How Australia’s data centre lobby is winning its fight to to freely pollute
Australia’s government announced months ago they aim to curb the fast-growing data centre industry’s risks to the power grid; namely impacts around emissions, power bills and grid reliability. The rapid growth of data centres in regions like the US and Ireland has affected all three, and it was nice to hear the Australian government worried about preventing those same problems.
There are a range of ways data centres impact environment but here, I am going to talk here purely about greenhouse gas emissions: the pollution created by coal and gas power stations that drives much of the shocking physical disaster we see playing out across the world. If we don’t stop using these power stations ASAP, the number of people killed by them will be far higher than if we phase them out quickly. It is good to save lives, and it is bad to allow death.
Coal and gas power plants are deadly machines, but we do not act like they are deadly. I think the rise of the AI industry has shone light on a long-running problem, where this wartime-scale catastrophe is treated as a non-urgent side quest. This attitude comes from those you’d expect, like high polluting companies. It also comes, sometimes, from those who purport to act in defence of environment and climate.
The data centre industry has seen itself assisted by centre-left governments and politicians, and even some renewable energy and climate groups, because those people have become desensitised to the acute danger of coal and gas, and far too complacent about renewable energy. They assume it is inevitable, infinite, cheap and unstoppable, but big tech became immediately addicted to fossil fuels when its energy hunger exploded three years ago.
I think a lot of this has emerged through greenwashing (fake climate action). I very strongly contend that much of the industry’s most fervent greenwashing messages have been successfully targeted at centre and centre-left groups; intended to mute their criticism and win their support. Now, we wake up in 2026, and the most prominent climate groups are happily teaming up with the worst offenders to offer high-emissions versions of work previously done better by people.
The resulting pollution, all of it entirely avoidable, adds another heat-trapping blanket layer and that heat kills a massive number of people. The latest update to the Australian government’s data centre regulations lays this all out pretty nicely. Let’s dig in.
Not helping
Australia has a target to reach 82% renewable energy in its power sector by the year 2030. It is behind, but not by a terrifying amount. At the current rate of change, it’ll hit the target four years late.
Of all the power generation required (and this matches demand exactly) in the year 2030, 82% of it should come from stuff that isn’t coal or gas. Pretty straightforward.
But what happens when you add extra demand? In that case, the question arises of what increases to meet that new demand. If it is anything less than 82% renewable energy, the percentage target will be missed.
To still achieve an 82% renewable energy target with higher demand, Australia now has to ensure at least 82% of that additional gap-filling energy is renewable.
What this means is that new data centres make the 82% harder to achieve, because the finish line has been shifted further away from the track, and now you need to build extra renewables just to achieve the same outcome. The problem, of course, is that Australia was already not building enough renewable energy, and now it has to build even more to hit the same target.
Importantly – even if the 82% target is met with this new demand, you still end up with higher greenhouse gas emissions, because the other 18% of the gap is new fossil fuels. This is a design problem with percentage targets, but it’s worth mentioning because I was under the impression eliminating fossil fuels was the reason for this whole energy transition targets thing.
While Australia isn’t too far from the percentage target, the NEM’s emissions are actually ending up way higher than the grid operator’s previous models:
It’s easy to look at the black line and shrug: it’s going down, what’s your problem? The problem is that greenhouse gases accumulate in the atmosphere. The 2024 system plan modelled 317 megatonnes of pollution through the next two years, but it ended up at 355, 12% higher than laid out in the scenario.
The fact data centres will make this pre-existing underperformance worse without actual, real new renewable support is not my opinion. This is just the hard reality of the energy grid and some extremely basic math.
But Australia’s energy minister claimed the exact opposite in a recent speech:
“Data centres aren’t a challenge to 82%, they’re a pathway to delivery”
“Renewables need what data centres bring — certainty, investment, and load that can bend when the grid needs it to. So data centres aren’t a challenge to 82%. Handled well, they help us get there”
So we’re clear: the only way data centres could “help” achieve the 82% renewable energy target is if they built significantly more renewable energy than the demand they added. This raises two extremely fundamental questions there seems to be extreme hesitation to ask: (a) are the companies willing to do exactly this, and (b) if they’re not, do Bowen’s policies force them to do exactly this?
