It’s no secret that pricey Big Clouds like AWS, Azure and Google dominate cloud computing. Most cutting-edge AI apps are powered by these clouds, partially due to big-tech’s scale and their customer base (startups which have easy access to venture capital). In fact, there has been a data center construction boom caused by hyperscalers because of the AI bubble.

But what happens if and when the AI bubble crashes? Many AI startups will be priced out of the market. This is because, for now, they are supported by not only venture capital, but also the substantial account credit they obtain from Big Clouds if they’re venture capital backed.

But once the bubble crashes, the venture capital will dry out, and the hyperscalers will become too costly for startups. Surviving startups will be forced to optimize their models. Remember DeepSeek? They optimized their models after US trade sanctions prevented China from buying high-end Nvidia GPUs.

We’re conditioned to think that innovation happens when we have the latest and greatest tech. But in reality, the best innovation happens within constraints. Constraints make engineers seek creative solutions and optimizations. This is what happened in the noughties when dot-com companies were forced to move to low-end Dell and HP servers. How?

A Dot-com Lesson

In the 90s, most well-funded dot-com companies like Amazon and E*Trade used Sun servers. Sun servers were expensive. But since venture capital was easy and interest rates were low, Sun was the choice.

And it wasn’t just Sun. If you were a dot-com company, you also had to use Cisco routers and Oracle databases.

While Linux and MySQL would have worked just fine for many 90s dot-com companies, most startups wanted maximum access to funding and talent. Using Sun, Cisco and Oracle were the surefire path to both, because talent and funding were based on the tech stack.

In comparison, cash-strapped internet firms like small dial-up ISPs and web hosts ran Linux and FreeBSD due to the high prices of Sun servers. Despite the use of ‘low-end’ technology, they ended up with the technological lead.

When the dot-com bubble burst and interest rates rose, Sun servers, while available dirt-cheap on eBay, were no longer a long-term solution because the used server supply would ultimately dry out. Most surviving dot-com startups migrated to Linux on Intel.

And it wasn’t just Amazon and E*Trade who left Sun. Other web survivors like eBay followed suit. Subsequently, social media startups like Facebook, Twitter and YouTube ran on Linux instead of Sun.

While some 90s-era firms like Google always ran Linux, for others, using Sun became a liability. While porting your Perl back-end to Linux took time and money, before the rise of cloud computing, using Linux and Intel ended up saving server costs when a web company needed capacity.

Today’s AI Comparisons

AI is today’s frenzy. Companies are all-in on AI, despite current AI being considered unprofitable by many. But if and when the bubble bursts, cloud and AI optimization will happen the way optimization happened for the dot-com firms.

Outside of Big Tech itself, many mobile app and AI companies will abandon large clouds like AWS for lower-cost services like Fourplex or alternatively will run their own clusters. They’ll optimize their apps to run on any cloud, not just on Big Tech. This will then allow Tier 2 and 3 clouds to win contracts.

Yes, the current AI-induced RAM shortage hurts our short-to-medium term expansion. But if the shortage subsides, our long-term prospects are bright.

Web hosting and VPN providers already need Tier 2 and 3 providers, due to the per-gigabyte fees charged by Big Tech. When the AI bubble bursts, others will come on board, boosting a more competitive hosting industry in the process.