Cloud Computing In 2025 Is Broken

Cloud platforms were built for a world where software was big and slow. You had a frontend, a backend, a database. You shipped one thing. You paid for one thing.

That world is gone.

AI changed everything.

Now a solo dev can vibecode ten tools a week, hell, probably 10 per day. One to scrape docs. One to automate your relationship. One to monitor uptime. One to chat with your CRM. None of them are "apps" in the traditional sense. They are workers. Scripts. Bots. Agents. Call them what you want. Ephemeral services that live for a day or two.

The cost of writing software has collapsed and is still going down.

But hosting? Still expensive. Still rigid. Still built for a world that doesn’t exist anymore.

Cloud Economics Did Not Evolve

Most platforms still charge you per app or per service. Heroku, Render, DigitalOcean.

That was fine when you had a few big services.

But if you want to launch 50 tools, the costs explode. Not because you're using more CPU. But because the platform charges you as if each tool is an entire company.

This model punishes velocity.

This model punishes experimentation.

And that makes it incompatible with the way AI-first developers work (also called vibecoders).

Some Platforms Tried

Fly.io made a bet on per-usage billing and auto-shutdowns. It brought serverless-style pricing to containers. A smart move. But their product had to pivot multiple times, and they are still struggling to find their place. (That is at least my interpretation of their product.)

Railway is pretty good at letting you scale to zero AND having usage based billing. It helps. But if you have many services running even occasionally, the costs stack up fast. And if you have many services running all the time, you are still paying for idle resources.

These platforms saw the shift. They responded. But most are still trapped by old assumptions. Or VC incentives. Or both.

Why Serverless Isn’t the Answer Either

On paper, serverless sounds like the perfect fit. Scale to zero. Pay per request. No idle costs. Yuhu!

But in practice? It’s harder than it looks.

A good serverless architecture means splitting code into tiny parts. Managing AWS Lambda configuration hell. Wiring up APIs. Coordinating queues. Dealing with observability in a maze of functions.

If you're a big company with a platform team, that’s fine. If you're vibecoding your way through AI-generated tools every night? It’s just overhead and most vibe-coders wouldn't get it even if they tried (not entirely their fault)

And LLMs don’t think in Lambdas.

They think in “what’s the shortest path to shipping something that works.”

Which is usually a monolith, not some "scale to infinity" infrastructure.

Why This Isn’t Just Microservices All Over Again

If this sounds a bit like the microservices hype of the 2010s, here’s the difference:

Microservices were a solution to organizational complexity. They let teams move independently. But they also required infrastructure, coordination, contracts, DevOps.

The cost was worth it if you had 500 engineers. Not if you’re solo or in a 3-person team.

This isn't microservices. This isn't serverless. It's a new mode of building where software is cheap to create, fast to throw away, and built by people who don’t want to think (and understand) about infra. The tooling and pricing models need to reflect that.

The Real Problem: Nobody Wants to Run Servers Anymore

Sure, you could rent a $5 VM. That’s the dream. Full control. Maximum flexibility.

But fast builders (read, vibecoders who don't know what they are doing) don’t want to manage firewalls or logs or SSL. They don’t want to remember which ports are open. They don’t want to maintain anything.

They want to paste some code, hit deploy, and forget.

So if you ask me, we’re left with a strange gap.

• Serverless is too fragmented
• PaaS is too expensive
• Self-hosting is too manual

This new wave of builders needs something in between.

A Better Model: Pay for Capacity, Not Units

Instead of paying per app, what if you just rented a box?

One price. As many services as that box can handle. No extra charges for being creative.

That pricing model changes everything. It turns the cloud into a sandbox. You’re no longer penalized for building small, disposable tools. You’re rewarded for doing more, because the price per app goes down with every app.

Some platforms are starting to go that route. That includes us at Sliplane. But this is bigger than us.

(I think this is a big reason why Coolify is also experiencing the boom it currently does)

This is a change in how software is made.

And most of the industry isn’t ready.

The Fork in the Road

VC-backed platforms need high ARPU. They need predictable, scalable revenue. That means charging per app. It means charging for every container, even if it sleeps.

But vibecoders don't care about shareholder value. They just want to build.

AI encourages chaotic creativity. The best developers lean into that. They generate. They tinker. They throw away. They rebuild.

The cloud needs to get out of the way.

And the only way that happens is by changing how we price and think about hosting entirely.

Closing Thought

We’re entering a new era. The unit of software is no longer "an app." It’s a swarm of vibecoded disposable things.

If your hosting model doesn't reflect that, you're not ready for what's next.

— Jonas Co-Founder, Sliplane

PS: Yes, I am aware how biased I am and that I would benefit from this becoming the universal truth :)