Data Centre In Space (Part 3): Why The Real Solution To AI Infra Load May Still Be On Earth

The AI industry is exploring the feasibility of orbital data centres as a possible solution for widespread environmental issues associated with the power- and water-hungry terrestrial structures. However, the concept not only requires breakthroughs to solve engineering bottlenecks like thermal management and ground communication but also needs to wait for a few years when space launch becomes much cheaper and starts making financial sense. In the meantime, the data centres may either continue to put pressure on natural resources or face public opposition for the same reason.

While space companies continue to look towards the sky for answers, some industry experts believe that the real solution is on Earth, and does not require breakthroughs. Embedded in boring practical measures like renewable energy, closed-loop cooling, and smarter infrastructure, they believe that terrestrial data centres can be made sustainable without venturing into orbit.

IS THE ‘REAL’ SOLUTION CLOSER TO HOME?

According to Abhishek Agarwal, President, Judge India Solutions, space facilities make for good headlines, but neither is solving anything this decade. “The real fixes are sitting in plain sight already,” he says and points towards practical solutions based on technologies that are already available today:

1. Using liquid cooling instead of the evaporative systems that drink millions of litres of water a day.
2. Building near actual renewable power instead of wherever land happens to be cheap.
3. Capturing waste heat and routing it into nearby homes or industry, which Nordic operators have been doing for years now.

“None of this requires a breakthrough. It requires operators choosing to build properly where they already are, instead of hunting for an exotic location to dodge public scrutiny,” he adds.

Cost Economics of Orbital Data Centres (ETV Bharat Creative)

Manoj Dhanda, Founder and CEO, Utho Cloud, also shares a similar view. He calls orbital data centres a distraction from the real work that needs to happen on the ground, arguing that the industry should instead focus on three immediate priorities:

“First, take renewable energy commitments seriously. Not carbon offsets, not green certificates purchased on paper — actual renewable energy powering actual data centres. The gap between what hyperscalers claim and what they actually deliver on clean energy is significant, and it needs scrutiny.

Second, invest in efficiency before capacity. The AI industry's default response to more demand is more hardware. But a meaningful portion of AI compute is wasted on inefficient model architectures, over-provisioned infrastructure, and workloads that run 24/7 when they only need to run for four hours. Software-level efficiency improvements can reduce energy consumption substantially before you ever add another server.

Third, and this is something we feel strongly about: locating data centres where the energy math makes sense. India has extraordinary renewable energy potential, particularly in solar. Building AI infrastructure in regions with abundant clean energy isn't just environmentally responsible — it's economically smart in the long run.”

“The US is facing protests because they scaled fast and asked questions later. India has an opportunity to build its AI infrastructure era differently with sustainability as a design principle, not a PR statement. That window won't stay open forever,” he adds.

'Real' Solutions to the Problems Associated With Data Centres (ETV Bharat Creative)

Agarwal and Dhanda's suggestions are not without precedent. Several data centre projects around the world have already begun adopting measures to reduce their environmental impact, including Reliance and Meta's proposed 168 MW AI data centre in Jamnagar, Gujarat. Announced last month, the facility is expected to run entirely on renewable energy and use desalinated seawater for cooling, a design intended to minimise carbon emissions and reduce reliance on freshwater resources.

WHAT ABOUT UNDERWATER DATA CENTRES?

Apart from sending data centres into space, another possible fix being tested is submerging them into the ocean. Project Natick (by Microsoft) and Shanghai Lingang Project (by China’s HiCloud Technology and China Communications Construction) are the front runners in this area, but they deserve a thorough independent discussion of their own.

Shanghai Lingang Undersea Data Centre is jointly developed by HiCloud Technology and China Communications Construction (Image Credit: Shanghai Hailanyun Technology)

Abhishek Agarwal clubs both the orbital and underwater facilities as distractions but considers the latter to be a more sensible choice when asked to compare the two.

“Space sounds futuristic right up until you look at launch costs, the near impossibility of fixing anything that breaks, and radiation chewing through hardware faster than it would on Earth,” he says. “Underwater wins this comparison fairly easily. Microsoft already ran Project Natick and found server failure rates were actually lower underwater, helped by stable temperatures and zero human interference, with cooling that comes essentially free from the ocean itself.”

However, Agarwal also raises concerns about this approach, but argues that those challenges would be easier to address than the technical and operational problems associated with space-based data centres.

“Underwater still raises legitimate questions about marine life and local water temperature, but those are engineering problems with known paths forward. Space computing, for now, is mostly an expensive idea with a great press release attached to it,” Agarwal adds.

ORBITAL DATA CENTRE FOR SPECIALISED USE CASES?

While domain experts remain divided over whether space is the right answer to the data centre crisis, there is broader agreement that orbital computing, if it succeeds, will likely begin with specialised applications rather than replacing conventional facilities.

Ronak Kumar Samantray, founder and CEO of Hyderabad-based Indian space tech company TakeMe2Space, does not see orbital data centres as a replacement for terrestrial facilities. Instead, he believes they should exist as specialised workloads that are better suited to space, at least initially.

In the final part of this series, we explore this idea as well as India's orbital data centre ambitions that may start by processing data in space—not moving the cloud there.