The question comes up constantly now. Billions of dollars are flowing into AI companies. Valuations look disconnected from revenue. Every product has an AI feature whether it needs one or not. History suggests this ends badly. But does it?
The honest answer is more nuanced than either the hype or the scepticism suggests. Here is a clear-eyed look at what the AI bubble argument actually claims, where it is probably right, and where it misses something important.
The Case That a Bubble Exists
The numbers are genuinely extraordinary. Tens of billions of dollars have been invested in AI infrastructure, models, and applications in the space of a few years. Many of the companies receiving that investment are pre-revenue or generating revenue that is a small fraction of their valuation. The ratio of capital to proven business value is historically a warning sign.
The competitive dynamics also look concerning. Many AI products are built on top of the same underlying models. If your product's competitive advantage is primarily that it uses GPT-4 or Claude, that advantage disappears the moment your competitors add the same model. Differentiation through AI capability alone is hard to sustain.
There is also a real question about whether the revenue will ever match the infrastructure cost. Training and running large language models is enormously expensive. Many AI products charge less per user than it costs to serve that user. That works as a growth strategy when capital is cheap and unlimited. It stops working when investors start asking when the business becomes profitable.
Where the Bubble Argument Is Probably Right
Most AI startups will not survive. This was true of internet startups in 1999 and it will be true of AI startups now. The majority of companies receiving investment in a bubble environment are not building durable businesses. They are building in the direction the money is pointing.
AI valuations at the infrastructure and model layer are almost certainly inflated relative to near-term revenue potential. The cost of building and running frontier models is high, the competition is intense, and the customers willing to pay premium prices for access to the best models are a smaller market than the valuations imply.
There will be a correction. There always is. Whether it looks like the dotcom crash or something more gradual is hard to predict, but assuming that current valuations across the AI sector are rational is probably not a safe assumption.
Where the Bubble Argument Misses Something
The dotcom crash destroyed thousands of companies but it did not destroy the internet. Amazon lost 90% of its value and survived to become one of the most valuable companies in history. Google was founded during the crash and built one of the most profitable businesses ever created. The bubble popping did not mean the underlying technology was wrong. It meant the capital allocation was wrong.
AI is genuinely different from most previous technology hype cycles in one important way. The productivity improvements are real and measurable right now, not speculative future possibilities. Developers using AI coding tools are meaningfully more productive. Writers using AI assistants produce more output. Customer support operations using AI handle more volume with fewer people. The value being created is not theoretical.
That separates AI from technologies like VR and the metaverse, where the promised applications remained stubbornly in the future. AI tools are being used today by real people doing real work and producing measurable results.
What Happens If the Bubble Bursts
The most likely outcome of an AI bubble correction is not that AI goes away. It is that capital becomes more selective. The companies building real products with real revenue and real differentiation survive. The companies that were essentially riding the hype wave without a defensible business model do not.
Model providers that are spending more on compute than they are earning in revenue will face pressure to raise prices, cut capacity, or find a path to profitability. Some will manage it. Others will not. This could mean some of the AI APIs that current products depend on become more expensive or less reliable.
For founders building on top of AI capabilities, this is a real risk worth thinking about. Products that are entirely dependent on a single AI provider are exposed if that provider changes its pricing or terms. Building in flexibility about which models or providers you use is a reasonable hedge.
What This Means for Founders Right Now
The tools are real and they are genuinely useful. Using AI coding assistants, AI-powered features, and AI-driven workflows to build faster and operate more efficiently is not speculation. It is the current reality and it is likely to continue regardless of what happens to valuations.
What founders should be cautious about is building a business where AI capability is the only differentiator. If your product's value proposition is "we use AI to do X," and X can be replicated by any team that adds the same AI model, you do not have a durable business. The founders who will do well through whatever correction comes are those building products where AI enhances a fundamentally sound value proposition rather than being the value proposition itself.
The bubble, if it bursts, will clear out the noise. The products that solve real problems for real customers will be fine. That has been true of every technology cycle before this one and there is no reason to expect this one to be different.
The Honest Bottom Line
Some version of a correction is likely. Many AI companies will not survive it. The underlying technology will continue to develop and become more capable and more embedded in how work gets done. The founders building real products on solid foundations have less to worry about than the headlines suggest.
Focus on the problem you are solving and the customers you are serving. Use AI where it makes your product genuinely better. Do not make AI the entire story. That advice holds whether the bubble bursts tomorrow or keeps inflating for another three years.
If you are building a product and thinking through how to position it well for whatever comes next, we are happy to think through it with you at Cystall.