No AI data centres in my backyard, please

Last week, the Tasmanian Greens announced they’ll move for an urgent parliamentary inquiry into the AI data centres going up across the north of the state. I think this is a very good idea. These projects appear to have been approved without any specific regulations or parliamentary oversight, and the public has barely had the chance to have a say. The announcement also put a new number on the record: in Budget Estimates, under Greens questioning, Energy Minister Nick Duigan confirmed Firmus plans to draw 400 to 500 MW across two or three “AI factory” sites, which is more than Bell Bay Aluminium’s 335 MW, currently the state’s single biggest power user.

A lot of people seem convinced Tasmania can become some kind of “AI Island”. We’ve supposedly got cheap hydro power, and the world apparently wants somewhere clean to run artificial intelligence. So we sell ours as “green compute” and the investment rolls in. That’s the theory, anyway. The Premier loves it, and so does a Singaporean startup called Firmus.

One of the core problem with this is the power. We don’t uniformly have a surplus to sell. Last year we were importing electricity from Victoria just to keep our own lights on, and the government wants to hand about a sixth of the grid to a handful of foreign-owned sheds full of NVIDIA chips. What we get back is a few dozen permanent jobs, a stack of environmental claims nobody has audited, a great deal of noise, and a contract we’re not allowed to read. It’s a bad deal, and it’s being pushed and pushed and pushed.

And to be clear, this isn’t anti-AI. You can use AI, like it, even love it, and still think this plan is a bad deal for Tasmania. You can love electronics and still not want a lithium mine next door. Whether you’re for or against it barely matters here. What matters is whether handing a foreign company a big slice of our grid, water and goodwill is worth it for the people who actually live here. There are better ways to get a few dozen jobs, and most don’t rely on a huge chunk of our power grid, and lots of concerning impacts on things.

The spare power isn’t there

Let’s start with the Regulator’s own numbers. In the 2024-25 water year Tasmania imported 1,983 GWh from the mainland and exported just 433, a net 1,550 GWh flowing into the state, most of it made by burning brown coal and gas in Victoria. We were a net importer across Basslink for nearly every month of the year, hydro inflows were the third-lowest on record since we joined the national market, and we ran the gas plant to get through. The renewable surplus this whole scheme leans on wasn’t actually there. As one local writer put it, the grid is “tight. Not abundant and infinite. Tight.”

If we weigh that against the demand… well, it doesn’t look great. Firmus’s Launceston factory is contracted for 104 megawatts on its own, about what the entire Boyer paper mill draws, in a single building. Across its three sites at St Leonards, Bell Bay and Wesley Vale, Firmus is heading for 400 to 500 MW. The reporting on the Bell Bay plans put the combined Launceston-and-Bell-Bay figure at close to 15 per cent of everything the Tasmanian grid can supply, and even that only holds “unless new wind, solar or hydro generation is built.”

So this amazing AI data centre plan relies on a fleet of new generators that don’t exist yet, on a timeline nobody has guaranteed, which may or may not happen. Until they’re built, every megawatt that goes to a data centre is one taken off someone else.

Taken from the industries already in the queue

There’s already a queue for that power, too. Firmus was waved through for its 104 megawatts, but our own industry (industry that’s already here, doing useful things) has been told to wait. The Boyer paper mill, which employs around 300 people in Tasmania, has been trying to get off its coal-fired boiler. The conversion needs an extra 45 to 60 megawatts, and Hydro Tasmania told the mill the capacity wasn’t there.

So the rules bend for the right foreign investor, making something of spurious value, but not for an existing industry that already has some form of demonstrable value and jobs. Greens leader Dr Rosalie Woodruff put the contradiction to Parliament directly:

“How is it possible that established industries in Tasmania aren’t able to get additional power generation, but a Singaporean company who walks in and suggests a $2-billion investment are able to whistle up what they need?” — House of Assembly, 25 March 2026

Every gigawatt-hour we sign over to AI is one that can’t get Boyer off coal, expand Bell Bay Aluminium, or electrify our homes and cars and bring people’s bills down along the way. We’d be rationing power to Tasmanian workers and households so a foreign company can rent out AI compute to whoever pays the most for it. To say nothing of the supposed benefits of AI, which look shakier by the month.

Not many jobs

Unless one of these places hands something to locals directly, the whole case for it comes down to jobs and economic growth. And it doesn’t hand us anything direct: no free or faster AI, no product, nothing at the gate. So weigh what it asks for against what it gives back. It wants space, up to a sixth of the grid, water, money, new laws, a long list of environmental and neighbourhood effects, and a big slice of our limited links to the mainland, for both power and internet.

What comes back is jobs, and not many. The boosters lead with the big figures: up to 100 jobs, $2.1 billion. Wow! But strip out the construction crews, who pack up once the building’s done, and look at who runs the place day to day. Firmus’s own planning documents put Launceston at about 30 staff around the clock, and the bigger Bell Bay site at roughly 70 in the day plus 25 on each of the evening and night shifts. Call it 150-odd permanent jobs, for 400 to 500 MW of the grid.

