JuliaHub Raises $65 Million and Launches Dyad 3.0 to Bring Agentic AI to Industrial Digital Twins

# JuliaHub Raises $65 Million and Launches Dyad 3.0 to Bring Agentic AI to Industrial Digital Twins

A startup built on the Julia programming language just secured serious venture capital to prove that AI agents can do more than write code -- they can design jet engines, predict pump failures, and compress months of industrial R&D into minutes.

JuliaHub, the company commercializing the open-source Julia programming language created at MIT, announced a \$65 million Series B funding round on April 30 alongside the launch of Dyad 3.0, a platform the company calls the world's first agentic AI system purpose-built for hardware engineering. The round was led by Dorilton Capital, with participation from General Catalyst, AE Ventures, and former Snowflake CEO Bob Muglia, who also joined as a strategic technology investor.

The funding marks a significant bet on what JuliaHub and its backers are calling "Physical AI" -- the application of autonomous AI agents not to software tasks like writing emails or generating code, but to the deeply technical work of designing, simulating, and stress-testing physical machines and infrastructure. Think heat pumps, satellites, semiconductor fabrication lines, and water distribution networks.

Spec In, Design Out

At the heart of the announcement is Dyad 3.0, a cloud-based platform that deploys swarms of AI agents to automate the full cycle of industrial systems design. Unlike general-purpose AI tools, Dyad's modeling language was built from the ground up to be legible to AI agents while remaining grounded in the laws of physics. Its agents can reason about fluid dynamics, thermodynamics, gravitational forces, and control logic -- producing models that are not just plausible but physically valid.

"It's not about helping engineers complete one small task at a time," said Viral Shah, CEO of JuliaHub and co-creator of the Julia language. "It's agentic engineering at scale, where teams can feed a full specification to Dyad and have it design the complete system. Spec in. Design out."

That pitch -- automating not just individual calculations but entire design workflows -- positions JuliaHub squarely against incumbents like MathWorks, whose MATLAB and Simulink products have dominated engineering simulation for decades. JuliaHub argues that those legacy tools were never designed for AI-native workflows, and that Julia's computational speed and composability give Dyad a structural advantage in an era where AI agents need to run thousands of simulations in parallel.

Real-World Validation

JuliaHub is not pitching vaporware. Several Fortune 100 companies are already using Dyad and the broader Julia ecosystem across aerospace, automotive, HVAC, government, and utilities sectors.

One concrete case study stands out: working with water management firm Binnies and Williams Grand Prix Technologies, JuliaHub developed a Scientific Machine Learning-powered digital twin that uses just four sensor inputs to predict pump faults in water distribution systems with over 90 percent accuracy. In another demonstration, Dyad automated the entire design process for a system of model-predictive controllers used in chemical manufacturing plants -- a task that would typically consume months of painstaking manual work by specialist engineers.

These are the kinds of results that attract investor attention in a market where digital twin technology is surging. The global digital twin market is projected to exceed \$110 billion by 2030, and industrial companies are under relentless pressure to shorten product development cycles while meeting tighter safety and sustainability standards.

The Investor Thesis

Daniel Freeman, who led the Series B for Dorilton Capital, framed the investment in terms of convergence. "Systems modeling is strategically important where physics, control logic, and AI converge," Freeman said. "JuliaHub has built something extraordinary with Dyad. JuliaHub has the potential to become one of the defining companies in Physical AI."

The involvement of Bob Muglia -- who spent two decades as a senior executive at Microsoft before serving as CEO of Snowflake during its hypergrowth phase -- adds a different dimension to the story. Muglia told Axios that JuliaHub aims to harness agentic coding tools while preserving the mathematical rigor that physical engineering demands. "The physical world needs mathematical certainty attached to what it's doing," Muglia said, drawing a sharp line between the probabilistic outputs of large language models and the deterministic requirements of engineering simulation.

What to Watch

JuliaHub's bet is that the same agentic AI revolution transforming software development will inevitably reach hardware engineering -- and that whoever builds the right foundation first will own the category. With Dyad 3.0, the company is not just offering a better simulation tool; it is proposing a fundamentally different workflow where AI agents handle the iterative grunt work of design exploration while human engineers focus on specification, judgment, and oversight.

The challenge will be execution at scale. Enterprise sales cycles in aerospace and automotive are long, regulatory requirements are demanding, and incumbent tools have decades of trust built into engineering workflows. But with \$65 million in fresh capital, Fortune 100 customers already on board, and a programming language with over one million developers worldwide, JuliaHub has assembled the pieces for a serious run at one of AI's most consequential and least hyped frontiers.

If Dyad delivers on its promise, the implications extend well beyond JuliaHub's bottom line. Compressing industrial design cycles from months to minutes would reshape how everything from consumer appliances to aerospace systems gets built -- and it would prove that agentic AI's most transformative applications may not be in the digital world at all, but in the physical one.