The reality is that tech companies and data centre developers fight aggressively to avoid doing anything more than superficial incentivisation of renewables, when compared to their ballooning demand growth. They use loopholes and offsets to fake renewable energy use, while they’re actively building gas plants at a completely unbelievable rate.
Australia’s climate minister is talking about a scenario where tech companies would have to actively set out to build an unprecedented amount of new renewable energy, but the industry’s action always point to them trying to actually build as little as possible. Does that not strike you as an extremely weird and confusing disconnect?
Why the weird eagerness to act like an explosion in data centre demand will be a significant boost to climate action? Why turn to an industry happy to take two steps backwards, and beg them to fund one step forward?
The data centre industry recognises that the reflexive desire to paint their industry as a ‘climate solution’ gives them material lobbying power. Specifically, the power to lobby for extremely weak regulations rather than strong ones.
They knew a while ago that their core task here was ensuring they were not forced to actually cover new demand with real, truly ‘additional’ renewables that wouldn’t have otherwise been built. They truly do not care if the fuel used to meet their voracious energy appetite is deadly and unsafe.
What they do care about is the presence of neutered pseudo-regulation that allows them to claim climate compliance without any actual responsibilities: not to cover their own new demand gap; nor to actually drive the energy transition. Greenwashing in 2026 embeds itself into laws and legislation.
I recently discovered they’ve been entirely successful.
Bad advice, weak rules
The Australian Energy Markets Commission (AEMC) is an independent statutory body set up to provide advice to governments. When it publishes advice, governments listen. And it was asked to provide advice on how the government should try to ensure new data centres are met with ‘100% renewable energy’ rather than allowing fossil fuels to expand to fill that blue space above.
Last week it delivered the answer (included in Bowen’s speech referenced above): data centres will be required to purchase the renewable energy equivalent of carbon offsets.
I have to do a bit of explaining here. All of this boils down to a straightforward question: how do big new data centres actually ensure the gap gets filled with new renewable energy they’ve brought into existence from their own actions? The answers exist on a spectrum:
| Tactic | In very plain English | In inaccessible jargony nerd language | Does it help renewables get built |
|---|---|---|---|
| On-site renewables | You just build renewable energy yourself. Literally next to or very close to your data centre | The construction of non-emitting facilities on premises, fully funded by the operator with extremely high additionality | ✅✅✅ Yes |
| Long term power contracts (PPAs) – developing projects | You told someone who wants to build a wind farm you’ll buy their power for a long time That allowed them to build their wind farm, and because you caused that, you get to claim their power as your own ‘clean’ supply | A long-term contractual arrangement between a purchaser of energy and a generator of energy, to buy power at a guaranteed price over a long time period. This can assist the generator in acquiring finance and approvals to reach financial close and eventual construction | ✅✅❌ sometimes, but if the project would’ve gone ahead anyway, then no |
| PPAs – existing projects | You told a pre-existing wind farm you’ll buy their power. It doesn’t help them get built, because they ALREADY exist | A long-term contractual arrangement between a purchaser of energy and a generator of energy, to buy power at a guaranteed price over a long time period. This does not assist the generator in acquiring finance or reaching construction because it ALREADY EXISTS | ❌❌❌ No, it ALREADY EXISTS |
| Renewable credits – existing projects | Someone generated clean energy and sold a certificate to prove it. You buying that paper somehow undoes the bad energy you used. Your actions make no difference to the project coming online because it ALREADY EXISTS | The creation of a certificate guaranteeing the generation of one megawatt hour of zero-emissions energy traded as an unbundled certificate and purchased by energy users This does not assist the generator in acquiring finance or reaching construction because it ALREADY EXISTS | ❌❌❌ No, it ALREADY EXISTS |
| Renewable credits – “new” projects | Someone plans to generate clean energy from a planned project in the future, and you agree to buy all the certificates they create | The creation of a certificate guaranteeing the generation of one megawatt hour of zero-emissions energy traded as an unbundled certificate and purchased by energy users – for a project that isn’t built yet and is seeking finance | ✅❌❌ sometimes, but if the project would’ve gone ahead anyway, then no (and it’s not clear these certificates make the projects more likely) |
I know that is overwhelming, but all of these bar the first just boil down to “I helped fund the project, so I get to claim its emissions-free power as my own”, in different ways.
