<div id="readability-page-1"><div><div><div><div><p>In America, surging electricity demand from data centres used for artificial intelligence has sparked a scramble for gas-fired power – and even the biggest manufacturers can’t keep up.</p><p>It’s created a gap in the market for the next best option: repurposed jet engines.</p><p>This week, Boom, a supersonic jet start-up building a modern Concorde, unveiled a modified version of its Symphony engine that is designed to be able to be deployed quickly to data centres across the US.</p><p>“Supersonic technology is an accelerant – of course, for faster flight, but now for artificial intelligence as well,” said Blake Scholl, Boom’s chief executive.</p><p>His company is the latest to cash in as technology giants, including xAI, OpenAI, Google, Amazon and others, scramble for power to supply their ever-growing needs.</p><p>Data centres used to train artificial intelligence software consumed just under 200 terawatt-hours of electricity in the US last year.</p><p>But that figure will rocket to around 600 terawatt-hours by 2035, according to the International Energy Agency (IEA), potentially taking their share of total American power consumption from 4pc to more than 10pc.</p></div></div><div><div><p>Much of this will have to come from “behind-the-meter”, or off-grid, generation because the American grid is too congested to handle so much extra demand, particularly as it is operational 24/7.</p><p>This is what has led several tech giants to explore nuclear power to meet their needs.</p><p>Yet with the exception of recently retired plants that can be brought back online, nuclear cannot be deployed at the speed that Silicon Valley billionaires such as Elon Musk and Sam Altman require right now.</p><h2>Gas guzzlers</h2><p>It means that, at least in the short-term, companies are piling into gas instead. The IEA estimates that around 260 terawatts, or two fifths of the total power consumed by data centres, will come from gas-fired turbines in a decade.</p><p>But two thirds of the world’s gas turbines globally come from just three major manufacturers: GE Vernova, Siemens Energy and Mitsubishi.</p><p>They have already seen orders for large turbines surge to their highest level since 2011 and are struggling to keep up with backlogs that now extend back five years or more, according to experts.</p></div></div><div><div><p>These bottlenecks have prompted increasingly desperate tech companies, worried about falling behind in the AI race, to settle for converted jet engines, or “aeroderivative” gas turbines, to power their facilities instead.</p><p>For example, Altman’s OpenAI, the company behind ChatGPT, has reportedly acquired 29 GE Vernova LM2500XPRESS turbines, an aeroderivative model, for its planned “Stargate” data centre in Abilene, Texas.</p><p>Together, these would have a combined capacity of nearly one gigawatt – roughly equivalent to a nuclear reactor.</p><p>Boom is also getting in on the action with the 42-megawatt Symphony engine. The company says it has agreed to supply 29 turbines to data centre developer Crusoe and already has a $1.25bn (£940m) order backlog.</p></div></div><div><div><p>Scholl says he’s been “flooded with requests” for the equipment, and Boom is now drawing up plans for a “superfactory” to churn out several gigawatts’ worth of turbines.</p><p>Yet while this kind of kit is fast and easy to deploy, it is far less efficient than a purpose-built combined-cycle gas turbine (CCGT) for power generation. These recycle hot air to drive a secondary turbine to gain extra efficiency, hence the name.</p><p>According to GE Vernova, the jet-derived turbines used by OpenAI typically have an energy efficiency of about 40pc, compared with upwards of 64pc for the company’s 9HA CCGT turbine.</p><h2>Colossal problem?</h2><p>Over time, that means spending vastly bigger sums on fuel than would otherwise be necessary – and belching out more carbon emissions as well.</p><p>This is one reason why xAI, the AI business founded by Elon Musk, has been embroiled in a dispute with residents in South Memphis, Tennessee, over its “Colossus” data centre.</p><p>Outside the facility, a row of 15 fixed gas turbines is lined up and provides power 24/7. But campaigners in the Southern Environmental Law Centre (SELC) say Musk’s company installed the plants without permission and are seeking to block him from securing retrospective approval.</p><p>They have taken the local authority, Shelby County Health Department, to court and argue that the turbines “pump out smog-forming pollution, nitrogen oxides or NOx, and harmful chemicals like formaldehyde” into the atmosphere.</p></div></div><div><div><p>It could put residents at higher risk of asthma, heart disease and cancer, SELC says.</p><p>“Our health shouldn’t be threatened at the hands of billionaires who circumvent the law,” Derrick Johnson, the group’s president, said recently.</p><p>The controversy looks only set to grow. Musk has set out plans for the Colossus site to more than triple its consumption from just under 300 megawatts to one gigawatt – an increase that will need to be at least partly powered with more gas. xAI has also said some will come from solar farms.</p><p>“I can’t emphasise enough, we need more electricity,” Musk said two years ago. “However much you think we need, it’s more than that, I assure you, and we need it as fast as possible.”</p><p>That is, unless the rush into AI turns out to have something else in common with supersonic jets: hot air.</p></div></div></div></div></div>

OpenAI

<p>Tech giants in America are turning to repurposed jet engines to meet the surging electricity demand for AI data centers. Boom, a supersonic jet start-up, unveiled a modified Symphony engine for quick deployment. AI data centers consumed nearly 200 terawatt-hours last year, expected to rise to 600 terawatt-hours by 2035. Gas-fired turbines are being used due to grid congestion, with companies like OpenAI acquiring aeroderivative turbines. However, these turbines are less efficient and more polluting, leading to environmental disputes.</p>

- The surge in electricity demand from AI data centres in the US has led to a reliance on gas-fired power and repurposed jet engines due to supply constraints from major turbine manufacturers.  
- Companies like Boom are rapidly developing modified jet engines to meet the urgent power needs of tech giants, while the International Energy Agency projects a significant increase in electricity consumption by these data centres.  
- However, the use of less efficient jet-derived turbines raises concerns over higher fuel costs and increased carbon emissions, prompting legal disputes over environmental impacts in local communities.

Power-hungry tech giants harness jet engines to fire up AI