𝗧𝗵𝗶𝘀 𝗠𝗮𝘆 𝗯𝗲 𝘁𝗵𝗲 𝗕𝗲𝘀𝘁 𝗜𝗻𝗱𝗶𝗰𝗮𝘁𝗼𝗿 𝗼𝗳 𝗔𝗜’𝘀 𝗙𝘂𝘁𝘂𝗿𝗲 𝗣𝗼𝘄𝗲𝗿 𝗡𝗲𝗲𝗱𝘀

A🧡435-word💛sub💚2.5-minute💙read.

I’ve lost count on how many articles I’ve read on the future power needs of AI. Every research firm and many utilities have weighed in, any naturally so. That’s what they do.

The real answer is that nobody knows. The market is in its early stages and as such, the technology is evolving quickly. More efficient chips and data centers designs are in the works. When those efficiencies are garnered will play a major role in determining how much power will ultimately be required.

Another complicating issue was pointed out by 𝗣𝗲𝘁𝗲𝗿 𝗞𝗲𝗹𝗹𝘆-𝗗𝗲𝘁𝘄𝗶𝗹𝗲𝗿, an electric industry guru. He noted that there are many “zombie” data center interconnection applications, meaning that companies submit multiple applications in the hope of landing one viable one. This strategy tends to create an upward forecasting bias.

Here’s a sample of forecasts:

▶ 𝗘𝗹𝗲𝗰𝘁𝗿𝗶𝗰𝗶𝘁𝘆 𝗶𝗻𝗱𝘂𝘀𝘁𝗿𝘆 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝗳𝗶𝗿𝗺 𝗘𝗣𝗥𝗜: forecasts an annual growth rate of 3.7% to 15% between 2023 and 2030. Wow, that really nails it. It’s somewhere between a lot and really a lot.

▶ 𝗚𝗼𝗹𝗱𝗺𝗮𝗻 𝗦𝗮𝗰𝗵𝘀 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵: estimates data center power demand will grow 160% by 2030, and that worldwide consumption will increase from 1-2% of overall power today to 3-4% by 2030.

▶ 𝗜𝗻𝘁𝗲𝗿𝗻𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗘𝗻𝗲𝗿𝗴𝘆 𝗔𝗴𝗲𝗻𝗰𝘆: the IEA projects a 4% annually rise in global electricity demand through 2027.

▶ 𝗚𝗿𝗶𝗱 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝗶𝗲𝘀: believes U.S. electric demand could rise 456% over the next five years.

You get the idea.

Putting the professional forecasters aside, a recent 𝙐𝙎𝘼 𝙏𝙤𝙙𝙖𝙮 𝙉𝙚𝙩𝙬𝙤𝙧𝙠 article may have hit upon a better indicator of future demand.

The article noted that some of the world’s major gas-turbine manufacturers are not scrambling to ramp manufacturing.

There’s talk about nuclear powering data centers, but that will take a decade to materialize. Others say solar + storage will do the trick. Call me a skeptic. That leaves natural gas as the primary fuel source for the next decade or so.

So why are the companies that stand to directly profit from the AI boom not falling over themselves to increase capacity, especially considering that capacity is currently sold out? Additionally, in the last year, lead times have doubled and orders placed today aren’t likely to be delivered any earlier than 2028.

The answer is that the gas turbine industry is worried that the power demand boom - driven by AI and the desire to electrify everything - could go bust sooner rather than later.

What do they know that we don’t? Maybe nothing, and maybe the industry simply has a conservative bent, but it’ something worth watching.

#ai #electricgrid #naturalgas #gastrubines #gasturbine #datacenters #nuclearpower #nuclearenergy #solarplusstorage