While I like unquietly pontificating on policy, I admit it, so don’t rush to my defence. I do still, even at my advanced age, from time to time, like getting my hands dirty by doing actual projects. It keeps me grounded and informs my policy ideas with real-world experience. Oh, and it pays the bills.
Recently I’ve been part of a team seeking to develop data centres, the key real estate typology of the moment that achieves all the following: data centres do the job they are intended to perform efficiently; they do so as environmentally sustainably as possible; and, if they are close to an existing or new settlement that they fit into the character of that area but also, crucially, provide significant local community benefits.
None of this is easy, and all of it is necessary as these things are coming and they are big. So we are faced with the choice of not closing our eyes and hoping they go away, but of using our skills and values to minimise their negative externalities and maximise the positive impacts we can leverage from them.
Spinifex is an opinion column. If you would like to contribute, contact us to ask for a detailed brief.
Informed by this outlook and influenced by a podcast I did on AI and the Next City with the great futurist and architect, Professor Carlo Ratti of Massachusetts Institute of Technology, I have been part of a team that submitted a prototype and model for “data-centre enabled urbanism” to the ongoing Venice Architectural Biennale. Sydney innovation has gone international.
Underpinning our objective was the desire to turn something communities can often reject into something they might embrace. Basically, if a data centre is just an enormous, sterile black box that sucks up electricity and water and offends the eye as well as the property values of concerned householders, objections will be vociferous and understandably so.
Some such centres are indeed beasts, and the bigger ones currently in more exurban sites can be the size of several football fields, with some needing a gigawatt of energy – enough to light up a reasonable-sized town.
Butif they can be elegantly designed, become more compact, contribute to the quality of places and landscaping and provide spill-over benefits, such as warm water for local home heating systems and amenities such as pools, and the capacity to use excess grey water from new local housing development in the running of the data-centre, they can be seen as adding significantly to local community amenity.
That is to say, bringing advantages beyond their function as what the UK government terms necessary national infrastructure”, and prioritised in a new planning system as a key enabler of the new digital and AI economy.
Two things became clearer to me while working on this pretty leading edge initiative
One is that the energy needs of AI and the data centres required to fully enable it, pose massive new challenges to the achievement of net zero goals and timetables.
This moment feels like the old joke from that spoof British History book, 1066 and All That, when its authors said that “just when the English thought they’d solved the Irish Question, the Irish changed the Question”.
That is to say that just when renewables had begun to play a significant role in the quantum of electricity needed, the quantum needed to be massively increased due to AI and data centres.
So big is this quantum that I feel that planners and policy-makers, certainly in the UK but also in Australia, are a tad in denial as to the challenges posed to their net zero and Paris Accords commitments and the dates by which success was meant to have been achieved.
Necessity being the mother of invention, has inspired the data centre project I am involved in. But I suspect some bigger innovation from the government itself will now be required.
This brings me to the second thing I personally think is now clearer, though this is something that currently divides Australia from the rest of the Anglosphere.
That is, it’s now probable that there will be a new push to consider nuclear energy to meet the excess demand we are now beginning to experience in the early stages of this AI Era.
In Australia, nothing seems likely to move decades-old opposition to the nuclear energy sector, and I understand the reasons behind this. Partly opposition came because of nuclear energy’s former affiliations with nuclear weapons – often denied, but there were interrelationships and overlaps in the UK in reality – but also because of the need to dispose of the challenging and lastingly toxic waste produced from nuclear energy in a country where many worry about the adverse impacts on both the environment and sites of significance for Indigenous communities.
Australia’s Climate Change Authority is currently warning that the surging energy demands of artificial intelligence. It has explicitly flagged the explosive growth of AI in data centres as a risk that could strain the rollout of renewable power.
Meeting radical national targets to cut emissions by 62-70 per cent by 2035 hinges, the authority says, on a rapid expansion of renewable energy as well as greater electrification of sectors such as transport and heavy industry.
As this is just not feasible politically in Australia, it doesn’t mention the nuclear option at all.
In the US, we know that there are no federal government boundaries now on exploring nuclear energy, but also that private companies such as Microsoft and Amazon are looking at the potential for next-gen small modular reactors (SMRS) to be used to support data centre activity and the needs of AI deployment.
In the UK, we see similarly ambitious cuts to energy by 2035 as Australia, we see Ed Miliband, the progressive Energy Secretary, expressing robust support for nuclear energy as a component of the country’s clean energy strategy and specifically for the development of small modular reactors.
He says his “door is open” for new nuclear projects, with him emphasising the need for a diverse mix of energy that includes nuclear, wind, solar and other renewable sources to achieve the UK’s energy goals and ensure its energy security.
That door is currently not open in Australia, but one wonders how long that opposition will survive in the face of exploding new real world demand from AI. We shall see – and I will report on it either way!
[The Fifth Estate has published many articles that outline the unfeasibility of nuclear power in Australia. Changing the legislation would alone require years of political persuasion, and before enormous costs and lead times are considered (perhaps 15 or 20 years). Asking governments to bear the costs is also more likely to spawn a new industry more invested in huge consulting fees than eventual success, while distracting time and money from finding alternative net zero solutions – Ed]
