Musk Built the "Fastest Supercomputer" While Trump Throws $500B at AI

So Musk apparently built what Nvidia's CEO calls the "fastest supercomputer on the planet" in Memphis. Cool story. Jensen Huang said it, which means it's probably fast at running Nvidia's chips, shocking nobody.

Meanwhile, Trump decided to throw $500 billion at AI with something called the "Stargate program." Because apparently we need to be the "AI capital of the world" now. They're claiming this could create a $20 trillion AI economy by 2030, which is the kind of made-up number that sounds impressive until you remember these are the same people who promised us flying cars by 2020.

Musk says his new AI system will uncover "the deepest secrets of the universe," which is exactly what someone says when they need to justify spending billions on more GPUs. This is supposedly "AI 2.0" - not like that peasant ChatGPT that just knows stuff, but real AI that'll solve all our problems. Right.

Here's what we know for sure: Musk has a massive computer in Memphis burning through more electricity than Iceland uses annually. The thing needs its own substation just to keep the lights on. Whether it actually does anything more useful than generating slightly better chatbot responses that still can't count the Rs in "strawberry" remains to be seen.

The real story here isn't the technology - it's that only a handful of people can afford to play this game anymore. Building these systems costs billions, which means AI research is rapidly becoming the exclusive playground of billionaires and nation-states. Google's Q1 2024 results showed massive AI infrastructure spending, while Amazon poured $75 billion into data centers. Even Meta's AI spending is reaching astronomical levels. Microsoft's AI investments are already cratering their margins. Nvidia's data center revenue continues to explode as everyone races to buy more chips. OpenAI's infrastructure costs are bleeding money faster than a startup with no business model. Great for humanity, I'm sure.

Oh, and someone's throwing around numbers like $14 quadrillion in future AI value. That's not a real number, by the way. That's what happens when consultants multiply "revolutionary technology" by "total addressable market" by "hopium." Meanwhile, Microsoft's AI investments are already cratering their margins and they're still trying to figure out how to make money from it.

But hey, at least the electricity bills are real.

So Trump's throwing $500 billion at AI with something called "Stargate" because apparently we're in a dick-measuring contest with China and can't let them have the biggest robot. This is being sold as the "largest AI investment in U.S. history," which is like saying it's the largest investment in something that didn't really exist five years ago.

The pitch sounds familiar: invest massive taxpayer money in technology that might transform everything, just like the internet/electricity/the wheel/fire. Remember when the federal government was going to build a nationwide fiber network that would change everything? Yeah, most of us are still on Comcast. Or how about the high-speed rail that was supposed to revolutionize transport? California's still working on it, $100 billion later. Don't forget the national EV charging network that was going to be everywhere by now.

What are we actually buying with $500 billion? Data centers, chip factories, and probably a lot of consultants explaining why we need more data centers and chip factories. The goal is to beat China at building AI systems, because nothing says "American innovation" like massive government spending to match what another country's government is doing.

They're throwing around numbers like AI will add $20 trillion to the economy by 2030. For context, the entire U.S. GDP is around $26 trillion. So apparently AI is going to almost double the economy in six years. This is the same logic that gave us the dot-com bubble, the housing bubble, and the crypto bubble. Remember when blockchain was going to fix everything? Or when the metaverse was the future? Now we're doing it again, but with AI chatbots that can't do math.

The plan involves giving taxpayer money to companies like xAI and OpenAI so they'll promise to keep their servers in America instead of moving them somewhere with cheaper electricity. It's basically corporate welfare with extra steps and a national security veneer.

Here's what we know will happen: contractors will get rich building facilities that may or may not work as advertised. Politicians will hold ribbon-cutting ceremonies. And in five years, we'll discover that most of the "breakthrough AI" is just better chatbots that can write press releases about how revolutionary they are.

But hey, at least that $14 quadrillion in long-term impact sounds impressive in PowerPoint presentations to Congress.