The honest way to weigh that is jobs per megawatt, and on that measure a data centre is about the worst deal going. Bell Bay Aluminium employs close to 500 people on 335 MW. The Boyer paper mill employs around 300 on roughly what Firmus’s Launceston site draws on its own. That’s something like five times as many jobs per megawatt at the smelter as at Firmus, and closer to ten at the mill. Firmus wants about as much power as the two of them combined, to employ a fraction of the people.

It doesn’t seed anything else, either. A smelter or a mill buys from local suppliers and anchors a supply chain; Bell Bay Aluminium alone supports almost 300 local businesses. A data centre is a sealed box of imported hardware. It doesn’t start a new industry or feed an existing one, so nothing much grows up around it. And there’s no guarantee even the few jobs go to locals: Firmus is a Singapore-based company running a highly specialised operation, and the open question, as the Tasmanian Times put it, is whether Tasmania ends up just “a provider of green energy and a handful of jobs”. There are much cheaper ways to make a few dozen jobs.

The claim that data centres transform local economies doesn’t hold up either. The most careful study going, from Brookings, found the first big data centre in a county lifts private employment by 4 to 5 per cent over five or six years, a real effect but a modest one, and that industry’s own estimates overstate it threefold. The honest number gets tripled before it reaches a press release.

These places are built to run on a skeleton crew. That’d be fine if we were drowning in spare power, but we’re importing it.

The green credentials are unverified

The environmental case is a non-trivial portion of the sales pitch, and almost none of it has been independently checked. Firmus says the Launceston site will use about as much water as 50 households a year, “up to 99% less than a typical data centre”, running dry cooling except on the ten hottest days. That figure’s on the company’s own FAQ page, describing a building that doesn’t really exist yet. Its Bell Bay application, meanwhile, concedes up to 22,000 kilolitres of cooling water in a peak year, “most of which would be lost to evaporation”, and lists diesel generators on site for backup.

There’s definitely some real carbon concerns, there. Add hundreds of megawatts of demand in a dry year while you’re importing fossil power from Victoria, and that power doesn’t turn up clean. At the margin you’re pulling Victorian coal across the cable and calling it green. “Clean AI” only earns the label once the matching renewables are built, and they don’t seem like they’re anywhere near it.

Investors have their doubts too. One analyst flagged Firmus spruiking a Power Usage Effectiveness of 1.03, close to the best in the world if it stands up. It’s self-reported, and nobody independent has checked it.

Spare a thought for the neighbours

Wherever these things get built near people, the people tend to end up hating them. Data centres run hot and loud around the clock: banks of cooling fans, chillers and backup generators that, as the US Environmental and Energy Study Institute puts it, throw off high- and low-frequency noise that “can be heard for hundreds of feet”. The low hum is the nasty part. It never stops. Residents near these sites report wrecked sleep and falling property values, there’s a wave of noise-nuisance lawsuits underway (including one in West Michigan), and a quiet suburb in Chandler, Arizona spent the better part of a decade fighting one. Then there’s the air: VCU researchers found Northern Virginia’s data-centre diesel generators now rival a power plant for emissions.

To be fair to Firmus, siting matters, and at Bell Bay it works in their favour. The George Town campus goes on the old Gunns pulp-mill land, which Firmus rightly calls “well separated from homes and residential areas”. It’s a heavy-industrial zone, and a server shed is hardly the worst neighbour that stretch of the Tamar has had. The full suburban horror story probably doesn’t apply there.

St Leonards is a different matter. That one sits in Launceston’s eastern suburbs, close enough that the council bolted noise-attenuation conditions onto the permit, and locals had already raised concerns about cooling-system chemicals. The Bell Bay plans, industrial zone or not, still list diesel generators on site. None of it is a deal-breaker on its own. But with no Tasmania-specific noise or air rules for these places, residents are left trusting the standard council conditions to hold the line, the same kind of conditions that, everywhere else, have ended in court.

Higher bills, including here

Modelling consistently indicates data centres will push up local power bills. Work reported by The New Daily and the Climate Council has data-centre demand pushing household power prices up by as much as 26 per cent within a decade, with wholesale prices in New South Wales up 26 per cent and Victoria 23 per cent by 2035 if the build keeps outrunning new renewable supply, which it is. This is on top of existing regular increases, which we’re already subject to.

And we don’t dodge it by hosting only a little of the load! The Clean Energy Finance Corporation’s Getting the balance right report found that even where barely any data-centre demand is added here, “some price impacts are still expected as their generation is exported to other states to meet demand.” We’re wired into a national market, so the pressure reaches Tasmanian households whether the sheds

[truncated for AI cost control]