Linda Lescuyer climbs the stairs to the roof of Data4's Marcoussis campus, south of Paris, and stops in front of a clear tank holding a thousand litres of green soup. It is not a science experiment. It is a balance sheet wearing a wetsuit. The liquid is Chlorella, a single-celled algae, and the heat keeping it alive comes straight from the servers humming one floor below.
This is the bio-circular data center: a facility that stops treating its waste heat as a problem to be vented and starts treating it as a feedstock. Data4, a European colocation operator, built the first module with the Universite Paris-Saclay and the startup Blue Planet Ecosystems. The question it poses is sharper than it looks. If AI infrastructure is going to consume a rising share of the grid, what exactly should happen to the heat it throws off?
The heat data centers throw away
Data centers do not mostly fail by overheating. They fail by wasting. The electricity that enters a server mostly leaves as warmth, and most operators still let that warmth dissipate into the air. District heating recovers some of it, but as Lescuyer's own team puts it, that route captures only about 20 percent of the heat a facility produces. The rest is lost.
The company estimates that roughly 18 terawatt-hours of energy sit unused across its French data centers in a year. That is not a rounding error on a sustainability report. It is a resource the size of a mid-sized city's annual demand, vented to the sky because nobody has agreed on what to do with it. Algae is one answer, and an awkward one, because it turns a thermal problem into a biological one.
A thousand-litre lung on the roof
The Marcoussis module runs a simple loop. Server heat is captured and routed to a culture of Chlorella. The algae photosynthesize, pulling in carbon dioxide and converting it into biomass. Patrick Duvaut, vice-president of the Universite Paris-Saclay and president of its foundation, has described the capture efficiency as up to twenty times that of a tree for an equivalent surface area. The comparison is flattering to algae and unflattering to our current defaults.
It's only a question of engineering. This is where it's very virtuous, and then we add another step of circularity because we can convert algae into biomass or into biomolecules or into biohydrogen. This is a first. Nobody has done that before.— Patrick Duvaut, Vice-President, Universite Paris-Saclay
The pilot is deliberately small. A thousand litres is a proof of concept, not a product. The interesting number is what comes next.
What twenty kilograms a day buys
Data4 and Paris-Saclay have sketched the scale-up: modules of roughly 600 litres mounted across about 900 square metres of data center wall. At that size the system is projected to produce around 20 kilograms of algae and capture about 36 kilograms of carbon dioxide every day. Per data center, the capture estimate lands near 13 tonnes of CO2 a year, which the partnership extrapolates to roughly 3,900 tonnes annually across all its French sites.
Independent analysis pegs the capture efficiency of the underlying photosynthesis mechanism at up to 90 percent. The biomass is not sequestered and forgotten. It feeds short supply chains: agri-food ingredients, cosmetics, eventually higher-value bioproducts. A single rooftop unit, in other words, becomes a small vertical farm whose only fuel is heat that was free to discard.
Nearly 18 terawatt-hours of energy are available in our data centers in France, but are currently unused. Through this project, we aim to transform this heat into energy.— Linda Lescuyer, Head of Innovation, Data4 Group
The economics are not yet obvious, and that is the point worth sitting with. Algae as food or cosmetics carries margin. Algae as fuel competes with everything else on the grid. The same module can look like a circular-economy win or a vanity project depending on which output you price.
The catch the spreadsheet hides
Biological carbon capture has a habit of looking better on a rooftop than in a portfolio. The algae still need light, nutrients, and constant monitoring. Scaling from one tank to a wall of them means solving contamination, harvest logistics, and the dull truth that biomass is heavy and perishable. District heating wins on simplicity: pipe heat to homes and the market already exists. Algae asks you to invent a market for the product.
The company is not alone in chasing this. Algiecel, a Danish startup, ships modular photobioreactors that turn captured CO2 into microalgae biomass for feed and food. Blue Planet Systems, on a different track, mineralizes CO2 into construction aggregate. The convergence is the signal: carbon that used to be a liability is being reshaped into something someone will pay for, and the data center is becoming one of the cheapest places to source the heat that drives the reaction.
With this project, we are demonstrating that it is possible to transform the waste heat from digital infrastructures into a living resource. This module is not just a tool for capturing CO2, it is a technological building block towards automated urban aquaculture.— Blue Planet Ecosystems team
Why this belongs on an investor radar
The reason to watch is not the algae. It is the repositioning of the data center from pure energy consumer to a node that exports both compute and recovered resources. As we wrote in August, the agentic-AI wave is pushing data center capital expenditure to levels that make every wasted joule expensive. When hyperscalers are signing power purchase agreements to feed their racks, the operator who monetizes the exhaust gains a second revenue line the others leave on the table.
The open questions are real. Will carbon-utilization credits ever price high enough to matter? Can biomass find offtake before the modules are built? Does the energy balance survive at scale? None of those are settled. But the direction is: waste heat is becoming a first-class asset, and the biology to use it is no longer science fiction. The bio-circular data center is early, small, and easy to dismiss. So were most infrastructure shifts worth owning before they compounded.