Microsoft just posted massive earnings. Revenue hit $81.3 billion for the quarter, up 17%. Cloud revenue topped $50 billion for the first time ever.
So why did the stock tank the next day? Investors are spooked. Microsoft spent $72.4 billion on data centers in just six months. That’s almost as much as the entire previous year. And the big question hanging over everything: Is anyone actually using this AI?
CEO Satya Nadella spent most of the earnings call trying to convince everyone the answer is yes. But the numbers he shared? They’re surprisingly vague for a company that usually loves precision.
The Spending Panic Is Real
Microsoft dumped $88.2 billion into capital expenditures last year. This year, they’ve already burned through $72.4 billion in the first half alone.
Most of that money builds AI infrastructure. Data centers to serve OpenAI. Computing power for Anthropic. Cloud capacity for enterprises demanding AI tools.
But here’s what scared Wall Street. Both Azure cloud growth and Microsoft 365 subscription growth came in below expectations. Not by much. But enough to trigger alarm bells.
UBS analyst Karl Keirstead put it bluntly in his research note. “The fact that BOTH Azure and the M365 segments fell a bit short is the key negative we’re hearing.” (He still recommends buying the stock, though.)
Plus, reports surfaced months ago suggesting people weren’t actually using Copilot despite Microsoft weaving it into everything. That narrative clearly bothered Nadella. He spent considerable time on the call pushing back.
The Usage Numbers Don’t Add Up
Nadella claimed daily users of consumer Copilot products grew “nearly 3x year-over-year.” Sounds impressive. But what does that actually mean?
Microsoft counts AI chats, news feeds, search, browsing, shopping, and “operating system integrations” in this metric. That’s a lot of different things lumped together. And Nadella didn’t specify actual user counts.
Microsoft only told TechCrunch after the call that total Copilot users hit 150 million. But that includes both commercial and consumer users. Last year’s annual report claimed 100 million monthly active users. So growth happened. Just not as explosive as the “3x” language suggests.
Moreover, the company splits consumer and enterprise numbers when things look good. The fact they’re combining them here? That tells its own story.
GitHub Copilot Actually Works
Not all the news was murky. GitHub Copilot, Microsoft’s coding assistant, now has 4.7 million paid subscribers. That’s up 75% year-over-year.
This appears to be a genuine success. Developers actually pay for it. They use it daily. It delivers clear value by writing code faster.
Last year’s annual report mentioned 20 million total GitHub Copilot users including free tier. But the paid subscriber growth suggests real demand exists for this particular AI tool.
The difference? GitHub Copilot solves a specific problem developers face constantly. It’s not AI for AI’s sake.
Enterprise Adoption Remains Slow
Microsoft 365 Copilot, the AI assistant for office workers, now has 15 million paid seats. Companies buy licenses for their employees to use Copilot in Word, Excel, and Outlook.
But here’s the context that matters. Microsoft has 450 million total paid Microsoft 365 seats. So Copilot penetration is about 3.3% of the base.
That’s after significant marketing push and integration into core products. After spending billions to build the infrastructure. After making it dead simple to deploy.
The slow uptake suggests companies aren’t convinced employees need AI assistants badly enough to pay extra. Or employees aren’t finding enough value to demand it.
Healthcare AI Shows Promise
Dragon Copilot, Microsoft’s healthcare AI for medical professionals, reached 100,000 providers. It documented 21 million patient encounters this quarter, up threefold year-over-year.
This competes with superhot startup Harvey and other medical AI tools. The growth trajectory looks strong. Healthcare providers clearly see value in automating documentation.
But again, Microsoft didn’t break out revenue numbers. We don’t know if this translates to meaningful profits or just more infrastructure costs.
The Capacity Argument
Nadella and CFO Amy Hood made one key claim during the call. They said demand for AI services far outstrips their data center supply.
According to them, all the new equipment is essentially booked to capacity for its entire lifespan. If true, that justifies the massive spending. You can’t sell what you don’t have capacity to deliver.
But this raises questions. If demand is so overwhelming, why aren’t Azure and Microsoft 365 growing faster? Why is enterprise Copilot adoption still under 4% of potential customers?
The math doesn’t quite add up. Unless the capacity is mostly going to a few giant customers like OpenAI rather than broad-based enterprise demand.
Wall Street Wants Proof
Investors aren’t being unreasonable. Microsoft is spending historic amounts on AI infrastructure. The company needs to show that investment will generate returns.
Vague growth percentages without absolute numbers don’t cut it. “Nearly 3x” daily users means nothing without knowing the starting point. 150 million total users sounds big until you realize it includes every free tier user who tried Copilot once.
The market wants to see enterprise customers paying significant money for AI tools. They want proof that companies will spend enough on AI to justify the infrastructure costs.
Right now, GitHub Copilot and Dragon Copilot show genuine traction. But the broader enterprise market seems hesitant. That hesitation cost Microsoft billions in market cap after earnings.
The AI Spending Arms Race Continues
Microsoft isn’t alone in this bet. Google, Amazon, and others are pouring similar amounts into AI infrastructure. Everyone’s convinced the next computing platform is AI-powered.
But someone has to be wrong. Not everyone can generate returns on this level of spending. The market can only support so many AI platforms with different strengths.
Microsoft’s advantage is its massive enterprise customer base. They already use Office 365, Azure, and other Microsoft products. Adding AI should be simple.
Yet that’s not happening fast enough to satisfy investors. The question remains whether it’s a timing issue or a demand issue.
Your guess is as good as mine. But I’d want to see much clearer usage numbers before betting billions more on AI infrastructure expansion.
Comments (0)