Essay by Eric Worrall
They knew a decade ago – “… Battery technology is not yet at the stage where it can fully absorb that firming need and requirement. …”
Google says it needs gas to power big AI data centres
Jenny Wiggins Senior reporter
Oct 13, 2026 – 6.21pmGas is needed to help power data centres that process artificial intelligence because renewable energy and battery storage are not reliable enough, says US technology giant Google, which is searching for development sites in Australia.
“Gas has a role to play,” said Alexander Smith, Google’s principal for global infrastructure and energy. “Battery technology is not yet at the stage where it can fully absorb that firming need and requirement.”
The federal government is urging data centre projects to invest in new renewable energy as it draws up new rules expected to be passed into law next year to ease fears that developers’ need for energy will increase carbon emissions.
But although developers are keen to trumpet their use of renewables, the biggest data centres – such as Anthropic’s proposed $30 billion project in Queensland’s Western Downs Digital Park – are expected to draw on gas to provide consistent power.
…
Read more: https://www.afr.com/policy/energy-and-climate/google-says-it-needs-gas-to-power-big-ai-data-centres-20261013-p61569
Google management had solid evidence renewables are not fit for purpose more than a decade ago, but chose to ignore their own researchers. They pushed green fantasies instead of genuine energy solutions.
From 2014;
All these tech giants pride themselves on their rational outlook and evidence based thinking, yet despite mountains of adverse evidence, they used their vast influence to sabotage Western economies with their green fantasies.
Only now that AI has forced them to grapple with the need for industrial levels of reliable energy, you know, the kind of energy manufacturing and industry needs, have silicon valley green stalwarts like Google finally admitted that battery firmed renewables alone are not fit for purpose.
If you need to back up renewables with fossil fuel, your energy system is an expensive failure. The cost of building and maintaining duplicated energy infrastructure, a renewable system which only works some of the time, and a fossil fuel system capable of providing 100% backup for when the renewables fail, which spends much of its time on standby, this cost far outweighs the cost of just building and maintaining the fossil fuel power system.
The reason duplicated power systems are so expensive is standby cost. Standby cost is the insurmountable burden which drives up energy bills in green energy dominated locales.
People who build and operate power stations expect the same profit regardless of how frequently their system is used. They have construction mortgages which have to be serviced, employees who expect to be paid full time, not just when they are needed, and equipment which has to be maintained and kept running on rolling standby, ready to jump in at a moment’s notice when the renewables fail. They don’t have to build and operate their power stations in places where they won’t be used full time – there is a long queue of paying customers who are desperate for their attention. This standby cost is what makes it impossible for a renewable energy system with fossil fuel backup to be cheaper than a fossil fuel power system.
It is not too late – even now Google and other major electronic frontier organisations could make an effort to undo all the damage they have done over the last few decades by pushing their green net zero fantasies.
But despite admitting they need reliable energy for their AI systems, that renewables + battery backup are not a viable solution for powering data centers, Google’s search engine and AI still delivers an over optimistic impression that renewables are the cheapest solution (see above), and search rank penalises or shadow bans climate and energy realist material from sites like WUWT.
We keep hearing radical greens claim that Exxon knew about the alleged climate crisis, but in my opinion what we have here is a case of Google knew. Google has known since at least 2014 that renewables are not a viable replacement for fossil fuel.
My question – which state or organisation will be the first to sue big tech for continuing to push green fantasies, when their own research people told them back in 2014 that renewables don’t work? How much taxpayer money has been squandered because ordinary people and politicians trusted Google search to provide unbiased search results?
What did Google know and when did they know it? (couldn’t resist)
#GoogleKnew
This was obvious to all who could reason no later than a couple of years ago when the techcos started talking about long term contracts with nuclear power companies while lying their asses off professing their commitment to renewables, and I commented so here on WUWT.
For why on earth would they have started to consider nuclear power if their engineers and economists had not already thoroughly analysed and evaluated the potential of renewables and come to the firm conclusion they were not up to the task?
And we should conclude that it’s responsible for much of the cost of the renewable scam?
Has anyone notified Sher Edling of this terrific opportunity for law suits?
“They pushed green fantasies instead of genuine energy solutions.”
Good little cult members …..
😉
And Google reached agreement with the UN to promote UN climate articles above all else.
Renewables alone are inadequate to power modern technological society without Fossil back-up to firm generation.
Renewables need 100% Fossil backup available 100% of the time.
Fossil generation needs 0% Renewable back-up to be available 100% of the time.
