Microsoft is turning away customers looking to rent GPUs for training new AI models as it shifts its focus to inferencing—a more profitable use of its AI infrastructure. During the Q1 2025 earnings call, CEO Satya Nadella explained that prioritizing inferencing workloads is part of the company’s strategy to fund the massive investments in AI and data center infrastructure, which reached $20 billion this quarter.
Training, essentially involves massive amounts of data be ingested and process to create a new model. Inferencing takes an existing model and fine tunes it to improve its performance, often in a particular domain or context. Given that even a company the size of Microsoft, even with unlimited funds, has a limit to how fast it can build out new infrastructure.
So using those facilities on the most profitable work, has a big impact on earnings. Additionally, the work of inferencing often has linkages to Microsoft’s new AI products.
Nadella emphasized that the real demand lies in powering AI services like GitHub Copilot and M365 Copilot, rather than renting GPU capacity for model training. By focusing on inferencing, Microsoft aims to capitalize on its AI offerings’ growing popularity, with its AI business expected to hit a $10 billion annual run rate next quarter—the fastest new product to reach this milestone in the company’s history.
Microsoft’s Intelligent Cloud segment, which includes Azure, saw significant growth, with Azure revenue up 33% year on year. However, rising infrastructure costs contributed to an overall expense increase of 12% for the quarter. Despite these rising costs, Microsoft highlighted strong revenue growth in areas like Azure, Microsoft 365, and LinkedIn, and pointed to the increasing number of $10 million-plus contracts as evidence of continued enterprise interest in its cloud offerings.
Interestingly, Windows, Microsoft’s cornerstone, has seen a minor decrease in revenue. Which makes the new AI services a much needed boost to company revenues.
Security was also a major focus during the call, with Nadella reassuring investors that Microsoft is prioritizing security across its operations, dedicating the equivalent of 34,000 full-time engineers to address key security challenges. The company has been proactive in applying what it learns from security incidents to enhance its products and develop new security offerings.
While focusing on AI inferencing has helped generate revenue to support its AI investments, the approach has not been without challenges, particularly in managing costs. Still, Microsoft’s leadership remains optimistic, forecasting continued growth for Azure and productivity solutions, while acknowledging some cyclical downturns in other product lines.