The AEMC’s recommendations basically involve the bottom two; the weakest way to make this claim. First, data centres can simply buy unbundled certificates from EXISTING renewable energy facilities, for some unspecified amount of time. Then there’s a gradual transition towards these certificates coming from “new” projects. “REGO” here means “renewable energy guarantee of origin”.
According to the AEMC, “This will drive investment in new renewable generation and improve investment attractiveness of renewables”. Australia’s renewable energy target worked off a similar basis: wind and solar created certificates, and the government forced energy retailers to buy those certificates at rising percentages, which drove a huge surge in renewable energy growth. But the 2010s were a different era, where this additional income made a material difference. It was also applied across all energy consumers, meaning the force of this effect was significant enough to drive massive renewable growth.
In this scenario, it seems essentially guaranteed that for many years, data centres will simply buy these offsets from existing renewable projects, driving no new renewable energy at all.
Analyst Tristan Edis recently warned schemes like this will actually deter projects from signing long-term agreements (noting he was talking about the certificates from the older scheme, but feasibly similar problems would occur with the new ‘REGO’ scheme):
Tristan Edis, an analyst at Green Energy Markets, said there was a risk the [Clean Energy Council] plan would backfire because a huge oversupply of [Large scale generation certificates, LGCs] and ultra-low prices could discourage data centres from entering into power purchase agreements. “Data centres would have no need at all to support any new projects in order to secure enough LGCs to cover their electricity demand,” he said.
It seems like it might help that there’s a transition to offsets from “new” projects – the word ‘new’ gets used 102 times in the AEMC’s full report. If data centres can only get offsets from new projects, that’ll help new projects get built, right?
But from what I can see, it doesn’t actually get defined anywhere. Does it mean a project that isn’t built yet? That was recently built? That’s five years old? Ten?
We have to go back to the basic question here: how do we know these offset-buying deals actually cause new projects to spring into existence? How do we know they’re additional to the existing pipeline, rather than just an accounting trick to fake that?
Buried on page 28 is a footnote that reveals that fundamentally, it DOES NOT matter whether this policy is actually driving new investment in renewable energy. Here’s the statement:
And the footnote:
This is a bit bonkers, because the word “additional” means absolutely nothing, if data centres do not have to prove their actions resulted in actual additionality. They could whip up a certificate deal with a project that would’ve gone ahead anyway, and you end up with a demand gap filled with fossil fuels, a richer renewable energy developer, and a data centre operator that can loudly shout about how clean it is.
Elsewhere, the AEMC highlights that the word ‘additional’ should be defined with ‘flexibility’, which has always been a term that, when you hear it, signifies that some absolutely rotten corporate wrongdoing is happening:
“What constitutes ‘additional’ is another important design choice, which we consider should be set by Ministers. The AEMC recommends flexibility here, with renewable and firming projects that are not yet at financial close able to be counted as additional. This includes projects that have won CIS contract but would benefit from additional financial support from data centres”
The “CIS” here is the government’s “Capacity investment scheme”: a government policy to bring renewable energy online. If they’ve won a contract with that scheme, how can the data centre be said to be ‘causing’ that project by buying up some greenwashing treat bags?
If a new renewable energy project isn’t brought into existence thanks to a data centre’s actions, then data centres are not filling the gap created with additional clean sources.
They are just piggy-backing off the existing pipeline. That means their new demand gets served by the existing mix: half fossil fuels, half renewables. That is a net worsening of greenhouse gas emissions. Contrast this with the recently-announced deal to keep the large Tomago smelter open, where the idea of being ‘renewable powered’ means direct long-term investment only in verifiably additional, new renewable projects currently stuck in development hell. I still don’t like the wording and we don’t know enough details, but honestly, at least it seems to involve actual new renewable capacity.
So why is the AEMC going with the weakest possible option, in the menu of potential approaches towards trying to actually force data centres to build new clean power sources?
