Five Major System Outages in One Week: Microsoft Leads as Older Infrastructure Struggles to Meet the AI Age’s Pace and Demands

17 November brought confirmation of an NSA network outage. Affecting NSANet, the agency’s internal network for transmitting classified information, it received the designation “Code Yellow,” the second-highest level in a three-tier alert system. Apple also suffered widespread disruption that day when an internal system update affected its employee portal, internal sign-in systems and remote VPN access, as well as consumer services including iCloud, the App Store and Apple Pay. Across numerous retail stores worldwide, sales and repair operations were interrupted.

TMGM Analysis: Financial Market News, Economic Calendar & Market Insights

Transaction processing at JPMorgan had slowed a day earlier because of a software bug in its on-premises hardware systems. Reports of outages on Downdetector exceeded 3,000. Customers at Bank of America, Wells Fargo and other institutions also reported related disruption. A further three days back, a communications failure had triggered a nationwide ground stop ordered by the FAA, resulting in more than 5,100 flight cancellations and nearly 10,000 delays. The first failure in this sequence came on 12 November: core Microsoft 365 services became unavailable across at least 70 countries. Simultaneous failures hit Outlook, Teams and Copilot, with disruption continuing for more than two days.

Five critical systems failed in six days, without any incident being identified as an attack or attributed to AI. Describing them as isolated technical failures with no malicious cause nevertheless exposes something harder to remedy than an attack.


Two-Day Microsoft Outage: The Failure Behind a “Capacity Upgrade”

A routine change in user-traffic routing initiated the outage. That is the account set out in Microsoft’s preliminary customer report of 15 November and Bloomberg’s reconstruction of events. On the evening of 11 November, an automated routing process moved Teams traffic onto another server system serving Outlook. An unrelated issue had already taken capacity offline. Together, these factors triggered a bug that had lain dormant in the company’s data centres since July. Surging traffic exhausted server memory, causing crashes.

Rather than containing the disruption, Microsoft’s automated failover system compounded it. Additional traffic went to servers the system deemed healthy, extending the failure. At the morning peak in business traffic on 12 November, more than 80% of Outlook servers were no longer responding. A further 15 hours of investigation followed before Microsoft began rolling out a fix.

The explanation offered by Microsoft Executive Vice President Ramanna exposed a testing weakness. Laboratory simulations of traffic patterns and hardware configurations differed from real-world operating conditions. Older, less powerful systems typically received the testing focus. On newer hardware, by contrast, the bug behind the crashes proved “more destructive”.

The staffing dimension calls for scrutiny equal to that given to the technology. A user identified as a Microsoft employee wrote on Blind, the anonymous workplace forum, that multiple rounds of redundancies and voluntary departures had affected Microsoft 365’s core infrastructure teams in recent years. According to the claim, engineers with long-standing responsibility for legacy systems had left, making rapid diagnosis of complex failures difficult for the remaining staff. A system rollback once achievable in around 45 minutes could now take five hours. Internal incident-management tools reportedly failed to load properly at one point during this outage. Engineers had to investigate code and service architectures inherited from departed colleagues, working across departments late into the night.

These claims remain without a public response from Microsoft. They are, however, compatible with both the exceptional duration of the disruption and the failure of attempted rollbacks to resolve it.


AI’s Pace Versus Infrastructure Renewal: The Growing Mismatch

Across all five incidents, AI received no attribution as the cause. In Microsoft’s case, the company expressly denied any involvement by “AI or agents”; the FAA and NSA likewise excluded AI as a factor. Still, the statements “not caused by AI” and “unrelated to the AI age” do not mean the same thing.

Two diverging curves explain the underlying pattern. The first tracks AI-driven increases in change frequency and workload intensity. Shifts in traffic patterns and hardware configurations accompany faster software updates, leaving testing environments and incident-response capabilities struggling to keep pace. Infrastructure renewal follows the second curve. For the FAA’s copper cables and routers, the timescale runs to decades. The NSA’s networks evolve over generations. Microsoft’s internal ability to maintain legacy systems, meanwhile, is declining quarter by quarter.

Unprecedented speed in finding large numbers of system vulnerabilities is now possible with AI. Companies and government agencies consequently face enormous pressure to fix them at once. Often, that means deploying multiple patches together and heavily straining systems never designed for changes at such speed. AI creating vulnerabilities is not the causal relationship at issue. Instead, AI accelerates change, while the systems tasked with absorbing it were built for a slower operating pace.

The mechanism is clearest at Microsoft. Routine traffic redirection activated a bug dormant since July; automated failover then intensified the failure. With expertise in legacy-system maintenance lost, recovery was unusually slow. No individual stage in this sequence was AI. Each, however, unfolded under pressure from the AI-driven expansion of cloud services.