#GoogleKnew
Kind of off topic. I just had my first meaningful exchange with AI. I am looking for a full roof skin for my 1983 Pontiac Grand Prix. On one of the parts sites I soon realized I was communicating with some AI program. Its responses and searches were blindingly fast, almost instantaneous. It couldn’t help me, it came up with the same stuff I came up with doing an internet search. In the end it did what I have always thought it would do, exactly what the shop wanted it to do. I asked it for recommendations for places to look. It basically told me to look at places that sold auto parts. I’d rather deal with people. On a side note it also suggested I call a number for more help, I did and instead of typing with AI I was talking to him, he goes by Alan while typing and Al when speaking. Kind of novel but not very helpful.
AI can be more useful than that, but if you want to create a useful AI you need to pay real money for the development of that service.
Of course you are right Eric but that isn’t my point, my point is AI did exactly what it was programmed to do.
The AI tools (Gemini) I use are super search engines. You must require it provide not only the analysis and summation but provide the source used to obtain the information, along with the link to that source. I often require it to include the page number where I can see the original source. If the response is prejudiced or is inaccurate, I will call out the errors and provide other sources with links supporting my position or upload the source document. Most of the time Gemini will apologize when in error on numbers and content and reprocess. Even when (e.g., climate change and CO2 levels being beneficiary) it seems to be programmed to toe the line it often works. Actually kind of fund debating these tools.
The PJM grid is also having a wake-up moment–they don’t have enough reliable power right now to meet demand if they comply with permit restrictions on their fossil units. They don’t have enough renewable power and the artificial restrictions are putting them at risk of blackouts in the region. Yesterday, prices were in the $3000/MWH for hours.
More power to them. Sooner or later the poop hits the fan. At some place up and down the line someone will get the blame, as they should.
Sadly, probably not the people who deserve it. My niece who lives in the BGE service territory has been told to blame data centers for her high rates.
A group of data centers consume a fairly constant and large amount of electricity and are therefore favored customers of utilities. Power generators can maximize their production and therefore their profits. Loudoun County Virginia has about 250 such centers, is building more and receives about 40% of their county budget from the centers taxes. That enables the county to keep property taxes low and the power provider to keep electricity rates low, much lower than nearby Maryland where I live.
BGE household rates run from 15 to 21 cents per KwH. This is about the same as charged by SMECO in my Maryland area to the south. In my area, there are no data centers. In the BGE area there are 18 to 36 centers operating or planned yet our household power rates are about the same. Blaming data centers for BGE rates seems unfounded. I agree that SMECO and BGE rates are too high but the reason(s) for this is not data centers. Indeed, it seems that if BGE has the power available, more would be better.
They lied and people died. I’m sure somewhere.
Batteries that charge from cheap solar at midday and discharge each afternoon to bridge the gap between solar failing and other generation kicking in are profitable.
Who would build a battery required for longer term that is on standby and not earning each day for the owner?
Similarly, anyone who builds a gas turbine for backup when unreliables fail would need to be subsidised with a guaranteed income.
Its more or less the same dilemma which makes fossil fuel firmed renewables so expensive, except since its batteries add a few zeroes to the cost.
A very simple sum, BUT LONG DESCRIPTION. Let’s say that Google wanted to run a very small server rack with an electrical load of 10kW, (when you think that some gaming PCs run into the near kW range, then you get an idea of how small this is). I’ve used Oz$ in this assessment US$ may be a little less, or not.
It needs to run 24 hours per day, so the electrical demand, for anyone day, would be 240kWh.
Let’s say that solar can be installed and it has no upper limit, really overdo it, install a silly amount, say 100kWp, it really is that cheap. That way, the solar panels will be providing enough to run the server and charge the batteries even when the sun is only a little over 10% of it’s normal brightness. Cost around $1/W, so $100k
Now, look at a winter day, (even fully cloud covered and dark with rain pending), the solar panels will be providing so much power that they are NOT a restriction during daylight hours. Actual supply on a dark day is around 10% of rated capacity, so 10kWp, (just enough to run the server).
So, for an 8 hour day, you need to cover 16 hours of darkness, or 160kWh of energy. Since no one wants to be caught short, let’s use a nice round number of 200kWh of storage for ONE day. Triple it to allow for rare events and down time in the charging system. So 600kWh is required.
A typical rack mounted, 48V nominal battery provides 5kWh, so you’ll need 120 of them. The cost, allowing for bulk deals would be around $240k, add another $60k for installation and the a couple of inverters with switchgear, etc. So $300k gets a 3 day, off grid energy storage system.
So for ONE small server that costs $10k, you require storage and solar that costs $400k.
