AI’s bubble trouble: Hype, myth and dotcom déjà vu

During my mid-20s, there was one popular advertisement that always fascinated me. Iconic Axe deodorant, promising a magical transformation the moment you spray it on. Suddenly, two gorgeous women would be irresistibly drawn to you. As a viewer seeking change in my life, I assumed that this was what I was missing. Of course, we discovered reality was far more mundane. Axe worked fine as a deodorant and did its fundamental job well, but obviously, the promised magic never happened.

This is how I am drawing parallels on how AI is currently being perceived. Many executives view AI as revolutionary magic that will transform business overnight; yet, most organisations are now discovering that, while AI will deliver good value, it is nowhere near the miraculous solution sold to them.

The AI position today aligns with the trend of the early days of the internet boom, when unprecedented investment drove the valuation of technology to new heights. As of today, we stand at a significant junction in technology, where, based on the current trajectory, we can be concerned that the AI trend is drawing parallels with the Dotcom bubble of 2000.

The current operating model of AI

As of Aug 2025, 78% of organizations have started using AI in at least one business area. The figure was 55% last year. Currently, the global AI market is valued at just over $233 billion and is expected to grow as machine learning is commanding 62% of AI investments. Enterprises are lining up to integrate AI across various IT operations (36% adoption), as well as marketing and sales, with 71% regularly using GenAI.

Most of those initiatives are primarily driven by AI hype built in the market and the organisation’s willingness to commit to AI in the near future. There is only one big issue here: “Problem of Scale”. Can these companies implement AI as promised and scale it for the entire organisation at the same efficiency? Currently, only 28% of enterprise applications are effectively connected with the AI ecosystem. This creates significant data silos, limiting the effectiveness of AI. Despite its widespread adoption, 74% of companies struggle to scale the execution of AI in production use cases. This highlights a substantial gap between the company’s ambition and ability to execute.

The cost of AI failures

With the ambition to be ahead of the technology curve, companies have already invested a significant amount in AI. But the failure rates of AI projects are staggering. 85% of AI initiatives fail to deliver the promised value of the project. This rate is nearly double that of regular IT projects. IBM lost $4 billion on Watson for Oncology, and in 2024, McDonald’s closed its AI drive-thru as it was persistently failing on the order process. The financial impact of scrapping these projects is severe. 42% of companies terminated most AI initiatives in 2025.  

The cost of a small AI automation project starts at $10,000 and can scale up to $10 million or more for enterprise solutions. Take a moment to absorb that OpenAI is projected to lose $5 billion in 2024 alone. Despite generating $3.7 billion in revenue from its subscription for ChatGPT, it was barely covering the $3 billion training cost of OpenAI models.

Drawing parallels with the dotcom bubble

This is where I am drawing a comparison between the dotcom bubble and the current AI hype. In 2025, 64% of US venture capital was invested in AI. OpenAI was valued at over $300 billion. This occurred even though they had registered losses for consecutive years before this valuation.  A similar concentration of investment was observed in the late 1990s. A significant amount of capital was flowing to dotcom companies that had unfinished projects and untested business models. Everything was being driven by future hope.

Today, we are seeing over 370 AI unicorns collectively valued at more than $1 trillion. The driving factor behind these startups is not business functionality, but rather technological promises.  According to the S&P 500, the AI sector accounts for 32% of the total market value.

Cost versus gain analysis

By the late 1990s, investments in NASDAQ in Dotcom companies were down 80%. But out of that financial disaster grew business giants like Amazon, Google and eBay. Those companies were among those that successfully melded innovation with traditional business practices.

Unlike traditional software companies, the cost structure of AI companies differs. AI gets its brain by training on an enormous amount of data. That training requires significant computing resources. The cost of OpenAI training alone exceeded their entire revenue stream. This unsustainable burn rate mimics the dotcom model, where unknown future potential is being prioritised over profitability.

The VC dependency problem

It’s not uncommon for companies to receive a boost from venture capital. However, in the case of AI startups, the pattern is slightly off. 87% of AI projects are concluded before they reach production, but funding continues to flow based on technological potential.

A similar pattern emerged during the dotcom era, when capitalists put business growth over profitability and companies burned through their funding without establishing sustainable revenue streams.

The recent funding round of OpenAI was the largest private tech deal ever, with a whopping $40 billion on the table. This happened when OpenAI was registering constant losses. AI companies have been valued at an unreasonably high level based on their future potential rather than their current performance.

Why AI expectations may face a reality check

Just like the dotcom bubble, the ongoing AI euphoria exhibits concerning signs. Inflated valuation, disconnection from business fundamentals, overoptimism among investors and limited evidence of future success are indications that AI fatigue is likely to set in as proof-of-concepts fail daily. The concentration of investment in AI has created a real systemic risk in the global economy. If key companies face setbacks, the entire sector is likely to experience a drastic correction.

As the backbone of AI is previously generated data, regulatory compliance is intensifying. Security of data and the Cost of maintaining such compliance are straining already struggling financial models. The EU AI Act and other similar frameworks are all set to impose burdens on startups operating on thin margins.

The future of AI by 2030

While bubble phenomena are concerning, no one doubts the transformative potential of AI. The global AI market is expected to reach $1.81 trillion by 2030. The adoption of AI is expected to become an economic driver, propelling global GDP to $15.7 trillion. Key sectors are expected to show growth as follows:

The healthcare industry is expected to generate a market of over $ 200 billion by enabling modern treatment patterns, early disease detection and personalising medication. The financial services market is expected to grow by $150 billion as fintech applications for fraud detection and risk assessment continue to be developed. The Indian manufacturing industry will upscale its adoption to 78% by leveraging AI for predictive maintenance and supply chain optimisation. Similarly, the automotive sector is also expected to see a rise of $400 billion market led by autonomous driving technology.

By 2030, 86% of businesses expect to undergo an AI transformation, and 70% of companies will be using AI in at least one business solution. Since we had already experienced the dotcom bubble, many companies have begun working on genuine business needs and established measures of KPIs for success.

Setting realistic expectations

AI progression is real, but concerns of the AI bubble are also not entirely mythical. The AI bubble is not predetermined to burst. Still, suppose companies want to avoid being part of another technology bubble burst. In that case, the key differentiator will be setting realistic expectations and focusing on a sustainable profit-making model, rather than relying solely on technological advancements.

Dotcom companies mostly lacked sustainable revenue streams, but successful AI implementation shows a clear value proposition and quantifiable returns. Regardless, companies must avoid the trap of chasing AI as mandated technology for the sake of being AI adopters; instead, they must focus on solving real-world problems with the technology solution that best fits their needs.

By the late 1990s, investments in NASDAQ in dotcom companies were down 80%. But out of that financial disaster grew business giants like Amazon, Google and eBay. Those companies were among those that successfully melded innovation with traditional business practices. They examined the problems first and utilized World Wide Web innovation to solve those problems, unlike many others who developed a solution and then looked for areas to apply it retrospectively. The AI revolution is real, but sustainable success requires an ambitious but realistic vision.

As we head our technology towards 2030, companies that equip AI with measured expectations, calculated investment and disciplined execution will realize its transformative value; a far better position than those caught in a speculative frenzy, likely to find themselves casualties of the next technology bubble.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?