Nadella warns against single-model AI strategies

Microsoft CEO Satya Nadella said enterprises that depend on a single AI model or outside platform risk losing control of critical assets and falling behind. In a CNN interview, he framed data sovereignty and multi-model infrastructure as central to corporate AI strategy, warning that prompts, response data and metadata should remain under enterprise control rather than flowing entirely to model providers.

Nadella argued that companies should build flexible internal systems, including an AI Gateway, to manage prompts and data while routing work to different models based on their strengths. Separating models from context, memory, development tools and data would let businesses switch providers if a model becomes unavailable or unsuitable, while preserving the material needed to train proprietary weights or open models later.

He also addressed Chinese models such as DeepSeek and Moonshot AI’s Kimi K3, saying U.S. hyperscaler infrastructure and trust in the U.S. technology ecosystem remain competitive advantages. His comments align with a broader enterprise shift away from vendor lock-in toward open-weight models and proprietary platforms, a position also echoed by Palantir CEO Alex Karp.