Where AI Goes When Earth Fills Up VIEW IN BROWSER  | BY KEITH KAPLAN
CEO, TRADESMITH |
The world’s biggest data center hub has hit a wall. Northern Virginia has the largest concentration of data centers on Earth. About 70% of global internet traffic passes through the facilities in this region at some point in its journey. For years, it was the ideal place to build: - Power was plentiful.
- Fiber networks were dense.
- Land was available.
- Permitting was predictable.
Then AI came along and broke the model. AI workloads are 5 to 10 times more power hungry per server rack than traditional enterprise computing. A single large AI data center can draw 100 to 300 megawatts of power – about the same amount as a city of 80,000 to 250,000 people. These facilities also need vast amounts of water. A large AI data center can consume 3 to 5 million gallons of water a day for cooling – about the same amount as a town of 30,000 to 50,000 people. That water has to be piped in, circulated, treated, and cooled again – using even more electricity. By 2023, utilities in Northern Virginia began warning that new data center projects were arriving faster than the power grid – and local water infrastructure – could support them. Some developments were delayed. Others stalled. Local governments started pushing back. And the same pressures are also showing up in major data center hubs in Dublin, Amsterdam, and Singapore. Most investors think AI is a software story. And in part, it is. But it’s increasingly an infrastructure story – one constrained by power, water, and geography. And those constraints are already starting to bite. Today, I’ll show you why some of the world’s biggest tech firms are testing a radical solution to this problem – not here on Earth but hundreds of miles above our heads in what’s known as low-Earth orbit. It’s the band of space where satellites circle the planet every 90 minutes or so. I’ll also recommend one stock that gives you exposure to this future without needing it to arrive on schedule. As you’ll see, this isn’t about betting on science fiction. It’s about understanding what happens to computing when Earth’s resources hit their limits – and owning a company positioned to win, whether that shift happens fast, slowly, or not at all. | Recommended Link | | | | The AI revolution is dead in the water without nuclear energy. And Luke Lango says the most important player in this drama hasn’t made their move yet. The United States Government. Luke believes the White House is about to take an unprecedented step: A direct equity stake in a specific U.S. nuclear company. This company holds the key to deploying nuclear power fast – years ahead of traditional plants. Click here to get the details on the company Luke believes is next in line for a government windfall. |  | | Space Is Vast… So Are Its Resources Strip a data center down to first principles, and only two requirements really matter: power and cooling. Everything else – land, permits, water rights, grid access – are byproducts of those two needs. On Earth, both are becoming binding constraints. Electricity has to be generated and transmitted on enormous scales. Cooling requires vast quantities of water. And every new facility has to be built somewhere – parcel by parcel, grid connection by grid connection, permit by permit. In space, those constraints look very different. Above Earth’s atmosphere, solar energy is constant and intense. Solar panels in orbit get roughly 30% more solar energy than panels on the ground. And depending on their orbit, satellites can remain in sunlight nearly 24 hours a day. There are no clouds, no nights, and no transmission losses from distant power plants. Cooling works differently in space, too. On Earth, data centers rely on water to carry heat away from chips. In orbit, there’s no surrounding air to trap heat in the first place. Excess energy can be radiated continuously into the vacuum of space. The result is a cooling environment that’s far more efficient than anything available on the ground. There’s also a networking advantage for certain workloads. A growing share of global data is already generated in orbit – by Earth-imaging satellites, communications networks, and defense systems. Today, much of that data is transmitted down to Earth for processing, then sent back up to guide systems operating in space. Processing some of this data closer to where it’s created reduces bandwidth needs, shortens transmission delays, and lowers costs. None of this means all AI computing is moving off Earth. Training massive models will still happen in terrestrial data centers. But the everyday work of AI – sorting data, analyzing images, making decisions – doesn’t always need to live in the same place. As power, cooling, and geography become limiting factors on the ground, space becomes the next logical extension of the Earth-bound computing stack. This Is No Longer Theoretical Over the past year, some of the world’s most advanced tech companies have started testing whether AI systems can operate hundreds of miles above Earth – using real hardware, real software, and real workloads. Last year, for example, a Google-backed startup called Starcloud launched an experimental satellite carrying an Nvidia GPU and successfully ran modern AI models in orbit. The system processed real data in space, analyzed it using AI models, and sent results back to Earth – proving that today’s AI software can run off-planet without being tethered to a traditional data center. Nvidia has been running similar tests. In separate orbital experiments, teams flew its H100-class chips into orbit and demonstrated live AI workloads operating in space. The hardware survived launch, functioned reliably in orbit, and performed the kinds of tasks that dominate everyday AI use – image analysis, pattern recognition, and decision-making. There are still plenty of problems to solve. Cooling systems don’t scale effortlessly in space. Radiation shortens hardware lifespans. And the time it takes data to travel back and forth limits which kinds of work make sense off-planet. But those are engineering challenges – not physics problems. And one of the companies best positioned to navigate these challenges is Alphabet (GOOGL), Google’s parent company. Alphabet isn’t a space company. It doesn’t need orbital compute to work as an investment. What it does have is something far more important: control of the full AI stack. - Alphabet designs its own AI chips.
- It runs one of the world’s three hyperscale cloud platforms.
- It builds and deploys leading AI models.
- And it operates services – mapping, imagery, communications, and analytics – that naturally benefit from data generated in and around space.
If more computing stays on Earth, Alphabet keeps winning. If some workloads move closer to orbit, as I expect them to, Alphabet doesn’t need a new business model – it simply extends the one it already runs. And taken together with the other space-economy opportunities we explored in Part 1 and Part 2 of this series, Alphabet gives you a way to approach this theme without needing the space economy to arrive on a precise timeline. All the best, 
Keith Kaplan
CEO, TradeSmith P.S. For a limited time, TradeSmith Daily readers can enter GOOGL, NVDA… or any other ticker to fill out your Seasonality calendar for 2026. In my Prediction 2026 event earlier this week, I invited readers to try out TradeSmith’s new Seasonality tool and walked through why these cycles persist… how seasonality has continued to work even through crashes and bear markets… and how to use it to better manage risk. I also highlighted how our team combines seasonality with other technical signals to find higher-probability setups – and how to use it to step aside when the odds turn against you. Test drive our software tool and check out the following features: - Our new Seasonal Statistics page, which lets you dive into monthly, weekly, and even daily seasonal data to hone your trading edge
- An updated Seasonality algorithm that helps you find the most optimal trades
- A brand-new 5-stock portfolio strategy designed to keep you in the best patterns in an easy-to-follow way
The presentation will only be up for a few more days. So if you haven’t seen it yet, check it out here. | 
Post a Comment
Post a Comment