Microsoft CEO Satya Nadella has reinforced his recent warning to businesses about overreliance on artificial intelligence providers, arguing that companies which depend entirely on a single AI platform could jeopardize their long-term future.
Speaking on CNN’s Fareed Zakaria GPS on Sunday, Nadella expanded on comments he made earlier this month, emphasizing that organizations should maintain control over the data, prompts, and usage information generated through their AI systems instead of handing everything over to external AI vendors.
According to Nadella, businesses should adopt architectures where every interaction with an AI model remains under their ownership. This includes retaining metadata generated during AI usage, allowing organizations to eventually train their own AI models or fine-tune open-weight alternatives using their proprietary information.
He explained that this approach enables companies to preserve control over their intellectual assets while reducing dependence on any single AI provider.
“Any firm that doesn’t have this control,” Nadella said, “will not remain a firm because you’ve essentially outsourced your thinking.”
Maintain Control Over AI Infrastructure
Nadella argued that businesses should avoid building their entire AI strategy around one proprietary model or ecosystem. Instead, organizations should separate their AI infrastructure—including prompts, memory, and orchestration layers—from the underlying language models.
This architecture would allow enterprises to switch between different AI models depending on their strengths without disrupting operations if one provider changes pricing, policies, or availability.
He also advised businesses not to become overly dependent on proprietary AI coding assistants, such as Anthropic’s Claude Code or OpenAI’s ChatGPT Codex. By keeping coding workflows and AI management systems independent of any single vendor, companies can retain flexibility while taking advantage of multiple AI models.
According to Nadella, separating the application layer from the model itself ensures that organizations remain in control of their own technology roadmap rather than becoming locked into one ecosystem.
Microsoft’s Position
Nadella’s remarks are particularly noteworthy given Microsoft’s significant investments in both OpenAI and Anthropic, two of the world’s leading AI developers. Microsoft also provides AI infrastructure through its Azure cloud platform, including tools that help enterprises manage and deploy multiple AI models.
While Microsoft’s own business stands to benefit from companies adopting multi-model AI infrastructure, many industry observers agree that avoiding vendor lock-in has become an increasingly important consideration for enterprise AI adoption.
Growing Interest in Open AI Models
Many businesses are already moving toward open-weight AI models that allow them to customize systems using their own data while running models on private infrastructure. These models offer greater flexibility, lower operating costs in some cases, and more control over sensitive business information.
As organizations deploy multiple AI models simultaneously, there is growing demand for infrastructure capable of managing different providers without becoming tied to one platform.
A Competitive Risk
Beyond technical flexibility, Nadella also warned of a broader strategic concern.
He suggested that companies giving AI providers extensive access to their internal operations, workflows, and proprietary knowledge could eventually face competition from those same providers. As AI systems become more deeply integrated into business processes, model developers gain greater visibility into how companies operate.
The concern echoes warnings that have circulated throughout the startup ecosystem for several years.
In May, OpenAI CEO Sam Altman offered AI credits to startups participating in Y Combinator’s latest cohort. Following the announcement, investor Jason Calacanis cautioned founders that accepting those incentives could expose valuable business ideas to platform providers.
Calacanis warned that AI companies could potentially study how startups use their platforms before developing competing products within their own ecosystems.
Nadella’s comments suggest similar concerns now apply not only to startups but also to large enterprises adopting AI across their organizations.
Different Rules for Consumers
While Nadella urged businesses to safeguard their AI data, he drew a distinction between enterprise and consumer use.
When asked how everyday users should protect their information, Nadella acknowledged that personal data sharing is often part of the value exchange consumers accept when using free online services.
He noted that advertising-supported digital platforms have long operated on this model, where users receive free products or services in exchange for providing data that supports the platform’s business model.
As enterprise AI adoption accelerates, Nadella’s latest remarks highlight a growing industry debate over data ownership, vendor dependence, and the importance of maintaining long-term control over AI infrastructure.
