Hamish McRae’s argument in This is Money is that both episodes expose excessive dependence on systems—or people—that appear indispensable until circumstances change. For Windows users and IT administrators, that argument becomes valuable when translated into something more specific than “have a backup.”
A useful fallback must be accessible, appropriate to the failure and tested before the emergency. A recovery plan that nobody can retrieve is not much of a recovery plan.
Two disruptions, two different risks
Nvidia’s SEC filing establishes the precise leadership timeline. Its board accepted Huang’s resignation as president, chief executive and director on November 11, 2026, effective immediately. The company publicly announced the transition on November 12. Those dates describe the board action and subsequent announcement, not separate departures.
Nvidia appointed chief financial officer Colette Kress as interim CEO, retaining her finance responsibilities while the board searches for a permanent successor. The company said it would consider both internal and external candidates and appointed Harvey C. Jones board chair. This was an announced succession arrangement, not an announcement that Nvidia’s business had stopped operating.
Microsoft’s disruption was an operational event. The Associated Press reported problems with Outlook, Teams, Copilot and wider Microsoft 365 services. That matters because losing access to several tools can interrupt more than correspondence: it can interfere with the coordination needed to respond to the disruption itself.
The distinction is fundamental:
- Leadership concentration raises questions about succession, decision-making and institutional knowledge.
- Service concentration raises questions about availability, shared dependencies and continuity of work.
The comparison is useful as an analytical framework. It is not evidence that Huang’s departure caused Microsoft’s problems, or that a cloud outage proves anything about Nvidia’s future competitiveness.
What the Microsoft incident actually establishes
StatusStack’s compilation of Microsoft’s incident updates identifies EX1550543 for Exchange Online and MO1550821 for the broader Microsoft 365 issue. Its record says Microsoft marked both incidents restored in final updates posted late on November 13, Eastern Time. The recorded Exchange incident end time was 11:10 a.m. EST on November 13, distinct from the later publication of the final update.
According to Microsoft’s explanation reproduced there, an issue associated with an earlier network-capacity upgrade surfaced following a subsequent network change. Microsoft described this as a technical problem, not one related to cyber activity or AI agents, and said a post-incident report would follow. That is an attributed provider explanation—not an independently completed root-cause investigation.
The same incident record describes mailbox access problems across Outlook on the web, desktop Outlook’s MAPI connections, Exchange ActiveSync and REST-based connections. Consequently, changing the application used to reach a mailbox was not a guaranteed escape route. Nor does the record establish that every user, tenant or Microsoft product was unavailable throughout the incident.
For administrators, the lesson is to distinguish a service-side failure from a local application fault before making disruptive changes. Microsoft’s own troubleshooting guidance recommends checking service health before spending time investigating a problem that may already be known and under repair.
A spinning Outlook window may look like a PC problem. The sensible first question is whether the fault actually lives on that PC.
Measure impact by workflow, not just application names
There is corroboration that the disruption reached essential services. The Associated Press reported that the outage affected the UK’s National Health Service and threatened doctors’ ability to email prescriptions to pharmacies. That supports a concrete operational concern, but not a blanket assertion that every affected public-service organisation reverted to paper records.
In the United States, the Southern District of New York issued a November 13 notice stating that the November 12 Microsoft outage interfered with email access and the ability to send and receive messages. This supplies a direct institutional example beyond anecdotal complaints about unresponsive applications.
For continuity planning, “email unavailable” is only the beginning of the description. Ask what email enables in your organisation:
- Does it deliver instructions that must be acknowledged?
- Is it the only way a customer can report an urgent problem?
- Does it carry approvals needed before another team can proceed?
- Is the emergency contact list stored inside the unavailable mailbox?
These are planning questions, not claims that every organisation suffered those consequences. Their purpose is to identify where an inconvenience becomes a service interruption.
The practical unit of resilience is the essential task, not the application icon.
First response: establish what is failing
Microsoft provides a documented route for checking Microsoft 365 health. An administrator should:
- Sign in to the Microsoft 365 admin center with an account permitted to view service health. Microsoft identifies Service Support admins and Helpdesk admins among the supported roles.
- Open Health > Service health, or select the Service health card on the home dashboard.
- Examine the relevant incident or advisory and compare its scope with the symptoms being reported.
- Use Issue history when investigating an incident that has already been resolved.
- If the problem is not listed, use Report an issue and complete the form. If admin-center sign-in is unavailable, Microsoft recommends its public service-status page as an alternative.
For this incident, retain EX1550543 and MO1550821 in the internal incident record so that later findings can be matched to the correct provider reports.
Microsoft also explains that the authenticated Service Health view knows which tenants an administrator manages and therefore provides a more targeted picture than a general status announcement. A public page is useful, but tenant-specific information is preferable when available.
As an operational recommendation, record what staff experienced, when it began, which tasks stopped and what subsequently recovered. The goal is to build a usable account of business impact, rather than preserve a collection of increasingly exasperated screenshots.
A practical Windows safeguard: prepare essential files
One concrete preparation is available through OneDrive Files On-Demand for Windows. Microsoft distinguishes online-only files from files explicitly retained on the device. Seeing a filename in File Explorer does not necessarily mean its contents are available without a connection.
For a small, approved set of essential documents:
- While connected and signed in to OneDrive, locate the file or folder in File Explorer.
