Amazon's $200B AI Bet Is Breaking AWS Reliability — What That Means for Every No-Code App

Amazon's free cash flow just fell off a cliff. From $26 billion to $1.2 billion in twelve months. A 95% collapse. Not because the business is shrinking — AWS revenue is actually up 28% year over year, its fastest growth since 2021. The cash is vanishing because Amazon is spending it faster than it comes in. $43.2 billion in Q1 alone. On pace for $200 billion across 2026. Almost all of it going into AI infrastructure: data centres, networking, and custom silicon.

Then, on July 16, US-EAST-1 went down. Not a brief blip. A multi-service outage that stretched beyond 15 hours, with cascading failures across DynamoDB, Lambda, and EC2. Services that tens of thousands of applications depend on. Including, almost certainly, whatever no-code platform you're building on.

These two stories are connected. And the connection matters more for no-code builders than for anyone else.

Why reliability suffers when you're building at this scale

Amazon isn't cutting corners. They're building as fast as the supply chain allows, and Andy Jassy has been explicit that most new capacity is spoken for before it comes online. The problem isn't underinvestment. It's that building at this pace creates operational fragility that even Amazon's famously robust infrastructure can't fully absorb.

When you're simultaneously expanding data centre capacity across multiple continents, rolling out custom silicon (Trainium, which Jassy claims would be a $50 billion ARR business if it were standalone), and maintaining the existing global infrastructure that runs a significant chunk of the internet, something gives. In this case, it was DNS management for DynamoDB in the most critical AWS region.

The May 7 overheating incident at a US-EAST-1 data centre was a warning shot. Servers shut themselves down. Power was lost. The July 16-17 outage confirmed the pattern. AWS is becoming less reliable as it scales AI infrastructure faster than any cloud provider has ever scaled anything.

Long-term debt tells the same story from a different angle. It jumped from $65.6 billion to $119.1 billion year over year. Amazon is essentially betting its balance sheet that AI infrastructure demand will justify the spend. If that bet is right, Amazon owns the next decade of compute. If it's wrong, the consequences cascade through every service that runs on AWS.

What this means for your no-code app

Most no-code platforms run on AWS. Bubble, WeWeb, Xano, Bildr — their infrastructure sits on top of AWS services. When DynamoDB goes down, your Bubble app's database might go down with it. When Lambda has increased error rates, your Xano backend functions might fail silently. You don't get an alert. Your users get a broken experience.

You're not going to fix this at your layer. But you should understand your exposure.

Ask your platform: what's the failover story? Do they run multi-region? If US-EAST-1 has a bad day, does your application keep working or do your users see error pages? Most no-code platforms run single-region because multi-region architecture is expensive and complex. That's a reasonable trade-off for most use cases. But if your application is business-critical — if downtime costs you real money — you need to know whether "single region" means your entire app is one AWS outage away from being unavailable.

Also worth checking: does your platform use DynamoDB? It's the service at the centre of the July 16-17 outage. Platforms that rely heavily on DynamoDB for session management, real-time data, or caching were disproportionately affected.

The counterintuitive argument for governed platforms

Here's where things get interesting. Platforms that abstract the infrastructure layer — the ones that handle hosting, scaling, and reliability for you — actually reduce your exposure to AWS outages. Not because they're immune. Because their engineering teams are the ones getting paged at 3 a.m., not you.

Stacker, for instance, manages the infrastructure so you don't have to think about regions, failover, or database replication. If AWS has a problem, Stacker's ops team deals with it. That's not a guarantee of zero downtime. It's an argument that infrastructure abstraction is worth more when the underlying infrastructure is demonstrably less reliable than it was two years ago.

The same logic applies to platforms like Retool and Airtable. They sit on top of cloud infrastructure but insulate you from most of its operational complexity. When AWS sneezes, these platforms catch the cold so your users don't have to.

This is the bit that gets lost in the "build vs. buy" conversations. Building on raw AWS gives you maximum control. But it also gives you maximum exposure to AWS's reliability problems. Building on a governed platform means you trade some control for operational insulation. In 2026, with AWS reliability demonstrably wobblier than it was in 2024, that trade looks smarter than it used to.

The bigger picture: cloud isn't a utility, it's a bet

We've spent fifteen years treating cloud infrastructure like electricity. Plug in, it works, you don't think about it. The AI buildout is breaking that assumption.

Amazon's $200 billion capex programme isn't just big. It's distorting. The company is spending more on infrastructure in one year than most countries spend on their entire defence budgets. The combined hyperscaler capex for 2026 — Amazon, Microsoft, Google, Meta — is projected at $610 billion, up 70% year over year. That's not normal growth. That's a land grab.

The supply constraint is real and it affects you now. OpenAI CFO Sarah Fryer described the current situation as "a vertical wall of demand with compute being the bottleneck." GPU rental prices are up 40% over the last six months. When Amazon says they're building as fast as they can, they're implicitly saying that if you need capacity today, you may not get as much as you want at the price you expect.

For no-code builders, the takeaway isn't "avoid AWS." That's not practical, and it's not the point. The takeaway is that cloud reliability is going to be choppier for the next 18 to 24 months as hyperscalers run their infrastructure hot while simultaneously building at unprecedented scale. Choose platforms that give you operational insulation. Understand your single points of failure. And for business-critical applications, know what happens when — not if — the next US-EAST-1 outage hits.

Cloud isn't a utility any more. It's an AI infrastructure arms race. And the ricochets land on your uptime.