adapts these foundation models using relatively small datasets, allowing the system to generate physically meaningful design variations even starting from something as simple as a single bolt. To make these designs usable for simulation, PhysicsX also developed a system that automatically repairs rough or imperfect 3D geometries, ensuring production-grade physics solvers can process them without crashing.
How the PhysicsX AI Jet Engine Design Process Works
Since standard physics solvers can’t directly guide AI generation on their own, PhysicsX trains differentiable surrogate models to act as fast, approximate physics judges throughout the design process. These surrogate models work alongside an SDEdit diffusion loop, which iteratively mutates candidate designs and retains only those variations that successfully pass physics-based validation checks.
To validate this entire framework, researchers tested it against the SimJEB titanium jet engine bracket benchmark, initializing their optimization process directly from the previous winning human-designed solution. The AI ultimately discovered a novel split-support geometry that met every required stress constraint while cutting overall weight by 18.5%, without relying on any traditional CAD design rules, demonstrating that generative 3D AI can genuinely expand engineering possibilities beyond conventional human design limits.
The Broader Mission Behind PhysicsX
Founded in 2020, PhysicsX describes its core mission as accelerating hardware innovation by fundamentally overhauling how industrial engineering and manufacturing processes work. The company is building a comprehensive software stack designed to bring deep physics AI capabilities across the entire engineering lifecycle, partnering with major organizations across aerospace and defense, automotive, semiconductors, materials, and energy sectors.
Jacomo Corbo, co-founder and CEO of PhysicsX, explained the broader significance of the breakthrough. “Almost every hard problem in the physical economy, better aircraft, better chips, better engines, better energy systems, comes down to how fast and how well engineers and machine operators can work through the underlying physics,” Corbo said. “For decades, that has been the binding constraint on hardware innovation.”
According to Corbo, physics AI directly addresses that longstanding bottleneck. “Physics AI removes it. We are giving engineers the ability to explore thousands of designs where they once managed a handful, in seconds rather than weeks, across the most demanding industries in the world,” he said, adding that the company’s recent funding will help extend this capability to more engineers while pushing toward increasingly capable large physics models.
Democratizing Advanced Engineering Capabilities
Robin Tuluie, founder and chairman of PhysicsX, emphasized how this technology could reshape access to advanced simulation capabilities more broadly. “High-fidelity physics simulation has always been powerful, but it has also been slow, costly and the preserve of a small group of specialists,” Tuluie said.
He argued that physics AI fundamentally changes that dynamic across multiple dimensions, making high-fidelity simulation significantly more efficient while also improving on traditional simulation results by incorporating real-world data directly into the company’s large physics models. “We believe in the democratisation of this technology to broad technical profiles across an industrial organisation, engineers, designers and operators who previously couldn’t run these analyses themselves,” Tuluie explained. “As that capability spreads, its utility compounds across the business. That’s the change we’re driving.”
As PhysicsX AI jet engine design research continues advancing, the company’s approach suggests generative AI could play an increasingly central role in solving complex physical engineering challenges that have traditionally required years of specialized human expertise to address.
Source: This article is based on reporting by Adam Pond for FDi Intelligence.