No wonder that Google says No.
And don’t mention the area required. Roof area = 100kW/150 = 667 m2 and the battery would take up 12 server racks, compare that to the single expensive PC/server, sitting in a single rack.
You’re going to need a bigger shed for the ancillary equipment. And that’s for ONE small server.
Absolutely – the fat tail is the killer. The cost of creating sufficient backup to match current reliability standards of around one day of outage every 10 years is unimaginable.
The Tesla Meagpack Battery holds 3MWh of energy similar to 30 Tesla 100KWh EV batteries.
Your 600MWh would take 6 Tesla 100MWh battery packs or 1/5 of a Megapack Battery.
Price?
About $1.5M (US) for a MegaPack, 1/5th of that would be $300k (US).
So my numbers look a little optimistic. I had $300k (Oz) for that amount of storage. But still good enough for a back of the envelope sum. Especially since the MegaPack has a weatherproof container, air conditioned, etc, etc. I was just looking at the batteries.
The Tesla EV battery packs cost $23,000 new so 6 would be $138,000
The Tesla EV pack is not really in a shape or voltage format ready for connection to a grid load. I could imagine that the cost of mounting those batteries and the inverters/chargers required would take you close to the rack mounted systems, (remember, I’m working in Oz dollars which are close to Pesos in comparison, US$138k is around Oz$230k). Not a long way off my numbers and you still have the need t make it mount it in a rack or other safe place.
Go let Google’s head of infrastructure know he got his sums wrong ;-).
Bryan got the outcome correct, there really is a need for a lot of batteries for this small server. Actually a disturbingly large battery when you consider the relatively low electrical demand.
How the hell could industry, (think electric arc furnace), ever go off grid/solar?
Fair point, my bad.
Something is wrong with your units, I think you dropped in MWh when you should have kWh, (twice), but you got the answer correct.
My calculated energy storage was 600kWh, that would be 6 of the Tesla EV batteries which is about 1/5th of the usable energy from a MegaPack.
Oops, correct that last statement should have been in KWh.
At $23,000 each, it would be cheaper to buy and stack 6 100KWh Tesla EV Batteries.
That 600KWh storage would cost $138,000 (Plus racking, installation and cooling)
One cannot overlook the other deficiencies of solar power. Yes, inverters convert the solar DC to AC but cannot respond to sudden changes in frequency or voltage as the load varies so some other machinery is needed. More $$$.
A server rack doesn’t need any of the ‘other machinery’, the load is going to be a converter that changes the mains voltage into low voltage high current. There are no massive spikes due to motor starts or similar.
In fact, the best system would be for a low voltage system from start to finish. It is much easier to convert a 48V system down to the 12V, 5V, 3.3V, 1.8V, (and others, all DC), than it is to first step it up to 230V AC, (or other, chose your country), and then have to convert it back to the DC voltages required.
There would still be a need for massive PC power supplies, (rated for the 10kW server), but you could supply them from a 48V busbar system relatively easily. And actually, (with regard to the lower voltage), more safely, just as long as you manage the potential arc flash that large busbars can ‘offer’.
Did you factor in the costs of the subsidies that reduce the price for the solar and battery? Those are significant.
Subsidies do not alter the cost. They alter who pays for which part.
Well said. And no subsidies are included above, unless they are included PRIOR to my pricing, eg the battery importer has claimed them.
And here is my view, as I built few battery solar systems by myself. I will count in Euro, should be similar enough to $.
10kWh system taking 240kWh daily, with price for electricity at 15 cents will consume 36 daily. This is 13140 per year. 65700 per 5 years.
As person you can get 30kWh battery LiFePO4 16 cells MB56 620Ah each with DIY box for around 3000. For 10kWh system, consuming 240kWh daily, I would recommend 1.5 day backup, e.g. 360kWh. That is 12 boxes for 36000.
You need one inverter capable of 10kW, that is around 2000 for Victron, 1000 for Axpert.
For panels, 1kW panels is in my climate capable to create 6kWh average during summer, 4kWh autumn, spring, 1kWh in winter.
So let’s go for fastest return, count with 4kWh daily per kWh of installed solar panel. That means you need 240kWh/4, 60kWp installed. That is roughly 100 of 600Wp bifacial panels, 100 per piece. (I bought four 580W 60Eur per piece for my father).
That is 10000.
You need 20 MPPT 3kW charge controllers for panel for 7000.
Another 10000 for construction, support, cables.
That is 63000.
For building I guess between 10000 – 20000, around 2 weeks for 2 skilled people.
