Satya Nadella says Microsoft will cut AI costs and rebuild around agents as it bets on a ‘frontier ecosystem’
In his annual letter, the Microsoft CEO says customers will judge AI by cost per outcome, as Azure crosses USD 100 billion and Copilot tops 30 million seats.
Microsoft Chairman and CEO Satya Nadella has said the company is working to bring down the cost of its AI services and widen its partner ecosystem, as artificial intelligence moves from pilots to mass adoption. In his annual shareholder letter, which drew fresh coverage in India this week, Nadella argued that customers will increasingly weigh what they spend on AI against the results they get, and said Microsoft is reshaping its products around AI agents rather than traditional apps.
The ‘cost-to-outcome’ frontier
Nadella wrote that inference, the process of running AI models to produce answers, has made marginal cost a central factor in software economics, something that was rarely true of traditional software. He said customers “will increasingly measure the cost of tokens against the outcomes they produce,” and called this the “cost-to-outcome” frontier that Microsoft is trying to optimise.
To get there, the company says it is speeding up the deployment of capacity while lowering infrastructure costs through improvements in datacentre design, custom silicon, systems software and models. The letter says Microsoft’s Maia 300 chip delivers 40 per cent more tokens per watt than leading GPU-based systems, and that Azure added 88 datacentres during the year.
From apps to agents
Nadella described a shift in computing from interactive applications to proactive agents that pursue goals set by users and come back for direction only when needed. Microsoft 365 Copilot, he noted, has evolved from a chat assistant into agents that can carry out long-running tasks. The company is also offering Agent 365 as a control plane to govern AI agents in the same way organisations manage employees.
A key design choice, according to the letter, is separating the “harness” (context, memory and actions) from the underlying model so that the best model can be used for each job. Microsoft’s Foundry platform now offers more than 11,000 models from partners such as OpenAI, Anthropic and Mistral alongside Microsoft’s own MAI family, and has crossed 100,000 customers.
A partner-first ‘frontier ecosystem’
Nadella framed Microsoft’s strategy as platform- and partner-first, saying the aim is an ecosystem in which others create more value than Microsoft captures itself. He also stressed that customers should retain ownership of the learning loops built on their proprietary data. He said he was confident of long-term returns on Microsoft’s heavy AI spending, citing strong demand and efficiency gains.
The numbers behind the push
Microsoft’s FY2026 results, as cited in the letter and reported by ANI, include:
- Revenue of USD 331.8 billion, up 18 per cent year-on-year
- Microsoft Cloud revenue of USD 214.4 billion, up 27 per cent
- Azure growth of 41 per cent, crossing USD 100 billion for the first time
- Operating income of USD 155.2 billion, up 21 per cent
- More than 30 million paid Microsoft 365 Copilot seats and 50 million GitHub Copilot users
Why it matters
As enterprises, including many in India, move AI from experiments into daily operations, the cost of running models at scale has become a major concern. Nadella’s emphasis on cheaper inference, swappable models and outcome-linked value signals how the world’s largest software company expects AI to be bought and priced in the coming years.
Sources
Related
Ultraviolette X-47 Venom launched at Rs 2.65 lakh with up to 343 km range on new VX-100 platform
SBI confirms 3,897 customer records found on dark web; probe under way
