On November 18, 2026, Microsoft CEO Satya Nadella warned at The New York Times' DealBook Summit that much of America's critical infrastructure is 'unpatchable.' 'Think about all the firmware and the chillers and the HVAC systems everywhere else,' Nadella said. 'So all of that, by the way, they're on the internet.' Chillers are the cooling units widely used in data centers; transportation systems, which Nadella also named, contain just as much discoverable firmware that cannot be patched.
The problem comes from two decades of the 'smartification' of the economy: vast numbers of devices that were once isolated were connected to the internet, without any update channel designed in. AI makes the situation more urgent: it can find vulnerabilities faster than human researchers, and it can help attackers turn a flaw into a working exploit faster too. Nadella's judgment is that the patching window has collapsed — once a fix is published, it has to reach production immediately, 'because it's being exploited in the wild.'
The Faster the Patch, the Less Stable the System
But 'patch immediately' has itself become a new risk. Nadella acknowledged that rushed rollouts create instability — and that is the backdrop to the recent wave of large-scale outages. Microsoft Outlook was among them; he blamed a 'latent bug' that produced a 'perfect storm,' and stressed that officials have denied any link between these outages and cyberattacks. In other words, the dilemma now is not a lack of patches, but that patching speed, stability, and coverage cannot all be had at once.
The message is practical: for companies, firmware, cooling, and building systems have to enter the security budget — not just servers and code; for ordinary people, the risk to essential services is shifting from 'will it be breached?' to 'how fast can it recover?' AI is helping defenders find flaws — two remote code execution flaws in NGINX that had hidden for eleven years were dug up by AI — and it is also shortening attackers' timelines. That is the backdrop to the frontier risk management framework OpenAI published this week: the window for defense will only keep narrowing.