The world’s most powerful computer fits into a shoe box, takes 80w, has an MTBF of 60+ years and is liquid cooled, yet it has no warranty and can take a life time to train, its the human brain. In recent years strides in AI are getting close to replicating this marvel which has taken millions of years to evolve, but what damage to the planet ?

In 2023 the BBC reported that AI globally could consume as much power as the Netherlands which is ironic considering this is one of the countries most at risk due to rising sea levels caused by global warming.

So with the recent COP climate summit struggling to get the world to bet zero but a relentless need for AI how can we square this circle without continuing planetary annihilation.

One potential answer is community cloud…

By distributing cloud to the edge and utilizing the new range of liquid cooled chips and chassis we can use the waste heat from data centers to power our communities.
Conventional wisdom is to build ever bigger data centers with 100mw and now 500mw being common place, this drives increased efficiency , but (quoting Einstein) “energy can not be created or destroyed it is only hanged from one form to another”. So when a data centre used 500mw of electricity to drive computers the energy is converted into 500mw of heat. This is generally wasted into the atmosphere. Some pioneering companies are now putting data centers on the sea bed to harness tidal cooling. But the pricipal still applies that this heat is not only being wasted but it is also making a tiny contribution to the planet’s warming.  343,400,000mwh is the UK estimated total domestic gas consumption for heating requirements. By comparison a 500mw data center consumes 4,380,000 mwh in one year.

UK data centre capacity is therefore estimated to be the equivalent of 10% of the UK domestic heating capacity by 2026.

Clearly connecting every data center to every house is an impossibility. We have also gone from PC’s and computers in homes to cloud based compute , the arguments and inertia to move all the way back to local computing in houses simply wont work either.

So is there a mid point where by we can maximize cloud computing and provide waste heat to heat our communities?  The economics above speak for them selves but when we consider energy sovereignty and global price fluctuations this becomes even more compelling.