Gigaton Lands $26M to Replace Heavy Industry’s Control Stack

Gigaton, the London AI company that until recently traded as Carbon Re, has raised $26 million in Series A funding to push autonomous industrial control software into cement, steel, glass and chemical plants. The round was led by Plural, an early-stage European fund, and takes total money raised past $35 million. What sets the company apart from the wave of plant-optimisation startups is its target: it wants to remove the control software those plants already run on, not bolt another layer on top.

That is a bigger swing than the funding headline suggests. The systems Gigaton is going after belong to a handful of industrial giants who have held the control room for a generation, and the plants in question cannot afford an hour of unplanned downtime.

Plural Leads a $26 Million Round to Replace the Control Room

The Series A was led by Plural with participation from 2150 and Semapa Next, the venture arm tied to Portuguese industrial group Semapa. Every existing backer came along, which tells you the early investors still like the thesis after watching the product run in live plants.

The returning names trace back to the company’s roots as a Cambridge and University College London spin-out:

• Planet A Ventures, a Berlin climate fund that screens deals on measurable emissions impact
• Cambridge Enterprise Ventures, the university’s commercialisation arm
• UCL Technology Fund, managed by AlbionVC alongside UCL Business
• Clean Growth Fund, a UK venture fund focused on low-carbon technology

The capital is earmarked for deploying the platform more widely and building the next version of it with what the company calls a select group of partners. The pitch to those backers is straightforward economics layered on top of carbon: heavy industry runs on tight margins, energy is the single biggest variable cost, and a few percent off the fuel bill across a fleet of plants compounds fast. European deep-tech investors have been writing these autonomy-scaling cheques across hard sectors, from Destinus raising €50 million to scale autonomous defence drones to robotics rounds aimed at the factory floor. Gigaton’s twist is that it is selling into facilities where the software has to be trusted with the controls, not just the dashboard.

Gigaton’s Pitch Is to Replace the Control Stack Outright

Most AI tools sold to plants are advisory. They watch the process, spot inefficiencies, and recommend a setpoint change that a human operator then types in. Gigaton’s argument is that the advisory model caps the upside, because the gains only land when an operator is awake, paying attention and willing to act.

How the System Actually Runs the Plant

Gigaton says its software is designed to replace the underlying control stack rather than sit above it. The platform runs inside existing plant infrastructure and uses simulations and predictive models to test an operational decision before it is applied. It can autonomously adjust parameters like fuel mix, kiln speed and oxygen levels, then show the control-room team the reasoning behind each move. It retrains on live plant data, so it keeps adapting as conditions drift.

The technology took roughly five years to build, developed alongside plant operators rather than in a lab. That detail matters in a sector where engineers have seen plenty of clever software that fell over the first time a raw-material feed changed.

The CEO’s Case for Going Autonomous Now

Josh Vernon, the company’s chief executive, frames the existing software base as the bottleneck.

The underlying software infrastructure most plants run on today was never built to manage the complexity plants are forced to deal with today. We have built Gigaton to deliver real cost and carbon savings now while building the AI infrastructure these industries need in a fully autonomous future.

Vernon said that at the funding announcement. The phrase carrying the weight is fully autonomous future. Today the system works with operators in the loop; the company is selling the destination as a control room that mostly runs itself, with people supervising rather than steering. You can see the product framing and the named customers on the company’s self-learning control system site.

What the Cement Majors Are Seeing

The proof points so far come from cement, which makes sense. Cement production throws off roughly 8% of global carbon dioxide emissions, and a kiln burning at around 1,450 degrees Celsius is exactly the kind of energy-hungry, hard-to-tune process where small control gains turn into real money.

Gigaton lists cement heavyweights among its users, including Heidelberg Materials, Holcim, Cimsa, Secil, Cimpor and Mannok. Its published results from a Heidelberg Materials deployment give a sense of the scale of the gains:

• 4% cut in the fuel cost index
• 2.2% lower specific heat consumption
• 33% reduction in variability of C3S (the alite phase that drives cement strength)
• 2% drop in fuel-derived carbon emissions
• Eight weeks from start to deployment

None of those are headline-grabbing on their own. Stacked across a multi-plant operator running continuously, a few points off heat consumption and a steadier product is the difference between a profitable kiln and a marginal one. The eight-week deployment number is the quieter selling point, because plant managers fear long, disruptive integrations more than they doubt the savings.

A Control Market Worth $24 Billion, Built on Legacy Code

To understand why replacing the control stack is ambitious, look at who owns it. The distributed control system (DCS, the networked computers that run a plant’s machinery in real time) market is worth about $23.89 billion in 2026 and is forecast to reach $30.78 billion by 2031, according to research firm Mordor Intelligence’s distributed control system market analysis. Five suppliers dominate it.

These systems are sticky for reasons that have nothing to do with how good the AI is. They are certified, insured, wired into safety interlocks, and staffed by engineers who trained on that specific vendor’s interface. A plant that has run Honeywell or ABB for twenty years has a control philosophy built around it. Gigaton is asking that plant to hand the actuators to a five-year-old software company. That is the moat, and it is made of trust and liability, not features.

Why Replacing a Plant’s Brain Is the Hard Sell

Process plants run continuously, and in a continuous operation any downtime means lost money by the hour. Engineers who have managed control-system migrations describe them as some of the most carefully planned projects a plant ever undertakes, precisely because the cost of getting the cutover wrong is a stopped line or an off-spec batch.

That is the wall Gigaton has to climb. Advisory tools win deals partly because they are low-risk: leave the existing controller in place, and if the AI is wrong the operator just ignores it. Taking over the control layer removes that safety net. The upside is larger, because autonomous control captures gains around the clock, but the buyer is now betting production stability on the new system.

The company’s answer is the structure of the rollout. It runs inside existing infrastructure, simulates decisions before acting, and keeps operators able to see and override the reasoning. That hedges the risk without erasing it. The harder questions arrive at scale, when the platform moves from a handful of supervised cement kilns into steel, glass and chemicals, each with its own safety regime and its own entrenched vendor. Winning a Heidelberg pilot is one thing. Displacing an incumbent across a fleet of plants is a multi-year sales cycle measured in audits and insurance sign-offs.

From Carbon Re to Gigaton

The rebrand from Carbon Re to Gigaton, completed earlier in the year, signals the shift. The old name pointed at carbon reduction; a gigatonne is the unit climate scientists reach for when they talk about the scale of change the planet needs. The company started life in 2021 as a joint Cambridge and UCL spin-out aimed at cutting cement emissions, a story laid out in the Cambridge Enterprise case study on its decarbonisation work, and one its early backers still hold positions in through funds like the Clean Growth Fund’s clean-technology portfolio.

The new name reflects a wider ambition: not a cement-emissions tool, but the control software heavy industry runs on. Whether that ambition survives contact with the chemicals and steel plants it has not yet entered is the open question the $26 million is meant to answer. The cement results are real. The next phase asks plant operators to trust the autonomous version with sectors where the margin for error is thinner still.