When the Cloud Went Down  ·  October 20, 2025

One Company Sneezed. Half the Internet Caught a Cold.

On an ordinary Monday in October, a single Amazon Web Services region in Northern Virginia failed — and within minutes, banks, brokerages, airlines, and the apps that run daily life went dark together. It was a trillion-dollar lesson in a risk most portfolios carry without knowing it: the single point of failure.

What Happened

The outage, hour by hour.

The failure began quietly — a DNS problem in a single database service, in a single region: us-east-1, the oldest and busiest corner of Amazon's cloud. But because so much of the modern internet routes through that one region, the local problem did not stay local.

Around 3:00 AM ET
Amazon's monitoring begins flagging elevated error rates in DynamoDB, a core database service in us-east-1. The root cause is later identified as a DNS resolution failure.
Morning
The failure cascades. Services that depend on that database — and services that depend on those services — begin failing in sequence. Snapchat, Fortnite, Venmo, Ring, Alexa, Robinhood, Coinbase, airlines, and major banks report disruptions. Users can't log in, can't pay, can't trade.
Midday
Amazon engineers identify and mitigate the DNS issue, but the backlog of failed requests takes hours to clear. Over a thousand companies report outages affecting tens of millions of users.
By Evening
Services are largely restored. The internet returns to normal — and most people move on without asking the question the day begged: how did one region of one company's infrastructure become a load-bearing wall for the global economy?
The Uncomfortable Math

Nobody chose this concentration. It accumulated.

No committee ever voted to route half the internet through Northern Virginia. It happened one reasonable decision at a time: us-east-1 was the first region, the cheapest, the default in the setup menu. Every individual choice made sense. The collective result was fragility that nobody designed and nobody noticed — until the day it mattered.

Markets concentrate the same way. Nobody votes to make a handful of technology companies the load-bearing wall of the S&P 500. It happens one index purchase at a time — each one reasonable, each one reinforcing the last — until the "diversified" fund in your retirement account is quietly running most of its risk through a few names.

~40%
Approximate share of the S&P 500 represented by its ten largest companies — a concentration level exceeding the dot-com peak.
1,000+
Companies reporting service disruptions from a single region's failure on October 20, 2025.
1
Number of points of failure it took — in the cloud, and potentially in a concentrated portfolio.
The Portfolio Parallel

Your portfolio may have its own us-east-1.

For the utility employees and retirees I work with, the single point of failure is often easy to name: the company stock. Decades of payroll deductions, 401(k) matches, and loyalty accumulate into a position that would fail the same architecture review Amazon is now conducting on itself.

For index investors, it's subtler. Owning "the market" through an S&P 500 fund feels like owning five hundred companies. Functionally, an outsized share of the risk runs through the same few mega-cap names — the same way a thousand different apps all ran through one database in one region. When those names stumble together, diversification that looked real on a statement stops behaving that way.

The lesson of October 20 is not that the cloud is fragile. It's that dependency is invisible until it fails — and that the time to map your dependencies is before the outage, not during it.

✓Map the dependency. What single stock, sector, or theme would hurt you most if it fell 50%? If the answer is "I'm not sure," that is itself the answer.
✓Check the overlap. Company stock, index funds, and tech-heavy growth funds often hold the same underlying risk three different ways. Overlap is concentration wearing a disguise.
✓Build a second region. Amazon's answer to us-east-1 is multi-region redundancy. A portfolio's answer is genuine diversification — assets that don't fail for the same reason at the same time.
✓Rehearse the outage. Stress-test the plan against the bad scenario while it's hypothetical. The middle of a failure is the worst possible time to design your recovery.

Amazon fixed its outage in a day. A retirement portfolio built on a single point of failure doesn't get a same-day recovery.

Find Your Single Point of Failure — Before It Finds You

A concentration review takes about an hour: what you own, where it overlaps, and what a failure in your largest dependency would actually do to your retirement timeline. No obligation — just the architecture review your portfolio has probably never had.

Schedule a Concentration Review
Bailey Financial Services, Inc.  ·  Watkinsville, Georgia  ·  Wilder@BaileyFS.net

Bailey Financial Services, Inc. is a state-registered investment adviser. Registration does not imply a certain level of skill or training. This page is educational in nature and does not constitute individualized investment advice. All investing involves risk, including the possible loss of principal. Index concentration figures are approximate and change over time. Past performance does not guarantee future results.