Artificial intelligence has delivered $10bn to Microsoft’s bottom line, but the business is not in cloud GPUs for third-party inference engines
Microsoft’s artificial intelligence (AI) business is set to hit an annual run rate for subscription renewals of $10bn.
For the quarter that ended September 30, the company reported revenue of $65.6bn, up 16%. The Microsoft Cloud business posted revenue of $38.9bn, an increase of 22%, while its productivity and business processes business reported a 12% increase in revenue to $28.3bn.
Server products and cloud services revenue increased 23%, driven by Azure and other cloud services revenue growth of 33%, and Microsoft Dynamics products and cloud services posted a revenue increase of 14%, driven by Dynamics 365 revenue growth of 18%.
According to CEO Satya Nadella, AI is the “fastest business” in the company’s history to reach the $10bn milestone.
“AI-driven transformation is changing work, work artefacts and workflow across every role, function, and business process,” he said. “We are expanding our opportunity and winning new customers as we help them apply our AI platforms and tools to drive new growth and operating leverage.”
During the earnings call, Nadella spoke about the uptake of generative AI (GenAI) among Microsoft’s customers. He said GenAI in the Microsoft Power Platform is helping customers use low-code/no-code tools to cut costs and development time. “To date, nearly 600,000 organisations have used AI-powered capabilities in Power Platform, up four times year-over-year,” said Nadella.
Looking at the GenAI technology in Microsoft 365 CoPilot, Nadella said the number of people who use Microsoft 365 Copilot daily more than doubled quarter-over-quarter.
When asked about future investments needed to innovate in foundational AI models, Nadella said the cost of training these models is limited by the monetisation of each generation of inference engine. “You can think of training essentially as building the next generation model so that then, you have a more capable model [which] drives more inference demand,” he said.
Given Moore’s Law, which stipulates that processing power doubles every 18 months to two years, Nadella said: “You really want to refresh your [datacentre] fleet every year.”
The equipment is then depreciated over the course of its useful lifespan. According to Nadella, demand for AI inference from Microsoft customers ultimately governs how much the company invests in training.
When asked about how the investment in AI inference shows up on Microsoft’s profit and loss account, he said: “One of the things that may not be as evident is that we are not actually selling raw GPUs [graphics processing units] for other people to train.”
He said using cloud-based GPUs to run AI inference workloads from third parties is a business Microsoft turns away from, as it’s largely being driven by tech startups with venture capital money.
“We really are not even participating in most of that, because we are going to the real demand, which is in the enterprise space, or our own products like GitHub Copilot or M365 Copilot,” said Nadella.