Beyond Betting on OpenAI! Microsoft CEO Nadella: AI Models Shouldn't Be Locked In — Customer Cross-Model Usage Soars 5x in Six Months

Microsoft (MSFT-US) surged over 8% in after-hours trading following a strong earnings report, with CEO Satya Nadella further outlining the company's AI strategy during the earnings call. He emphasized that enterprise customers should not depend on a single large language model but instead build platform architectures capable of seamlessly switching between different AI models—whether closed-weight or open-weight—based on specific needs.

Nadella stated that since the beginning of this year, the number of Microsoft customers using multiple AI models for development has increased fivefold, indicating a rapid shift among enterprises from "betting on a single model" to a "multi-model strategy." He stressed that applications should be decoupled from the models themselves, saying, "Your entire architecture must be separated from the model, meaning any model can be replaced at any time."

He further noted that while enterprises can certainly adopt the most advanced (frontier) AI models, there is no reason to rely on just one. "You should and can use frontier models, but you can also use multiple models simultaneously." He believes that what enterprises truly need is flexibility—to freely switch between the most suitable models based on cost, performance, security, and varying workloads.

Microsoft has been expanding its AI footprint in recent years, not only as a major strategic investor in OpenAI but also investing in Anthropic. The company has integrated models from both companies, along with numerous open-source models, into its Azure AI Foundry platform, enabling enterprises to deploy, manage, and compare different models through a single interface.

In fact, Microsoft has consistently moved toward a "model-agnostic" strategy over the past several quarters. According to previous earnings call data, Azure AI Foundry now offers the industry's most comprehensive model ecosystem, including not only OpenAI but also Anthropic, xAI, Mistral, and a wide range of open-source models. Enterprises can select different models for different tasks. Microsoft 365 Copilot has already implemented smart routing (auto routing), automatically choosing the most suitable AI model based on the task context to improve efficiency and reduce inference costs.

Microsoft’s recently released Q4 FY2026 earnings report further reflects the sustained surge in AI demand. Quarterly revenue reached $90.01 billion, an 18% year-over-year increase. Azure cloud services revenue grew 43% (in constant currency), exceeding market expectations. Intelligent Cloud revenue hit $39.31 billion, and Microsoft Cloud overall reached $59.3 billion in revenue—surpassing the $100 billion annual revenue mark for the first time. Copilot now has 30 million paid users, with AI-driven demand continuing to fuel growth in Azure and enterprise software.

Notably, one week before the earnings call, Microsoft joined NVIDIA (NVDA-US), Palantir (PLTR-US), Meta (META-US), IBM (IBM-US), and other tech firms in co-signing an open letter urging the U.S. government to avoid imposing excessive restrictions on open-weight AI models. They argued that open models enhance competitiveness, lower enterprise adoption costs, and promote technological transparency and innovation.

However, the AI industry remains divided on open models. Leading frontier model developers like OpenAI and Anthropic generally take a more cautious stance on fully open-weight models, citing risks such as cyberattacks, biosecurity threats, and malicious misuse. In contrast, infrastructure and cloud providers like Microsoft and NVIDIA favor a more open ecosystem, advocating for platforms that offer multiple model choices so enterprises can deploy the best AI solutions for their needs—without being locked into a single vendor.

Nadella’s remarks once again highlight Microsoft’s positioning of Azure as the "operating system" of the AI era. Its core competitive advantage lies not in betting on any single model company, but in building an open platform that can simultaneously accommodate OpenAI, Anthropic, and open-source models—enabling enterprise customers to continuously switch to the best models as technology evolves, maintaining maximum flexibility in a rapidly advancing AI market.