Let’s face it, everyone is obsessed with buying Nvidia chips and discussing the next big ChatGPT update. But almost no one is paying attention to the messy, physical side of this tech boom. At the G20 technology forum, Elon Musk pointed out a massive reality check that most retail investors are completely ignoring.  

According to Musk, by 2027, about 15 gigawatts of AI computing power could be completely useless. Why? Because we simply won’t have enough electricity to power them. To give you some context, 15 GW is essentially what ten large nuclear reactors produce. We’re looking at a future where billions of dollars’ worth of cutting-edge equipment is just sitting in warehouses, collecting dust because there’s nowhere to plug it in. Musk described it as a collision of two completely different growth rates. On the one hand, the production of AI equipment is growing at a blistering pace of about 40-50% per year. On the other hand, actual grid capacity outside of China is barely growing at 10-20% per year. The faster trend will inevitably wipe out the slower one. And to be fair, he’s not the only one panicking about this. Wall Street has been crunching the numbers lately, and the consensus is pretty grim. Goldman Sachs recently reported that data center power demand in the U.S. is likely to double by 2027. They also noted that historically, nearly half of all planned grid expansions have been delayed. Morgan Stanley predicts a 38 GW deficit in the U.S. by 2028. If AI were a country in its own right, some forecasters believe it could gobble up enough energy by 2035 to become the fourth-largest electricity consumer on the planet, behind only India. So how will the industry get out of this mess? First, big tech companies are already moving into nuclear power. Companies are literally trying to buy up nuclear power plants or invest in small modular reactors to power their data centers directly, bypassing the old public grid. 

Second, smart money is moving away from blindly chasing Nvidia. Investors are moving to companies that build the really boring infrastructure — high-voltage transformers, cables, and cooling systems. For example, GE Vernova already has a huge backlog of orders worth over $176 billion just for grid upgrades. Ultimately, this could be the biggest catalyst for DePIN crypto networks like Render or Akash. When tech giants can’t get local permits to build a massive power plant through government bureaucracy, decentralization becomes the only logical Plan B. If you can distribute heavy AI workloads across thousands of smaller global nodes instead of one giant data center, you bypass local power shortages entirely. We’re officially moving from the era of pure software to a time of brutal physical constraints. Access to electricity is now becoming far more valuable than silicon itself.   What do you think about this? Let me know in the comments :)