Three years ago, Satya Nadella was basically a superhero. He bet early on OpenAI, unveiled AI-powered Bing with all the fanfare of a Marvel premiere, and declared war on Google. Bill Gurley called it an “amazing shift in corporate reputation.” CNN made him CEO of the Year. The guy was untouchable.
Now? Microsoft’s stock is down 24% in a year, and Nadella’s legacy is looking less “visionary” and more “oops.”
Here’s the thing: Microsoft went all-in on AI, and it’s starting to look like they might’ve gone all-in on the wrong bet. Their flagship Copilot product lags behind ChatGPT and Claude. LinkedIn is drowning in AI-generated hustle porn. Xbox is hemorrhaging money. And internally? Employees are quietly asking why the company is burning through a record $190 billion this year on AI infrastructure when the returns are… questionable.
**The Core Problem: You Can’t Eat Your Own Lunch**
Microsoft’s real problem isn’t that AI is bad—it’s that AI is *too good* at replacing Microsoft’s core businesses. For decades, Microsoft owned the office. You opened Word, Excel, PowerPoint. That was your day. Now millions of workers are doing all that stuff directly inside AI tools. Gartner predicts AI will shake up the $58 billion productivity software market. Ouch.
The company’s trying to spin this as a win. “M365 Copilot adoption is huge!” executives say. Sure. But here’s where it gets messy: Microsoft doesn’t have enough computing power to go around.
**The GPU Shortage Nobody Talks About**
Despite spending $190 billion on data centers, Microsoft is *still* constrained. So they made a choice: feed their own AI products first, then give the leftovers to Azure customers. CFO Amy Hood basically admitted this during earnings, saying they prioritize “M365 Copilot and GitHub Copilot” before allocating remaining capacity to Azure.
Translation: We’re starving our cloud customers to feed our own AI habit.
This triggered a 10% stock drop. Investors realized Microsoft was essentially cannibalizing its most profitable business to chase an AI future that might not materialize. One executive even asked the uncomfortable question: “Why would Satya prioritize growing Adobe over growing M365?” (Spoiler: Nobody has a good answer.)
**The Desperation Play**
Microsoft got so desperate for computing capacity that they asked Amazon to bail them out. They explored leasing Oracle infrastructure. Now they’re “shopping for capacity everywhere”—Amazon, Google, you name it. One executive literally said: “We are shopping for capacity everywhere.”
That’s not a strategy. That’s panic.
**The Culture Shift**
Meanwhile, Nadella’s reshaping the company from the inside. He’s flattened leadership, promoted new executives, and overhauled the performance review system. Employees say it feels like the old Microsoft is back—the Steve Ballmer era where “you lead with a lot of fear and a billy club in your hand.”
**The Bottom Line**
Microsoft’s real problem isn’t that AI is a bad bet. It’s that they bet *everything* on AI while their core businesses—the ones that actually make money—are getting disrupted by the same technology. They’re caught between defending the past and building the future, and they’re running out of GPUs to do either.
Nadella went from superhero to cautionary tale in three years. His legacy won’t be defined by whether Microsoft can build the best AI. It’ll be defined by whether they can keep AI from destroying the company that made them one of the world’s most valuable.
That’s a much harder problem to solve.
