Microsoft has “tapped on the brakes” of its spending on artificial intelligence, reducing capital expenses by more than $1 billion in the first three months of 2025. But after 10 straight quarters of increased AI outlay, momentum continues to propel the sector forward with new data center commitments in 16 countries and the expansion of the Phi small language family with the debut of Phi-4 (that supports text, visual and voice inputs). The moves come on the heels of $70 billion in sales for the first three months of 2025, when profits were up 18 percent to $25.8 billion.

“The results far exceeded Wall Street expectations,” reports The New York Times, noting that revenue was up 13 percent for the most recent quarter when measured, like the profit figure, year-over-year.

Microsoft closed the January through March period, its fiscal Q3 2025, having “spent $21.4 billion on capital expenses, down more than $1 billion from the previous quarter,” NYT writes, calling the deceleration from what was originally projected to be $85 billion in capital expenses through the end of the current fiscal year in June “an indication that the tech industry’s appetite for spending on AI is not limitless.”

But the world’s appetite for AI is still robust. Microsoft CEO Satya Nadella “said on a call with investors that demand for cloud and artificial intelligence remained strong,” explaining “the company was tweaking its investments based on efficiency improvements,” per NYT.

Stressing the company’s deep ties with Europe, Microsoft President and Vice Chair Brad Smith said Microsoft is expanding data center operations in 16 European countries, which will more than double its European capacity between 2023 and 2027, resulting in “cloud operations in more than 200 data centers across the continent.”

The company’s cloud and AI operational EU build-out comes “as businesses in the bloc pressure policymakers to protect EU tech sovereignty,” writes The Wall Street Journal. “AI and cloud data centers represent the next stage of industrialization,” according to Microsoft.

Simultaneously, the company in a blog post introduced three successors to its Phi-3 small language model series: Phi-4-reasoning, Phi-4-reasoning-plus, and Phi-4-mini-reasoning.

Phi-4-reasoning is a 14-billion parameter open-weight reasoning model that rivals larger models on complex reasoning tasks while Phi-4-reasoning-plus “builds upon Phi-4-reasoning with reinforcement learning used to leverage more inference-time compute. Phi-4-mini-reasoning is a transformer-based language AI “designed to meet the demand for a compact reasoning model,” explains Microsoft.

TechCrunch says these open models are “competitive with OpenAI o3-mini on at least one benchmark.”

Related:
Microsoft Forecasts Strong Growth for Azure Cloud Business, Reuters, 4/30/25
Satya Nadella Says as Much as 30% of Microsoft Code Is Written by AI, CNBC, 4/29/25