The sun threw a tantrum. On Tuesday, November 28, 2023, it unleashed a massive M9.8 solar flare. That is just one step below the most intense classification, X-class. The eruption originated from sunspot AR3500. It hurled coronal mass into space at speeds exceeding 800 kilometers per second.
This wasn’t just light. It was matter. Heavy, charged particles headed straight for Earth.
The Cannibal Storm Approach
NASA models indicate the resulting Coronal Mass Ejection (CME) will arrive on Friday, December 1st. The impact window opens around 11:00 AM Paris time. If the timing holds, skywatchers in France might catch auroras. Not in the Arctic. But at mid-latitudes.
Here is the complication. The sun didn’t just send one wave. It sent several.
This CME is likely to catch up to slower ejecta launched hours earlier. Astronomers call this a cannibal CME. The faster storm swallows the slower ones. The combined force could trigger a G3 geomagnetic storm. Researchers classify this as “strong.”
G3 storms happen. They aren’t rare anomalies. We see about 200 per 11-year solar cycle. But they are significant. They push auroral boundaries far south.
Why Should You Care?
Strong storms disrupt more than just the view. They stress power grids. They confuse satellite navigation. They create radio blackouts.
These solar eruptions threaten Internet infrastructure. How? And why does it matter?
The energy from the sun interacts with Earth’s magnetosphere. This induces currents in long conductors. Power lines. Undersea cables. Substations.
For the average user, the disruption might seem invisible. A dropped signal. A glitching map app. A flickering light.
But the underlying mechanics are brutal. The sun is not static. It breathes. It expands. It contracts. And right now, it is breathing out.
Does this mean your phone will die? Probably not. But the systems supporting your connection are vulnerable.
The question isn’t whether the storm hits. It’s when. And what breaks first.
The Hidden Cost of Our Digital Convenience
We are living through a paradox. Every swipe, search, and stream feeds the insatiable hunger of large language models. These systems promise us a world where information is instant, accessible, and almost magic. But behind the sleek interface of your favorite AI assistant lies a physical reality that is heavy, hot, and resource-intensive. The environmental impact of generative AI isn’t just a footnote in tech reports; it is a growing crisis that demands our attention.
The Energy Bill of Intelligence
Training a single large language model can emit as much carbon as five cars over their entire lifetimes. And that is just the training phase. The inference phase—the actual act of answering your question—adds up fast. A single query might trigger thousands of calculations across data centers that run 24/7. These facilities require massive amounts of electricity to power servers and even more to cool them down. Water is another critical resource, often drawn from local supplies to keep the machinery from overheating.
The scale is hard to grasp until you look at the numbers. Tech giants are investing billions in new data centers, many of which are located in regions already facing water scarcity. The question isn’t whether we can afford the technology. It is whether we can afford the footprint.
Who Is Responsible?
The blame game usually starts with the user. “I only asked one question,” you might think. But the responsibility is shared across the entire stack. Developers optimize code for speed, often at the cost of efficiency. Data centers prioritize uptime over sustainability. Consumers demand immediacy. No single actor can solve this alone.
Some companies are beginning to invest in renewable energy sources. Others are experimenting with liquid cooling systems that use less water. These are steps in the right direction, but they feel like band-aids on a bleeding wound. The real solution requires a fundamental shift in how we design and deploy these models. Smaller, more efficient models that require less data and less compute could offer similar performance with a fraction of the environmental cost.
The Efficiency Trap
Efficiency improvements often lead to increased consumption, a phenomenon known as the Jevons Paradox. As AI becomes cheaper and faster to run, we use it more. More queries. More complex applications. More devices. The net result? Total energy use goes up, even if individual tasks become more efficient. We are caught in a loop where technological progress drives greater demand, which in turn requires even more infrastructure.
This isn’t just an environmental issue. It is an economic one. The cost of energy is rising. Water prices are fluctuating. Supply chains are fragile. The current trajectory is unsustainable, not just for the planet, but for the industry itself. If we want AI to remain a viable tool for innovation, we need to break the cycle.
What Can Actually Change?
Regulation is starting to catch up. The European Union’s AI Act includes provisions for transparency regarding energy consumption. This could force companies to disclose the carbon footprint of their models, making inefficiency a reputational risk. In the U.S., voluntary standards are emerging, but they lack teeth.
Consumers can also play a role. Choosing platforms that prioritize sustainability. Supporting open-source models that are often more transparent and efficient than closed, proprietary systems. Even small changes in behavior, like batching queries or using offline features, can add up.
But the real leverage lies
