AI: AI Data Center Power ramps and 'Bad Harmonics'. RTZ #584

...unintended consequences of the trillion plus dollar, multi-Gigawatt AI data center builds

The Bigger Picture, December 29, 2024

A core theme I’ve discussed a lot in 2024, is the accelerating pace of power demands of ever larger AI data centers. As potentially trillions are invested this decade to drive ever scaling AI models, both large and small. The demands for energy, nuclear and otherwise, are unprecedented in this AI Tech Wave. Taking the utility industry that used to barely grow in the low single digit percentages for decades, into possibly double digit growth of 10%++ annually for this decade and beyond.

This demand, besides the billions in new ‘Additionality’ impact of AI investments, has an unexpected impact on power for non-AI uses. That is the Bigger Picture I’d like to unpack today.

Bloomberg lays it out in a detailed piece, “AI Needs So Much Power, It’s Making Yours Worse”:

“AI data centers are multiplying across the US and sucking up huge amounts of power. New evidence shows they may also be distorting the normal flow of electricity for millions of Americans. This map shows readings from about 770,000 home sensors, with red zones indicating areas with the most distorted power.”

“The problem is threatening billions in damage to home appliances and aging power equipment, especially in areas like Chicago and "data center alley" in Northern Virginia, where distorted power readings are above recommended levels.”

“An exclusive Bloomberg analysis shows that more than three-quarters of highly-distorted power readings across the country are within 50 miles of significant data center activity. While many facilities are popping up near major US cities and adding stress to already fragile grids, this trend holds true in rural areas as well.”

“Every day, Americans reach into their refrigerators or turn on their dishwashers without much thought given to the electricity flowing through their homes. But a hidden problem now threatens these seemingly mundane tasks: distorted power supplies.”

Add a new term to our AI industry vocabulary:

“The term for the issue is “bad harmonics.” It may seem a bit esoteric, but you can think of it like the static that can be heard when a speaker’s volume is jacked up higher than it can handle. Electricity travels across high-voltage lines in waves, and when those wave patterns deviate from what’s considered ideal, it distorts the power that flows into homes. Bad harmonics can force home electronics to run hot, or even cause the motors in refrigerators and air conditioners to rattle. It’s an issue that can add up to billions of dollars in total damage.”

“More importantly, bad harmonics are symptomatic of much deeper problems that are engulfing the US grid.”

That should get our attention.

“Distorted waves are just one measure of broader power quality. When homes experience good, or stable, power quality, it means that the flow of electricity for lights and appliances is being delivered at an even and predictable pace. The worse power quality gets, the more the risk increases. Sudden surges or sags in electrical supplies can lead to sparks and even home fires. Left unaddressed, one problem can morph into another. That means the bad harmonics of today can be a sign of potential disaster down the road.”

Here’s how it’s all measured, tracked, and tabulated.

“Harmonics are a pretty good canary in the coal mine for early signs of stress and problems,” said Bob Marshall, chief executive officer of Whisker Labs Inc.”

“Whisker Labs tracks power quality in real-time using roughly 1 million residential sensors, which are spread so widely across the country that nearly 90% of US homes are within half a mile of one. A Bloomberg analysis of exclusive sensor data coupled with data from DC Byte, a market intelligence firm, showed a strong link between proximity to data centers and worsening harmonics.”

“More than half of the tracked households showing the worst distortions of power quality are located within 20 miles of significant data center activity, according to the analysis, which covered readings from February through October. US census figures show that about 3.7 million Americans live in the most-impacted areas.”

And there are long-term implications for grids here and abroad.

“Experts have been warning for some time now about the impact data centers will have on power grids across the globe. The AI boom has only underscored the issue: The digital economy is sucking up so much power that demand is now straining available supplies of electricity in many parts of the world, leading to concerns over price increases and even widespread outages. And that’s only projected to worsen as more data centers are built.”

“The new harmonics data shows how these problems are already starting to play out in real time across the US.”

“Read More: AI Is Already Wreaking Havoc on Global Power Systems”

“It’s an issue that goes beyond just whether or not there’s enough power to flip the lights on. Distortions mean that even as electricity is flowing to homes, the quality can be eroded enough to destroy appliances and increase vulnerability to electrical fires if there’s a voltage surge. Poorer power quality overall can also eventually lead to lights flickering along with brownouts and blackouts.”

“We need to understand those risks,” said Hasala Dharmawardena, a senior engineer of power systems modeling studies at the North American Electric Reliability Corp., where he is part of a team looking at data centers.”