Well, the AEMC happily admit in this document that they considered the stronger option of forcing data centres to enter into long-term power deals, which as we saw above is a notably less bad1 way of ensuring projects come online. They also explain who it was that convinced them not to recommend it:
“Data centres could be required to invest in additional capacity, such as signing PPAs with new renewable generators or directly investing in new firming capacity, to offset their demand. However, data centres expressed concerns that this would significantly increase costs (particularly for smaller data centres trying to match demand exactly) and slow down build times”
Oh no! Some of the richest companies on the planet, in one of the most wildly over-funded industries in recent history, would face increased costs 😱.
What an amazingly open admission of a high-polluting industry killing off regulation because they consider their own impatience a higher priority than atmospheric safety. It’s nice to have such a transparent demonstration of the industry lobbying for its own right to pollute.
So there you have it: data centres lobbied the AEMC to kill off even a middle-ground approach to forcing them to actually build new renewables. Now they can use a far dodgier and weaker version instead.
If offsets did not exist, it would be necessary for capitalism to invent them
Something important to note is that even if data centres were to be responsible for building renewable energy, that does not neutralise or undo the harm that they do cause. It is meaningful that both the AEMC and the data centre lobby itself uses the phrase “offset” to describe what is happening here.
Instead of meaningful cuts to climate pollution in the power sector, we get the strained, weird logic of moral neutralisation where funding some good inexplicably undoes the bad that was caused. This renewable energy offsetting suffers from the same fundamental problem as all offsets. A large number of technocratic nerds spend a lot of time trying to form a tight grip around extremely slippery parallel universes and counterfactuals, when they could be dedicating their time and energy to fighting against fossil fuel use instead.
This is the entire point of the neoliberal world of ‘market-based mechanisms’ as a replacement for actual protection of human society from the danger of fossil fuels. It is a pseudoscientific immune response created by capitalism to ensure annoying stuff like ‘do not kill your customers through pollution’ doesn’t impact plans for short-term gain.
It’s in these fuzzy counterfactuals that much of the problem lies. The financial incentive here is not for maximum emissions reductions and fastest fossil fuel elimination. For generators, it’s to sell as many certificates as possible. For data centre operators, it’s to buy just enough certificates that you fend off criticism and delay or destroy any chance of real regulation. Sometimes these motivators will drive climate action by accident, but it’s never the core goal. That means the default future here is a booming data centre industry that makes cumulative emissions far higher than they would’ve been without it.
I have always despised this dirty deal. ‘Let’s allow a little bit of greenwashing as a treat, in the hope the industry will hurl some pocket change at some form of climate action’. There has never been a deal like that where climate has ended up as the winning party. We should measure, report and criticise the pollution of these companies and that should never be adjusted based on how charitable they were feeling at the time.
Why parts of the climate movement love data centres and AI
It’s amazingly clear in the AEMC’s document that the danger posed by Australia’s energy system, whereby greenhouse gases kill and harm people, has been ignored. Despite this pollution being included in Australia’s National Energy Objective (NEO), and the AEMC citing that objective and mentioning emissions several times, no part of this plan actually analyses what happens to greenhouse gas emissions when you weaken data centre climate regulation.
This type of work has been done elsewhere: Irish academic Hannah Daly has done a bunch of vital analysis showing how this “glide path” approach allows for many years of unrestricted greenhouse gas emissions. Greenwashing, particularly when embedded deep into legislation, kills people.
What I see playing out in Australia, and many other parts of the world (like Canada) is this inexplicable desire to see data centres as a climate saviour, when the basic math of their exploding power demand and the behaviour of their lobbyists both show they’ll be nothing but a villain.
One nice example of this seriously problematic instinct is the constantly repeated claim in climate and energy spaces that data centres can ‘flex’ their demand to ‘help integrate renewable energy into the grid’ (as you saw in Bowen’s quote above). Even if that were to be the case, for the other 99.99% of the year, the data centres stubbornly push up demand and increase fossil fuel combustion, so the net effect is worse climate outcomes. But the grid operator recently quietly revealed that data centres basically aren’t even interested in actually doing demand flexibility in the first place:
This promise of “demand response” at key moments from AI (and Bitcoin people, who love this talking point) is like a man who idles his car outside a school 24 hours a day promising to turn it off for five minutes in the afternoon. AEMO’s data shows that he isn’t even turning it off for five minutes. Why does a revelation like this essentially have zero effect in the enthusiasm out there for panels, podcasts and press releases about how data centres are a climate solution?
