A reality check on feeding and housing AI
As the appetite for AI grows, NSSN Board member Dr Diana Day writes, so is the demand for new power-hungry data centres.
AI has entered mainstream consciousness in recent years, especially since ChatGPT was released in late 2022.
This technological marvel is not without real cost in resources and power.
AI in its multiple manifestations, and particularly the use of data-heavy generative AI, means it needs feeding and a place to live.
The old server in the back room of a large business will not cut it in 2024. Using AI is spectacularly hungry of energy and water for cooling.
It lives in large data centres through cloud computing services. Cloud services are already a big business and very profitable.
The demand for new data centres is increasing as appetite for AI grows.
Goodman Group’s Project Pluto plans multiple data centre buildings at one site in Sydney and appears to have a pipeline of more development interests.
Data centres demand climate-controlled data storage facilities.
They must run 24 hours a day, seven days a week, 365 days a year, so they need to be secured and protected from fire and flood.
And they are expanding in cities and towns in light urban and peri-agricultural belts.
There is advertising by many firms who offer the service.
Even the big international Blackstone investment monolith suggests this critical infrastructure has a demand outpacing energy provision.
Data centres are now vital to social function as we know it.
There is a rush for new locations and there may be local impacts in greenfield sites.
For example, Google opened its first data centre in Oregon at The Dalles, near the lower Columbia River, in 2006.
The site targeted a smallish town a bit down on its luck, and locals were concerned that their local streams would be targeted for cooling the facility and considered their niche orcharding agriculture at risk.
Most data centres tend to snuggle within the urban landscape to tap directly into water and power especially within the CBD or technology corridors and of course advertise secure sovereign protection.
As it transpired Google has modified the entire water cycle and created new water and sewer projects for the city including aquifer storage and recovery and as well transferring water rights from one of their owned sites to the city. Jobs were an added incentive.
There are efforts to be more environmentally sensitive, towards NetZero, by reducing large power draws, greenhouse gas emissions and water consumption.
More cloud providers are investing in sustainability initiatives including building design, renewable energy projects elsewhere, and carbon offset programs.
Green cloud certifications and regulations will become more significant, and some companies offer data centre sustainability advice.
The Australian Department of Climate Change Energy, the Environment and Water give a host of data centre guidelines on their website and have also developed the new NABERS [ National Australian Built Environment System] environmental criteria for data centres.
NEXTDC’s MI Melbourne Data Centre is the first to feature the NABERS 5-star rating.
More than one data centre has boasted using solar cells on rooftops for non-essential energy needs and recycling runoff water.
However, these efforts are outpaced by the demands for more computing power.
Emerging technology, such as quantum and edge computing, may close the computing demand gap and alleviate some of the pressure on data centres.
The NSSN is striving to innovate smart sensing technology towards NetZero by reducing size, weight, power and cost (SWAP-C) to process data and run AI at the edge.
Dr Diana Day is an NSSN Board member who has 30 years’ experience as an independent non-executive director of federal and state statutory authorities, the private sector, government R&D and advisory organisations, higher education, and social enterprises.