Orbital Industries and the Hardest Bet in Modern Hardware
There is a data point in this story that deserves a pause before we talk about funding rounds or language models: according to the CEO of Orbital Industries, developing a new cooling fluid for data centers would normally take ten years and one hundred million dollars. The company says it accomplished this in months, at a fraction of that cost. If that claim withstands validation by major chip manufacturers, we are not looking at a laboratory achievement. We are looking at a change in the speed at which hardware can exist.
Orbital Industries has just closed a Series B round of 50 million dollars led by venture capital firm Plural, with participation from NVentures (Nvidia's investment arm), Radical Ventures, Compound, and Fly Ventures. The company, which has offices in London and San Francisco and a team of around fifty people, was founded in 2022 under the name Orbital Materials. The name change is not cosmetic: it reflects an explicit bet on leaving the territory of applied science and entering that of industrial hardware at scale.
---
The Model the Others Did Not Choose
Context matters here. Over the past two years, a wave of startups has bet on using artificial intelligence to discover new materials. CuspAI raised one hundred million dollars in a Series A. Periodic Labs captured three hundred million in a seed round. The sector's thesis is relatively uniform: use machine learning models to identify novel compounds and then license that intellectual property to established chemical companies such as BASF or PPG.
Orbital Industries decided not to do that.
Jonathan Godwin, co-founder and CEO of the company — who spent five years at Google DeepMind working on artificial intelligence for science and advanced materials — articulated it with precision: "We are what is known as vertically integrated. We don't sell the software. We have hardware, manufacturing, and advanced materials teams, laboratories and things like that, and we use that software internally to develop new advanced materials and hardware devices, and we sell those devices."
That statement, delivered with apparent naturalness, describes an organizational decision of enormous weight. Godwin is not building a software company disguised as science. He is building a company that manufactures physical things, with all the risks that entails: supply chains, manufacturing at scale, qualification processes with industrial clients that can take years, environmental regulations, and capital-intensive costs.
The licensing model they chose to avoid has a very concrete virtue: it transfers the complexity of manufacturing to whoever already knows how to do it. The model Orbital chose instead concentrates that complexity upon itself. That can be a strength — it captures more value per unit, builds higher barriers to entry — or it can become the Achilles' heel if execution fractures at any point in the chain.
---
The Problem Orbital Is Attacking and Why It Matters Now
To understand why this bet makes sense at this precise moment in the industry, one must look at the problem the company is trying to solve.
Modern data centers, especially those designed for artificial intelligence workloads with high-density GPU racks, generate heat on a scale that conventional cooling systems cannot manage efficiently. Godwin described it in deliberately everyday terms: it is like compressing the energy of a supermarket inside a filing cabinet. The dielectric fluids historically used for liquid cooling contain PFAS — the so-called "forever chemicals" — which face growing regulatory restrictions in the United States and Europe due to their environmental and health impacts.
The convergence of these two problems — extreme thermal density and regulatory pressure on existing refrigerants — creates a window of genuine demand. Orbital used its artificial intelligence model, called Orb, to screen hundreds of thousands of molecular candidates and synthesize a family of cooling fluids that dispense with PFAS. The company says that Orb can simulate the quantum mechanical behavior of 100,000 atoms on a single GPU, at a speed approximately ten times faster than alternative models from Meta and Microsoft.
The cooling fluid, along with a cooling system that Orbital is also building, is designed to be deployed alongside the next generation of GPUs in 2027. If that timeline is met, it would be the first molecule designed by artificial intelligence to reach the commercial market in any industry. Godwin points out that in drug discovery — where startups have spent years using AI to identify molecular candidates — no drug discovered by artificial intelligence has completed clinical trials and reached the market. The difference is that industrial materials do not go through that clinical regulation, which significantly shortens the path to market.
The company's second product is a modular data center system, manufactured off-site and delivered as ready-to-deploy units, which according to Orbital can bring high-density computing capacity online in six months, compared to the up to three years required by conventional construction. Both products are commercialized under the Orbital IT brand.
---
What the Investor Architecture Reveals
When NVentures, Nvidia's venture capital arm, decides to participate in a round, it does not do so simply for financial return. It does so because it has a strategic interest in ensuring that the ecosystem around its chips functions properly. A PFAS-free cooling fluid that can be deployed alongside the next generation of GPUs is exactly the kind of infrastructure piece that Nvidia needs someone to solve. NVentures' participation does not guarantee a commercial contract, but it establishes a proximity that can accelerate the technical qualification processes with the world's largest chip manufacturer for artificial intelligence.
Ian Hogarth, partner at Plural and the individual leading the investment for that fund, framed the argument directly: the progress of artificial intelligence is being constrained by energy, heat, and infrastructure. Orbital attacks those constraints from within. Plural also holds a position in Proxima Fusion, the German fusion energy startup that has raised approximately two hundred million dollars in public and private capital. It is no coincidence that the same firm betting on nuclear fusion is also betting on a company that wants to redesign the materials with which critical physical infrastructure is built. There is a coherent portfolio thesis there, even if its realization horizon is measured in decades.
Godwin was explicit about his terminal ambition: to build the largest industrial conglomerate in Europe. He compared Orbital's position to that of the chemical giants that emerged a century ago — BASF, PPG and similar companies — and argued that those companies exist because they built deep competitive moats based on accumulated knowledge, manufacturing scale, and vertical integration. According to him, the only way to erode those moats is with sufficiently radical technological innovation. Artificial intelligence, in his reading, is that innovation.
The argument has logic, but it also has a trap worth naming. The industrial conglomerates of the twentieth century took decades to consolidate, with access to cheap capital over long periods and in very different regulatory and competitive environments. Orbital has fifty people, fifty million fresh dollars, and a product timeline that extends to 2027. The distance between the declared ambition and the current capability is not a communications defect: it is the most concrete operational risk the company faces.
---
When Laboratory Speed Collides with Industrial Slowness
There is a structural tension in Orbital's proposition that no funding round resolves on its own: the speed at which artificial intelligence can discover and synthesize new materials does not automatically transfer to the speed at which the industry qualifies, adopts, and scales the use of those materials.
Major chip manufacturers have qualification processes that can take between one and three years, even for products that technically work from day one. Hyperspecialized data centers have established suppliers, long-term contracts, and a risk tolerance that does not move at the pace of a startup. The company says it has already located a contract manufacturer to scale production of the cooling fluid, and that it is in the process of qualification with "leading chip providers." None of those providers are publicly identified, and the complexity of those processes cannot be compressed by technology alone.
That does not invalidate the bet. But it does reveal where the hardest variable to control lies: not in the laboratory, but in the organizational friction of its future customers. Godwin has a background in computational science and knows how to build a model that simulates one hundred thousand atoms on a single GPU. What will determine whether Orbital reaches 2027 with a product on the market is its capacity to navigate the decision-making processes of organizations that do not operate under the same assumptions of speed as a fifty-person startup.
The vertical integration that Orbital chose gives it control over its value chain. But it also gives it total responsibility over every point where that chain can fail. That demands a type of organizational maturity that cannot be built with AI models or venture capital: it is built through difficult conversations between science, manufacturing, and industrial sales teams that have radically different time horizons, vocabularies, and criteria for success.
If that integration is managed well, Orbital has a position that its larger competitors in the materials space would find difficult to imitate quickly. If it is managed poorly, the fifty million dollars will be spent on internal coordination before the first fluid reaches a production rack.
That is what makes this bet genuinely difficult, and genuinely interesting.
