Technology

Nadella Says Much of U.S. Critical Infrastructure Cannot Be Patched

Martin HollowayPublished just now4 min readBased on 1 source
Reading level
Nadella Says Much of U.S. Critical Infrastructure Cannot Be Patched
Photo by Briansmale / CC BY-SA 4.0

Microsoft CEO Satya Nadella said on November 18, 2026, at The New York Times' DealBook Summit, that a lot of U.S. critical infrastructure is unpatchable. TechCrunch reported his remarks the same day.

He cited firmware, chillers and HVAC systems as examples of unpatchable systems that sit on the internet. He also said transportation systems include discoverable systems running firmware that cannot be patched. The examples span physical plant, from building management hardware to vehicle and transit infrastructure. Each is a device class that typically ships with an update path that operators do not control or cannot easily use.

Nadella also said AI has made sophisticated cyberattacks much easier to carry out. On the update side, he said patches must now move from available to installed on live production systems instantly, because attackers are exploiting them in the wild, meaning against real targets right now. He said this rapid cycle creates instability.

He also addressed the recent global Microsoft Outlook outage. He attributed it to a latent bug, a dormant flaw that surfaced only when several conditions lined up at once, which he described as a perfect storm. Microsoft issued an incident response for the outage and would apply upstream architectural and operational learnings to fix the underlying problem. The fix is described as a plan, and the remarks did not describe it as complete.

The broader context here is a collision between two operating models. Enterprise patch processes are built around scheduled maintenance windows, staged rollouts and regression testing, the practice of re-running tests to confirm a change has not broken existing behavior. Nadella's framing is that exploitation now arrives on a timeline those processes were not designed for, and that some assets cannot accept a patch at all. Those two problems call for different responses. The first calls for faster change control that remains safe under production load. The second calls for knowing which devices exist, isolating them, and replacing them when no update path exists.

In my view, the most useful reading of the speech is operational rather than alarmist. The tension Nadella described is one that network and facilities teams already live with, often without the vocabulary to put it to a board. Naming it plainly, from the position of a platform vendor whose software runs across much of the enterprise, gives operators a clearer starting point. An organization cannot isolate or retire a device it has not yet discovered on its own network, so a discovery inventory is the least glamorous and most practical step available.

The same AI capability Nadella cited as lowering the cost of attacks also raises the question of how defenders can keep pace. That question is open, and the remarks did not answer it. What the speech did establish is a shared diagnosis: the patch cycle that protects a modern cloud estate does not map cleanly onto a chiller plant or a signaling controller. The encouraging part is that the problem is now stated in terms that engineers, procurement staff and regulators can act on.