Engineering and industrial companies are under pressure to design more complex systems faster while meeting strict efficiency and sustainability targets. Traditional simulation tools are slow, often need supercomputers, and don’t work well with modern AI workflows. This means design teams can spend hours or even days waiting for a single high-quality simulation, slowing testing and iteration.
BeyondMath, a London-based deeptech team, is tackling the problem head-on with a foundational AI model trained directly on first-principles physics. Their generative physics platform can deliver engineering-grade simulations in minutes, up to 1,000 times faster than the old-school supercomputers.
One standout is their $19M, three-year STRATA project with Honeywell. BeyondMath’s tech lets teams run thousands of design iterations for complex aircraft parts in seconds, paving the way for lighter, more efficient components that could save billions in fuel and slash aviation emissions.
Today, BeyondMath just closed an $18.5M seed round (including a $10M extension), led by Cambridge Innovation Capital and backed by UP.Partners, Insight Partners, and InMotion Ventures.
In a conversation with TFN, Alan Patterson, CEO and co-founder of BeyondMath, confirms, “To date, BeyondMath has raised $20.35 million in total funding. We are focused on deploying this capital to scale our technology and team; as a private company at this stage of growth, we are not disclosing our current valuation.”
Reduce design cycles from months to minutes
Founded in 2022 by AI experts Alan Patterson and Darren Garvey, BeyondMath was born out of pure frustration with slow, legacy simulation tools. Instead of relying on old simulation data, they built a foundational physics model that learns straight from the laws of physics.
Patterson shares, “The primary motivation for starting BeyondMath was a critical market need within engineering across all sectors. For decades, design engineering has been held back by the computational bottleneck of traditional simulation tools. These tools are prohibitively slow and expensive, often requiring multi-million dollar supercomputers and days of processing time for a single iteration.”
At the heart of BeyondMath’s platform is what they call the world’s largest foundational physics model: a generative AI trained on first-principles physics to simulate complex phenomena such as aerodynamics and thermal management.
The company aims to shrink design cycles from months to minutes and enable real-time iteration at scale, backed by high-profile partnerships with STRATA and Honeywell, and by record-breaking aerodynamic results.
Unlike traditional tools that solve each new scenario from scratch, BeyondMath’s model can quickly generate engineering-grade predictions after training, delivering significant speed improvements without sacrificing accuracy.
“Our model doesn’t need large amounts of simulation data generated by the customer to be useful; it already understands the inherent laws of physics. From an economic perspective, this means customers face drastically lower upfront costs to get started. We provide a 1,000x speed advantage that allows for instant simulation, enabling a level of iteration that traditional methods simply cannot match,” elaborates Patterson.
Key strengths and technical highlights include: a foundational physics model trained on core physics laws instead of just past simulation data, allowing it to work across areas like airflow, heat transfer, and fluid dynamics; speeds up simulations by up to 1,000 times, as shown on the DrivAerML dataset; engineering-grade accuracy designed for safety-critical fields like aerospace and automotive; and a wide range of applications including aerospace (like the STRATA project with Honeywell), automotive (including work with an F1 team), electronics, data centre design, and semiconductor manufacturing.
In automotive, BeyondMath’s platform lets teams test thousands of design options for vehicle aerodynamics and thermal management almost in real time, instead of being limited by the cost and delays of traditional CFD methods.
BeyondMath is carving out its own space alongside the new wave of physics-AI and simulation-AI startups, as well as the big CAE vendors using machine learning. While names like Ansys and Siemens still lead the traditional pack, and others are tinkering with ML for CFD and FEA, BeyondMath stands out by betting on one big foundational physics model.
What about diversity?
On diversity, Patterson notes, “BeyondMath is proud to be a truly global team. We currently have 10 different nationalities represented within our company, bringing together a wealth of perspectives from across the world to our headquarters in the UK. We prioritise building a multidisciplinary team that blends deep expertise in theoretical physics, machine learning, and industrial engineering. We believe that solving the world’s most complex physical problems requires this level of international diversity and academic variety.”
So, what’s next for BeyondMath?
After raising $18.5 million in its Seed round, BeyondMath plans to grow commercial use of its generative physics platform and expand research. The company aims to double its team this year by hiring more people in research, engineering, and customer support.
Geographically, BeyondMath will strengthen its presence at home while growing its customer base across Europe, the US, and Japan. It will focus on sectors where complex physics and strict performance demands meet: aerospace, automotive, electronics, data centres, and semiconductors.
On the product side, the team will keep expanding the range and detail of phenomena their foundational model can simulate and work on tighter integration with customers’ existing design, simulation, and PLM systems.
Patterson concludes, “Our objective is to become the foundational engine for all physics-based design and engineering globally”
