By Pauline Shangett, Chief Strategy Advisor at NOWNodes
On October 20, 2025, the internet faltered. For hours, countless apps, platforms, and services simply stopped working. Fortnite froze. Snapchat crashed. Alexa went silent. Even major banking and trading apps were down.
The cause wasn’t a cyberattack or a hack; it was a routine software update gone wrong in Amazon Web Services’ US-EAST-1 region, one of the most relied-upon pieces of digital infrastructure on the planet.
A small configuration change led to DNS failures that rippled across the global web, breaking everything from gaming to financial services to parts of the crypto ecosystem.
It was a moment of silence and a reminder that the “cloud” isn’t some ethereal, distributed network. It’s a collection of data centers owned by a few companies. And when one of them sneezes, the internet catches a cold.
For most of the world, the outage was an inconvenience. For Web3, it was an existential warning.
Summary
Centralized convenience, decentralized illusions
The modern internet runs on convenience. Platforms like AWS, Google Cloud, and Azure have made it much easier for companies to scale. Startups don’t need to buy racks of servers or run their own data centers anymore. You just pay for what you use, get an instance running in a few minutes, and can focus on building your product rather than dealing with hardware.
But this convenience comes with a catch: reliance. When your infrastructure lives somewhere else, you’re putting a lot of trust and control into someone else’s hands. The Web3 space, the world’s loudest advocate for decentralization, still leans heavily on centralized infrastructure. Many DApps, RPC endpoints, wallets, and validator nodes run on the same few providers, often in the same regions.
If a single cloud region fails, entire “decentralized” ecosystems grind to a halt. The irony is painful: we’re decentralizing finance and governance but centralizing the servers that keep them alive. When AWS goes down, it’s not just a matter of downtime; it also damages trust. If a decentralized system can’t withstand a single point of failure, can it really be called decentralized?
The hidden cost of centralization
Centralized infrastructure concentrates not only risk but also control. Cloud providers can, and do, throttle, suspend, or reprice services at will. They operate as invisible intermediaries with the power to affect everything from latency to liquidity.
For years, cloud computing was cheaper and more flexible than owning hardware. But as the “Big Three” clouds consolidated dominance, the market began to look less like innovation and more like oligopoly.
In 2024-2025, AWS compute costs increased by over 20%, with nearly 40% of companies reporting bill spikes exceeding 25%. The same services that once enabled startup agility now punish success with unpredictable scaling fees.
And when your product’s uptime and financial runway depend on a single provider’s business model, you’re not in control. You’re a tenant.
Hardware returns, not as nostalgia, but necessity
Owning servers might sound outdated, but in 2025, it’s becoming a strategic advantage.
The math isn’t too complicated. A physical server costing about $1,100, spread over ten years, comes out to roughly $110 a month. Compare that to cloud computing at scale, which can easily hit $2,000 to $7,000 a month. But the real benefit isn’t just about money. When you run your own hardware… well, you’re really the one in charge. You get to decide where the data lives. You figure out how redundancy should work. You can tweak things for speed, for security, or for compliance. No waiting for a cloud provider to roll out some new feature. No dealing with their API limits. You just do it your way.
And importantly, your service doesn’t vanish because one cloud region had a bad morning. Owning your infrastructure doesn’t mean going fully offline or building bunkers full of machines. It means designing for distribution, spreading your systems across providers, geographies, and hardware models so no single failure can take everything down.
So it’s not “cloud vs. metal.” It’s about control vs. fragility. Clouds will fail. Hardware can fail. But when systems are distributed, redundant, and supported by real engineers who understand failure as an inevitability, the overall architecture becomes antifragile.
Designing for failure: the distributed model
At NOWNodes, we design infrastructure around one assumption: failure will happen. That’s why our architecture is globally distributed across the European Union, the United States, and Asia, with data centers in Germany, Finland, the Netherlands, the U.S., and Singapore. Each location is selected not just for network performance but for political stability and operational safety.
Every critical system follows a 2N+1 redundancy model. That means if one system fails, another takes over instantly. If two fail, traffic still routes through the third. Downtime isn’t “avoided”; it’s absorbed.
We also test failure on purpose. Our engineers run controlled outages in mirrored environments to identify weak points before they break in production. Resilience doesn’t come from assumptions, it comes from actually practicing it. And it’s not just about machines; infrastructure involves people, processes, and a lot of little things working together.
When the AWS outage hit, thousands of teams scrambled in panic. Who could they call? What could they do? In Web3, even a short downtime can freeze transactions or lock people’s funds. That’s why we don’t just rely on chatbots or long ticket queues. Our engineers are on call all the time on Slack, Telegram, and live chat and usually respond in just a few minutes.
Most issues are resolved within hours, not days. Because resilience isn’t just about systems staying up; it’s about people staying present when they go down.
Infrastructure is trust
Infrastructure is rarely visible until it fails, right? Users don’t think about where their transactions are processed or how their wallet connects to a blockchain until it stops working. And when that happens, trust erodes instantly.
The AWS outage was a reminder of that invisible trust relationship between platforms and users. Even “smart” devices couldn’t escape it. There was a viral post about a smart bed that stopped working because AWS went down. Sounds ridiculous, but it’s actually a perfect metaphor. The more connected and “smart” our world gets, the bigger the mess when centralized systems fail.
That’s why decentralizing infrastructure isn’t just about ideology; it’s about functionality. It’s about ensuring your product, your blockchain, or your “smart bed” keeps working when the internet’s biggest provider takes a nap.So yes, use NOWNodes, and your smart bed won’t crash when AWS does.