“The grid has never faced the kinds of strain that comes with data centers. These city-sized users can pop up very quickly, within a year or two, which is much faster than grid planning usually happens. Even during population booms, the rise in power demand paled in comparison to the expected installation in the coming years of hundreds, perhaps thousands, of these facilities to power AI. That stress is adding to problems of aging infrastructure, extreme weather and the electrification of more parts of everyday life, such as the rise of electric vehicles.”

The scale of acceleration is part of the issue.

“It’s especially important to understand the power system impact from AI “because it is such a big hammer” on the grid, Dharmawardena said. “The data center is a very large load. Take your house and increase that by 10,000. That is the difference between your house and a data center.”

It’s a US problem for now, given US leadership in AI to date.

“The US today is by far the largest operator of data centers in the world, with Northern Virginia’s hub boasting more than twice as much operational capacity as Beijing, its next-biggest rival, according to estimates from Bloomberg Intelligence. But other countries are racing to build out their own facilities, with growth expected in nations including Saudi Arabia, Ireland and Malaysia, which will all face their own pressures on domestic power systems.”

“The problems in the US are compounded by the fact that not enough investment has gone into the grid to fortify it for the coming demand boom. For decades, US power use was largely flat. Now, it’s about to be turbocharged. The nation’s demand for electricity will surge almost 16% over the next five years, more than triple the estimate from a year ago, driven largely by new data centers, according to a recent report from Grid Strategies, a DC-based consulting firm.”

“The increase means that without major improvements to the grid and power equipment, harmonics issues seen today are likely to get worse.”

“Some of the biggest data-center clusters are near big cities, so that the facilities can tap into larger grids and the fiber networks that are often located close to consumers for latency issues. That underscores why harmonics are often worse in urban areas. But while the effect appears to be more severe where population density is higher, Bloomberg’s analysis shows that even in rural areas, sensors that are closer to significant data center activity are more likely to have distorted power.”

Part of the issue is of course the diversification of US power grid due to climate change, growth of EVs and other priorities.

“Other factors, like solar energy generation, the deployment of EVs and swings in industrial loads can all contribute to irregular wave patterns. Still, a Bloomberg analysis found that sensors with more distorted power weren’t necessarily closer to large-scale solar installationsl.”

The northeast in particular as the most acute issues:

“Almost nowhere are the correlations between data centers and bad harmonics as clear as in an area of Northern Virginia that’s been dubbed “data center alley,” the global center of the industry. The area is mostly located in Loudoun County, outside of Washington, DC, which saw its data-center capacity increase by 2% in 2024 to about 3,000 megawatts (MW).”

“Nationally, the data analysis showed that roughly 1.7% of sensors in the average county had at least one monthly reading that exceeded the 8%-threshold for bad harmonics. That share was more than four times higher in Loudoun County.”

““Much of “data center alley” is served by Dominion Energy Inc.’s Virginia utility.”

The industries are of course cycling through possible ways to address this rising set of issues. As well as leading tech companies like OpenAI et al:

“Some solutions are already in the works.”

“Most new data centers in Virginia require their own substation and transformer to be able to come online and this isolates them from nearby distribution circuits, limiting their impact on other customers, Dominion spokesman Ruby said. The Virginia utility is building a new transmission line into Loudoun County to address power quality and reliability constraints overall. Devices like filters and capacitors can help address harmonics issues around data centers.”

“But still, there’s a big hurdle when it comes to measuring the problem, which thus makes it challenging to address, according to more than half a dozen experts interviewed by Bloomberg News. Most utilities don’t have the capacity to track on the residential level because it would be too costly, said Dharmawardena of the North American Electric Reliability Corp.”

Then there’s the issue of communication and coordination across fragmented energy ecosystems and jurisdictions.

“A NERC task force focused on modeling data centers and other big users plans to release an analysis in 2025 for impacts on the transmission system. Oversight of local distribution systems under stress often fall to state regulators and utilities.”

“Dharmawardena compared bad harmonics to trash in the middle of the road, which can impact all drivers in the area. And the Whisker Labs data, he said, gives you a sense of “the overall condition of the road itself.”

The whole piece is worth reading in full, especially for the detailed charts and graphs accomanying the ebb and flow of this issue as the trillion plus dollar AI data center buildout accelerates in this AI Tech Wave.

There are always unintended consequences, both positive and negative in every tech wave, many times in industries that previously unaffected. Kind of like how PCs and spreadsheets in the 1980s, fundamentally up-ended Wall Street and finance, turbo-charging financial markets private equity firms, and of course institutionalized investment vehicles like mutual funds and ETFs.

And this one is not different. How Power is generated and consumed in the US and elsewhere is fundamentally changing. And that is the Bigger Picture to keep in mind as we move into 2025. Stay tuned.

(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here)