One of my nightmares during the Bitcoin hype cycle was renewable energy advocates actually falling for the talking points, and with AI slop, that nightmare has come true.
Another major part of the problem here is that Australia’s chief clean energy lobby group, the Clean Energy Council2, itself suggested the government should allow years of ineffectual renewable offsets (alongside longer-term contracts). They even allow a dangerously vague role for fossil gas to be used as ‘additional reliability’, as long as the emissions are “offset via certificates“. It has been extremely obvious for years the gas industry’s chief wedge tactic was selling it as just for ‘backup’ but ensuring it gets burned constantly; something even more clear in the data centres case. The CEC know this too, so leaving that door open is directly assisting Australia’s fossil fuel industry.
Australia’s clean energy groups operating on the thoughtless assumption data centres are a ‘climate solution’ has led to the incredible, absurd situation where the government clean tech loans body, the Clean Energy Finance Corporation (CEFC) announced they’re directly funding data centres that are “procuring” renewable energy:
“We haven’t done [a data centre deal], we haven’t announced one, but they are in our pipeline that we would be providing finance to data centres, obviously on the basis that they are procuring renewable energy”
Here is my challenge: someone please explain to me why a government loans body is lending cash to a data centre instead of just cutting out the middle-bro and lending directly to the renewable energy company.
The idea that tech companies or data centre developers are short of cash is formally impossible to make fun of. No industry in the history of anything has ever been less in need of money. It is not “lack of money” that causes them to turn to fossil fuels to power AI slop factories.
This definitely feels like it should’ve drawn more of a reaction than literally no reaction at all.
There’s also the now long-running idea that Australia should be a kind of ‘green export superpower’ (or some variation of that phrase). It’s been picked up by centrist governments in the UK and Canada; kind of a tantalising vision not of rapid fossil fuel elimination but the preservation of nationalistic power, influence, wealth and comfort in western nations freaking out they’re losing all those things to surging growth in Asia.
The projects and technologies behind these visions, like ‘green hydrogen‘ and massive export cables plugging the Northern Territory into Asia, are being cancelled and their websites are being find-and-replaced with data centres instead. Here in Norway, Fortescue’s Andrew Forrest is getting in trouble because planned green hydrogen projects are being replaced with data centres. Monash University (in a baffling partnership with climate-denying News Corp) is talking up the potential for Australia to export ‘clean AI’ to other countries, despite Australia still having a higher emissions intensity for data centres than America.
Again: why? Putting aside how all these technocratic visions ignore the actual harms produced by generative AI, what drives the instinct to paint big tech, data centres and AI as something we should treat as a climate solution alongside solar and batteries?
One of my theories here is simply that a huge number of middle-to-elder-millennial climate and energy professionals instantaneously developed problematic addictions to generative AI, sometimes verging on AI psychosis. We are old3 enough to be in middle-to-upper management now; the perfect mark for sycophantic slop-bots (as younger generations express deep skepticism and discontent). Constant exposure to an extremely effective conversation-faker makes a decent chunk of my generation wants to Apu-leap in front of data centres. But that’s probably only a fraction of the problem.
The deeper problem here is that, in contrast to the focused moral clarity of the late 2010s, we are now in an era of ‘side effect’ climate advocacy. The move away from fossil fuels is now presented as an unstoppable, infinite and technologically marvellous shift that spreads ‘energy sovereignty’ and cheap power bills. Prolonged fossil fuel use isn’t a mere bump on the road to the same point in time in the future; it’s a delay that does damage. That is why any of this is being done in the first place, and it’s being forgotten very often.
Climate groups deleting the word “climate” from their statements and insisting that nothing in the world could ever stop the energy transition breeds complacency and confusion. It doesn’t matter that the transition is ‘inevitable’: it is extremely slow-down-able and big tech is already succeeding in slowing it down.