- Right-click it and select Always keep on this device.
- Allow the download to finish.
- Look for the solid green circle with a white check mark, which Microsoft identifies as the always-available state.
- As a preparation exercise, disconnect from the network and confirm that the required document opens in the intended application.
This consumes local storage. Microsoft also warns that newly created files from the web or another device can appear as online-only, so do not assume yesterday’s check covers every future addition. If the intended file is not downloaded, restore connectivity and resolve that before relying on it.
Treat this as a targeted offline-access measure—not a promise that every protected document or cloud-dependent workflow will function during an outage.
Offline availability is not an independent backup
Microsoft documents that changes and deletions in a synchronised OneDrive folder are reflected online, and vice versa. A locally available synchronised file therefore serves a different purpose from a separately managed recovery copy.
That distinction suggests two separate tests: can someone open the information during a connectivity problem, and can the organisation recover an earlier valid copy after an unwanted change? Passing one test should not be treated as passing both.
Keep sensitive material within approved storage and access arrangements. Continuity planning should not become an excuse to scatter business records across personal accounts.
Build a fallback around essential work
The following is a planning exercise, not a Microsoft-issued workaround for the November incident.
First, choose the tasks that genuinely cannot wait. Avoid beginning with a demand to duplicate every application. Identify the minimum work required to keep customers informed, protect urgent operations and coordinate staff.
Second, assign a fallback owner. Someone should have authority to activate the alternative process, explain its limits and decide when normal working can resume. Otherwise, even a good procedure may sit unused while everyone waits for permission.
Third, define an alternative communication route. Ask whether staff can reach it without the affected mailbox, application or sign-in path. Do not merely nominate another channel and assume it is independent; test the scenario you intend it to survive.
Fourth, make the instructions retrievable. Select an approved way to retain emergency contacts and a short operational checklist outside the workflow whose failure triggers the plan. Include a review date and an owner.
Fifth, rehearse the return to normal. Decide how temporary records will be reconciled, how pending work will be checked and how duplicate actions will be avoided. Restarting an application is not the same as establishing that every interrupted task was completed.
A useful tabletop exercise is deliberately modest: “For the next hour, assume the normal mailbox and collaboration channel cannot be reached. Show how you would handle one urgent request.”
Judge success by whether people can complete that task safely and account for what they did. An impressive document is optional; a usable procedure is not.
Nvidia’s transition calls for evidence, not extrapolation
Nvidia’s announcement offered a continuity message: its board said it expected operations, customer commitments and the product roadmap to continue without interruption. That is the company’s stated expectation, not an independently demonstrated guarantee about future execution.
Equally, the resignation itself does not establish McRae’s prediction that Nvidia will become a less exceptional specialist business. The SEC filing documents a leadership change and interim appointments; it does not substantiate a forecast about the company’s eventual market position.
For technology buyers and developers, a more disciplined response is to identify what evidence would change a decision. Relevant questions might include whether contractual commitments change, whether a supported product’s lifecycle changes, or whether an announced delivery schedule is revised.
Those are questions to monitor, not developments established here.
The same restraint applies to the column’s investment argument. Its market predictions should remain attributed opinion, not be repackaged as an IT conclusion. Leadership succession and portfolio allocation are different decisions, even when both involve concentration risk.
The broader organisational lesson is worth retaining: ask where important knowledge and authority reside, and whether anyone else can exercise them. That question is just as useful for a small IT department as for a prominent technology supplier.
Household preparedness deserves a fair hearing
McRae’s criticism of the UK government’s wind-up-radio advice misses the practical rationale stated in its Be Ready guidance. The campaign recommends a battery-powered or wind-up FM radio for updates during outages, alongside a charged power bank and written emergency information.
The guidance recommends preparing for three days, building supplies gradually and checking what a household already owns. Its checklist includes drinking water, non-perishable food, a torch, warm clothing, copies of important documents, and cash in case electronic payments or ATMs are disrupted. It explicitly acknowledges that not everyone can take every suggested step.
These are general emergency preparations, not evidence that the Microsoft incident interrupted electricity or payment systems. The useful connection is narrower: some information and basic resources should remain accessible when the usual digital route is unavailable.
For readers outside the UK, the campaign is a British example, not a substitute for local emergency arrangements. The principle to borrow is proportionate preparation, not an assumption that every household faces the same risks.
Resilience is something to demonstrate
The strongest conclusion is not that cloud services should be abandoned or that a company cannot outlast a prominent founder. It is that confidence should be accompanied by a testable plan.
For a Windows user, that might mean verifying access to a few essential documents. For an administrator, it means knowing how to establish incident scope and keep one critical workflow moving. For an organisation evaluating a supplier, it means separating an unsettling announcement from evidence of changed commitments.
“Have a backup” is a slogan. Know what it preserves, how to reach it and how to return safely to normal is an operating principle.
References
- Nvidia and Microsoft meltdowns show us all that we must be more resilient in our everyday lives, says HAMISH MCRAE This is Money · 2026-11-15T07:00:53+00:00
- Microsoft 365 and Exchange Online outage, November 12, 2026 statusstack.com
- How to check Microsoft 365 service health - Microsoft 365 Enterprise | Microsoft Learn learn.microsoft.com