For those of you who would prefer to listen:
Microsoft is building a massive data center in West Texas and Chevron will provide the power. That was announced this week. The deal is set for 20 years. The facility will be on 2,000 acres in the Permian Basin, the heart of America’s Oil & Gas industry. Increasingly, Tech companies are building facilities closer to energy sources as the AI advancement continues.
There has been serious push back across the country with rising costs and reduced energy supplies as the AI growth has put tremendous stress on existing electric grids. New energy sources are being developed. Nuclear is making a major comeback. Presently, Natural Gas is supplying much of the AI advancement. Microsoft will have a dedicated fuel source from Chevron, which will reportedly not impact the energy supplies for Texans living in the area.
This AI revolution is soaking up vast quantities of natural resources. The amount of energy and water required is massive. Some data centers require energy supplies capable of powering big cities. And they operate 24×7. Existing electrical grids weren’t designed for such needs. It’s creating conflict in communities. It’s showing up this election cycle. Tech companies are increasingly forced to bring their own power.
Elon Musk has been calling for data centers in space for a while now. As the Earth-based data centers face the growing pressures and controversies over its land use, massive water consumption, and spiking energy demands, it appears that space is a legitimate alternative. That said, it would come with some serious engineering challenges. Now I’m no astrophysicist, but my curiosity had me digging into the out-of-this-world possibility. My primary sources for information were studies done by the European Space Agency (ESA) & the Aerospace company Thales Alenia Space and the Hewlett Packard Enterprise (HPE) pilot program at the International Space Station. What I found was pretty interesting and quite enlightening.
Here on Earth, data centers require massive electric grids and backup diesel generators. In space, they say power is the easiest problem to solve. The reason? There is constant sunlight. A data center in low orbit can access near-continuous solar energy. It wouldn’t be affected by weather, atmospheric filtering, or the daily cycle of day and night on Earth.
Space is cold. The temperature can reach an average of -455 degrees Fahrenheit. That’s really cold. But space is also a vacuum. On land, data centers reject heat via convection. That’s done by moving air or water past the servers. In a vacuum, convection is impossible. In space, heat can only be dissipated through radiation. The data center structures in space would need sophisticated sunshields, similar to the Webb Telescope, to prevent the sun from cooking the servers while they try to shed internal heat.
To prevent the semiconductors from overheating, all excess heat has to be transported to large radiators. They are mounting servers on various satellites to liquid-cooled cold plates. This liquid would pump the heat away from the servers out to giant fins that are pointed away from the sun. They radiate the heat into the deep freeze of dark space as infrared light. It’s pretty remarkable stuff, even though I don’t completely understand it. One thing is clear: Cooling is one of the greatest engineering challenges for space data.
Maintenance would be much different in space. You can’t easily send an IT tech to swap out a dead hard drive in orbit. Hardware would have to be built in modular blocks. If a server rack malfunctions or dies completely, autonomous robotic arms would have to swap out the module. If the situation was more complicated, the entire satellite might have to be de-orbited to burn up in the atmosphere and then get replaced. That can be expensive. Prototype servers have already been tested on the International Space Station to prove that the hardware can survive the environment. Launched in 2024, HPE’s Spaceborne Computer2 is up there now.
Before space, the ocean has been a strategic testing ground for data centers. In fact, the ocean is increasingly being sought to solve the massive power and water constraints caused by the AI boom. Data centers require Millions of gallons of water every day to evaporate and keep server rooms cool. Underwater data centers flip the script. Many in High Tech say that the ocean is the perfect environment for data storage. Inside tightly sealed ocean pods, the air is replaced with bone-dry, pressurized nitrogen gas. Since there are no human oxygen levels, no moisture, and no dust, hardware failure rates underwater are proving much lower than they are on land.
Microsoft pioneered this concept in 2013 with its experimental Project Natick off the coast of Scotland. It was a scientific success in proving that subsea data centers had a much lower failure rate, were 8x more reliable, and had far greater energy and resource efficiencies. That said, Microsoft discontinued the project in 2024 as it determined the ocean pods weren’t capable of generating the massive amount of infrastructure required to drive today’s AI revolution. Regulation and environmental issues played a role, too. That has slowed the advancement. There are some startups still tinkering with the concept. A company called NetworkOcean is working on a pilot project in San Francisco Bay for autonomous, wave-powered data centers. It’s early days for this, still very much in the developmental and regulatory phases.
China has taken the lead in ocean-based data storage. The Shanghai Lingang Undersea Data Center just launched in May. It is the world’s first commercial underwater AI data center directly powered by offshore wind turbines. It is located 6 miles off the coast of Shanghai and is submerged over 30 feet deep. This underwater data center has four levels of server racks. It is currently operating at 2.3 megawatts, with a planned scaling capacity up to 24 MW. For perspective, 24 megawatts would provide ample power for over 20,000 homes on an average day.
The AI race continues at a rapid pace. Silicon Valley companies are competing with each other, while China continues to advance its own ambitions. The amount of money and energy thrown in this direction is just mind-blowing. Over $2.5 Trillion is expected to be spent on AI this year. That would mark a 47% increase from last year. Half of it is coming from the U.S. Current estimates for AI spending next year already exceed $3 Trillion. AI has been the biggest driver of the Stock Market as well as America’s Economy. It’s also been a major driver of stress and confusion. There’s little sign of that changing anytime soon. The implications are profound, all the way around.
Despite all the hype and spending, the AI-trade has tripped quite a bit. Correction continues to be the theme as June and Q2 come to a close. The Tech-heavy NAS has led the declines, recording its worst week in over a year. Both the NAS and S&P went into the weekend with 5-day losing streaks. Conversely, the more value-oriented Dow hit a fresh, all-time high this week. The equal-weight S&P is an important tell, having hit all-time highs this week too. Advancers were nearly 2 to 1 versus decliners. This price action shows there has been underlying strength beyond Tech. Smaller stocks were carrying the load while the larger, more index-impactful stocks sold off. This is basic rotation. Importantly, money has rotated within the Market, not out of the Market. That’s a good thing.
Digging in further, the AI-trade, which has dominated the Stock Market for the better part of 4 years has been split of late. They’re not all going up together. The companies that are aggressively spending, commonly called the hyper-scalers, have experienced a decent sized sell-off. Many are Tech Titans. Conversely, the recipients of the massive investments, primarily semiconductor companies, have seen their stocks continue to soar. Perhaps this correction is a healthy rotation, sending money into previously underperforming sectors that look cheap while more expensive, high-flying Tech stocks come back to Earth. That’s certainly what it appears to be, at least at this stage. But we all know that Summer tends to be volatile for Wall Street, and the Market was ripe for a correction. What’s more, these days, seemingly anything can happen. We know it. We’re ready.
Have a nice weekend. We’ll be back, dark and early on Monday.
Mike
