Musk's Colossus data center for Grok is at the centre of an environmental row over air quality in South Memphis

The Grok website on a laptop computer arranged in New York, US, on Wednesday, Nov. 8, 2023. Elon Musk revealed his own artificial intelligence bot, dubbed Grok, claiming the prototype is already superior to ChatGPT 3.5 across several benchmarks.
(Image credit: Gabby Jones/Bloomberg via Getty Images)

If you've been keeping your finger on the pulse of AI developments, you'll know that there's growing concern about meeting the power demands of the data centers that train and run AI inference. Elon Musk's Colossus supercomputer that powers Grok, xAI's alternative to ChatGPT, tackles this by using over 30 methane gas turbine stations, but the construction has incurred fierce criticism over its impact on local air quality.

When Colossus fired up for the first time last year, it was already making headlines. Partly because it took a mere three months to build, but also because its appetite for water and electricity was raising eyebrows. In the case of the latter, the system uses anywhere between 50 and 150 MW of power, and in order to meet that demand, xAI installed a number of methane-burning gas turbines.

A report by the Southern Environmental Law Center (via Fudzilla) claims that these turbines were installed without permits and raises serious concerns over the potential impact these machines will have on the local air quality, due to their emissions. To make matters worse, it appears that xAI is now retrospectively applying for permits to use the turbines constantly.

Memphis news channel WREG writes that the mayor of Memphis, Paul Young, had argued that things weren't as bad as they seem, quoting him as saying, "There are 35 [turbines], but there are only 15 that are on. The other ones are stored on the site."

However, thermal camera footage taken by the Southern Environmental Law Center (SLEC) shows, for that moment in time at least, 33 of the turbines generating large amounts of heat—in other words, they were in heavy use.

Annotated photographs of the COLOSSUS electronic digital computer - The National Archives (United Kingdom).

The original Colossus 'super' computer wasn't quite as energy hungry as the new one (Image credit: The National Archives (United Kingdom), FO850/234.)

SLEC goes on to criticise xAI's approach to the Colossus project: "From the beginning, the company operated with a stunning lack of transparency that left impacted communities in the dark. Even many Memphis city officials were unaware of the facility’s plans and how it would be powered."

One might think that using fossil fuels is a poor choice—not just because of the environmental impact but also because of the rapid rise of renewable systems—but with President Trump rolling out a raft of executive orders favouring the return to fossil fuels, it's perhaps not hard to see why xAI chose gas turbines over anything else. However, it's unlikely to be a long-term solution.

There is no way to avoid the huge energy demands of any data center, and it's a problem that will only get worse, as more systems come online to meet the AI growth targets that Google, Meta, OpenAI, xAI, and Microsoft all have. Musk hopes to have Colossus expanded from 200,000 up to one million GPUs, and no amount of gas turbines can realistically hope to cover that.

So, xAI will have to rely on the local electricity network and battery storage systems to meet that level of demand. But that passes the problem of generating the electricity onto somebody else, and they may well resort to using fossil fuels, even if xAI doesn't.

AI, explained

OpenAI logo displayed on a phone screen and ChatGPT website displayed on a laptop screen are seen in this illustration photo taken in Krakow, Poland on December 5, 2022.

(Image credit: Jakub Porzycki/NurPhoto via Getty Images)

What is artificial general intelligence?: We dive into the lingo of AI and what the terms actually mean.

If you don't use or have no interest in Grok, you might think that this has no real impact on PC gaming. However, it's worth noting that AMD, Intel, and Nvidia are all heavily invested in building and using data centers to train and run AI inference for their graphics technologies. In the case of the latter, Team Green ran such a system flat-out for six years, just to improve DLSS.

While Nvidia's center is unlikely to have anything near the same energy demands as Colossus, it's a reminder that the cost of sustaining the growth of AI is more than mere dollars. Energy and the environment have a sizeable share of the bill, too.

TOPICS
Nick Evanson
Hardware Writer

Nick, gaming, and computers all first met in 1981, with the love affair starting on a Sinclair ZX81 in kit form and a book on ZX Basic. He ended up becoming a physics and IT teacher, but by the late 1990s decided it was time to cut his teeth writing for a long defunct UK tech site. He went on to do the same at Madonion, helping to write the help files for 3DMark and PCMark. After a short stint working at Beyond3D.com, Nick joined Futuremark (MadOnion rebranded) full-time, as editor-in-chief for its gaming and hardware section, YouGamers. After the site shutdown, he became an engineering and computing lecturer for many years, but missed the writing bug. Cue four years at TechSpot.com and over 100 long articles on anything and everything. He freely admits to being far too obsessed with GPUs and open world grindy RPGs, but who isn't these days? 

You must confirm your public display name before commenting

Please logout and then login again, you will then be prompted to enter your display name.