Nadella warns enterprises are ‘paying twice’ for AI as Hormuz tensions flare
Microsoft CEO Satya Nadella issued a blunt warning to enterprises this week: using proprietary AI models means you are “paying twice”—once with subscription fees and again by surrendering proprietary business knowledge [1]. In a Sunday blog post, Nadella argued that every prompt,
Microsoft CEO Satya Nadella issued a blunt warning to enterprises this week: using proprietary AI models means you are “paying twice”—once with subscription fees and again by surrendering proprietary business knowledge [1]. In a Sunday blog post, Nadella argued that every prompt, agent correction, and workflow refinement teaches model providers like OpenAI and Anthropic the inner workings of their customers’ operations. “Every correction is distilled into institutional know-how,” he wrote, calling it “the kind of knowledge a competitor could never buy” [1].
Nadella’s proposed fix is twofold. Companies should retain ownership of their prompts and feedback inside “proprietary learning environments,” and they should deploy “orchestration layers” that let them swap between models rather than lock in with one vendor [1]. The subtext is a push toward open-source and on-premise deployments. Solo.io CEO Idit Levine told TechCrunch that customers are increasingly asking whether an open-source model running on their own hardware can do “almost 90% of what the big one’s doing” at far lower cost [1]. Vercel’s AI gateway now routes 29% of its traffic to open models, suggesting the shift is already measurable [1].
The irony is hard to miss: Microsoft has invested billions in both OpenAI and Anthropic, yet its CEO is now urging caution about exactly those kinds of proprietary models [1].
Elsewhere in AI, Meta rolled out Muse Image, its most advanced image generator to date, across the Meta AI app, Instagram Stories, WhatsApp, and meta.ai [2]. Built by Meta Superintelligence Labs, Muse Image acts as an agent that can write code, search, self-refine outputs, and collaborate with Meta’s Muse Spark assistant [2]. A preview of Muse Video, with native audio support, is also coming soon [2].
On the world stage, the U.S.-Iran ceasefire agreed to in June is unraveling around the Strait of Hormuz. President Donald Trump notified Congress on July 10 that the Pentagon had resumed strikes against Iran beginning July 7, describing them as “limited, measured, planned, and executed” to minimize civilian casualties [3]. Iran’s Revolutionary Guard claims it has struck U.S. forces at a Jordan airbase and the U.S. Fifth Fleet in Bahrain, while Jordan says it intercepted four Iranian missiles in its airspace [3]. Brazilian President Luiz Inácio Lula da Silva slammed Trump’s plan to impose a 20% tariff on ships passing through Hormuz as “piracy” [3]. A new Reuters/Ipsos poll found 79% of Americans expect U.S. military involvement to continue for an extended period, while 58% oppose the war [3].
The through-line today: whether in AI or geopolitics, control of chokepoints—data or shipping lanes—is becoming the central battle.