So final number can be somewhere around 83000.
This with 100% save this is paid in 6.3 years.
If your climate is allowing only 75% save like mine, it will take 8.4 years.
Is it lot? I don’t know, I have solar system on my house, it has 4 years and it will pay off in 6 years.
Batteries are good for around 10 years.
It will not be 100% sufficient, you still need some backup from the grid.
“Battery technology is not yet at the stage where it can fully absorb that firming need and requirement”
When will battery technology be at the stage where it can fully absorb that firming need and requirement?
About the same time as the commercial fusion power plants are available 😉
10 years
Google said this a couple of years ago after researching it.
Yep – but they haven’t updated their optimistic public position. Its going to be a fascinating legal case when someone tries to squeeze some money out of them for presenting misleading information.
People invest their money and buy shares based on the promises made by company management and their press statements.
Some of these people may think that they had been told something which is clearly not the truth.
Legal cases have been won for a lot less. In Oz, a director that knowingly misrepresents a fact which could influence a financial decision can be held criminally liable. As a minimum they would be barred from ever being a director again if found guilty, at worst jail.
Google may want to get their financial statements updated and aligned real quick.
The big players are finally waking up….as far as AI goes… I’m a sceptic…..mainly in regard to the issue of safeguards……but when the big boys finally publish that they have woken up, you have to raise a glass to the fact that the green/woke fantasy is on it’s last legs.
The early industrial revolution was no different in terms of risk, steam boilers exploding was a significant cause of death – so much so it led to a British priest who was fed up with burying parishioners to design a much safer engine which doesn’t use steam.
Eventually the metallurgy caught up with the technology and the boilers stopped exploding so often. That safer engine is still in use today, steam engines not so much.
Do the clergy also work on fusion power?
Asking for a friend.
The clergy aren’t but AI is heavily involved in Fusion research and in the last two years was involved in all the big advances. It is learning far faster than humans can.
Yes, because prayers are needed to make it work.
People who push for wind and solar power plants discuss the need for “firming,” whatever that means. The word “firming” is most commonly used to describe “cosmetic skincare products or treatments designed to improve skin elasticity, reduce sagging, and create a tighter, more resilient appearance.” Use of the word when describing wind and solar power avoids the need to specifically state that some reliable on-demand source of electricity is required to accommodate their unreliability and erratic nature. That source usually is natural gas power plants because some such plants can accommodate rapid changes in demand. Coal, and even nuclear plants if appropriately designed, could provide such service as well. Batteries, often mentioned as a cure, cannot do the job.
Advocates also universally fail to mention that the cost of wind and solar power is not at all low because of the need for backup fossil fuel power to fill in the holes and accommodate sudden changes. The backup plants must be able to meet the full electricity demand whenever wind falls below about 10 mph or increases above about 35mph and when the sun il low or set which means they must be as powerful as they would be if meeting all demand at all times. The combination gives us one power system (wind, solar, backup) for the price of at least 2. And whether this combination actually reduces CO2 emissions is problematic. Variable and part load operation of the usual backup, natural gas turbines, greatly reduces their fuel efficiency. For example, at hot standby, as some turbines must be to meet the variations, as much as 40% of their full power full load will be consumed just to keep the machine at ready-to-go temperatures. This is one of the reasons why gas turbines have not been adopted for car or truck propulsion. Part and variable load operation of gas turbine power plants also increases maintenance costs.
“Orders for new natural gas fired power plants surged to 130GW in 2025, a 25 year high, with US data centre demand a major factor”
IEA ‘World Energy Investment 2026’ (May 2026)
My question is, when will the lawfare, “Google Knew” commence?
And then for how many other industries does it also not work for? When I look at Europe today, particularly France, which was in recent years declared the true workers paradise, and how it seems to be cratering before our eyes, and we can clearly see that they only need to fight back and the crisis will end. I mean really fight back; unapologetically hospitalize some people, crush the cell phones of the vultures crowding around and gleefully cheering the mayhem. Take the ring leaders and hang them, with only a military trial. It would end. A culture can continue to exist with one generation wiped from the slates, but it cannot continue when it cannot go about its daily functions.
.
Anyone who has worked on the operations and maintenance of a data centre, even 20 years ago, knew that battery technology is, and would only be, good enough as an emergency back-up, given the energy draw for typical processors. The 2003 blackout, and many power outages since, have confirmed that gas back-ups were what was needed to keep a large data centre humming. The only real difference with these new data centres is scale, and there still are no solar or wind facilities of scale and endurance to operate a 7/24 data centre.