The data centre boom is also reminder that none of the people in the tech or data centre industry are economically rational: they will never give the slightest shit that solar and batteries are cheaper than the filthy, spluttering fossil fuel plant crouched in the corner of their land parcel. AI is a mouth hungry for fossil fuels for the same reasons Bitcoin: desperation, greed and anxiety are the best friends coal and gas ever had.
The mad behaviour of AI proponents building country-sized gas-fired power plants and murdering regulation is a direct threat to the passive ‘side effect’ climate advocacy worldview.
If the energy transition was inevitable, this massive new power demand would be defaulting to cheap and boundless “electrotech”. But the energy transition is not inevitable. We have to work hard, for the rest of our lives, to make it happen. I feel like we collectively understood that a decade ago, but in 2026, it feels distant and hazy.
When you’re all in on the narrative of a self-driving, self-propelling energy transition, you will be naturally compelled to dismiss the climate threat of AI while aggressively overstating its potential climate benefits. It is no coincidence that the most forthright data centre and AI advocacy has come from the same groups that have erased climate as a focus area, and become focused on ‘side effect’ advocacy instead. And all of this has consequences.
Australia’s coal closure delays are getting worse
Let’s go back to Australia’s story.
The damage done by fossil fuels is the worst engineering failure in the power sector’s entire history. No other pollution failures will ever match the death toll and financial cost of the atmospheric ruination caused by coal and gas fired power stations. Yet the power sector’s blank-faced shrug when it comes to the urgency of stopping this safety failure ASAP means it gets addressed at a pace that is glacially slow.
Demand projections that lurch upwards each iteration mean the media, industry and government case for keeping Australia’s disgustingly polluting coal plants open to belch more deadly pollution for another half-decade gets stronger.
Victoria’s Yallourn plant is next on the production line of this death-extension project, with the media articles already kicking off the “oh my god, BLACKOUTS” festival that surround every single coal plant extension in the country. NSW is gearing up to performance same rhetorical dance for the coal gods. The general idea here is that a few decimal points probability of increased blackouts is signficantly worse than the several thousand people who will lose their lives to the atmospheric consequences of burning coal for a half-decade longer. It’s anti-vax level absurdity, but in Australia it’s perfectly normal.
Renewable energy is not getting built fast enough, and projected demand is growing. Two lines moving further away from each other, creating a bigger gap to be filled by humanity’s deadliest fuel product. Stare this situation in the face: if things were going well, it would be nearly impossible to enact these harmful coal plant extensions.
Australia is frequently cited as basically a decarbonised solar paradise, around the world. The fact that it still has the world’s highest per-capita coal power reliance is not well known. The reality is the government took its foot off the accelerator for climate policy, and the infrastructure correspondingly slowed down to the ‘natural’ free market pace. When you delete climate as a priority, and you’ve gone all in on presenting renewables as “inevitable”, you’re forced to look away from how bad this situation is.
Australia’s power industry scoffs and sneers at any thought of climate as an urgent, immediate ongoing concern (whether that’s seeing fossil fuels as inevitable, or seeing renewables as inevitable). Instead of pushing back against that story, the renewable energy lobby group is arguing in favour of weak renewable certificates, gas and offsets for that gas. Governments are obsessed with framing data centres as a ‘climate solution’ that ‘help the energy transition’. Big tech and data centre developers care solely about their short term gains and will set fire to pretty much anything to fuel those gains.
A perfect storm. In these conditions, it’s stunningly easy for data centres to dump yet another big, heavy stone onto an energy transition that was already too sluggish. The price will be paid by those who never asked for any of this.
- Companies like Microsoft and Google have moved away from renewable offset certificates, and are at least claiming emissions reductions using long-term deals instead. Microsoft’s ‘adjusted’ power emissions actually rose massively last year, purely because they stopped relying on dodgy renewable offset certificates! ↩︎
- The former head of the CEC, Kane Thornton, left to join data centre company CDC, who claim to be 100% renewable solely through the purchase of renewable energy certificates. ↩︎
- I think ‘boomer’ should be a rolling phrase that applies to the worst generation of any mix of generations. We are the boomers now, and we have our own special weaknesses and young people will correctly hate us for being such slop-addicts. ↩︎