With over two decades in tech, I'm increasingly concerned about unfulfilled AI investment promises. Many will not materialise, leading to industry consolidation. This post explores the parallels with the dotcom bubble and why a gradual 'deflation' is likely.
AI Investment Promises: Consolidation on the Horizon?
As a developer with over two decades in the industry, I've been watching the AI landscape with increasing concern. The current fervour surrounding Artificial Intelligence, particularly Large Language Models (LLMs), has led to a surge in promised investments. However, I'm seeing a worrying trend: many of these investments seem unlikely to materialise, paving the way for significant industry consolidation.
Having witnessed the dotcom era firsthand, I see undeniable parallels. But I also see fundamental differences that suggest the outcome this time might be less of a 'pop' and more of a 'deflation'. Let’s explore the similarities, the crucial distinctions, and what I believe will remain when the hype finally settles.
Echoes of the Dotcom Boom
The most striking similarity is the investment pattern. We're seeing companies with minimal revenue, and sometimes questionable business models, achieve staggering valuations. The focus is on potential, on capturing a slice of a future market, rather than on current profitability. Many companies seem to be over-hyping their abilities, making bold claims about how their AI will revolutionise an industry without a clear, demonstrable path to generating that income. This was the hallmark of the dotcom bubble: sell the dream, get the funding, and figure out the business model later.
Just like in the late 90s, there's a gold rush mentality. Everyone is scrambling to integrate 'AI' into their product name and marketing materials, whether it's a core feature or a superficial addition. The fear of being left behind is palpable, driving a cycle of hype and investment that can feel disconnected from tangible value.
But This Isn't 1999
Despite the similarities, the technological landscape is fundamentally different. The dotcom boom was built on a nascent internet, with companies investing massive capital into their own physical servers and infrastructure. Today, we have a mature, cloud-native ecosystem. With platforms like AWS, a small team can spin up a globally scalable application in an afternoon. This drastically lowers the barrier to entry.
This accessibility is a double-edged sword. While it allows for rapid innovation, it also makes it easy to create what I call 'thin wrappers'. Many new 'AI' companies are essentially just putting a pretty user interface on top of an API from OpenAI, Google, or Anthropic. They aren't creating their own value; their success is entirely tied to the success and failure of the model they rely on. This is a critical red flag.
The Sobering Reality: Why We Haven't Seen Transformative Change Yet
For all the hype, many businesses are struggling to see the transformative return on investment that investors expect. In my own experience building AI-integrated workflows, like an AI Powered Event Management CMS or a Context Aware AI Guidance System, I've run into the same core obstacles time and again.
The Legacy System Bottleneck
The biggest blocker is often not the AI itself, but everything it needs to connect to. Many established businesses, especially in older industries, are running on legacy systems. These monolithic applications, sometimes 15 or 20 years old, were never designed with modularity or APIs in mind. Integrating a modern, LLM-based tool often requires massive, expensive rewrites of systems that, from the business's perspective, 'work fine'. They are understandably hesitant to invest heavily in overhauling a critical system for a technology they don't fully understand.
The Knowledge Gap in Leadership
This leads to the second major hurdle: business leaders often don't grasp the technology's limitations. They hear about AI's incredible capabilities but don't see the complex, unglamorous engineering work required to make it useful. They don't understand the costs, the 'hallucinations', or the need for constant prompt engineering and refinement. This disconnect between expectation and reality creates friction and stalls projects before they can deliver real value.
A Deflation, Not a Pop
So, will the bubble burst? I believe a sudden 'pop' is less likely than a gradual deflation driven by unmet expectations and subsequent industry consolidation. The market will eventually demand real, sustainable value, and when it does, many companies will be exposed, not through a dramatic crash, but through a slow decline in funding and increasing acquisition by larger players.
We're already seeing signs of this. Funding rounds are becoming smaller and more scrutinised. Investors are asking tougher questions about profitability and long-term viability. The era of easy money for AI startups is coming to an end. This isn't necessarily a bad thing. It's a sign that the market is maturing and becoming more discerning.
Here are the red flags I look for:
- Thin Wrappers: As I mentioned, if a service is just a skin on a major model, it has no defensible moat. When you evaluate a tool, check for signs like consistently high latency or a lack of domain-specific accuracy. These often indicate the service has little control over the underlying model. Furthermore, these services operate on razor-thin margins, as the cost of inference from the foundation model provider can easily outweigh the value they add.
- Over-promising: We've seen examples like Builder.ai, which made grand claims that didn't align with reality. Companies that promise the world are often hiding a weak foundation.
- The 'Wizard of Oz' Method: Some services use low-paid human workers to mimic AI, a model AWS Mechanical Turk has long exemplified. This is not a scalable or honest approach to building an AI product.
In this environment, I see a parallel in how different AI leaders are positioned. A giant like OpenAI is the clear technology leader, but its route to market is deeply tied to Microsoft's ecosystem. This gives it immense reach but also creates a complex codependency. This contrasts with Google, which controls the entire stack from silicon to model to application, and a company like Anthropic, which is carving out a niche with a strong focus on safety and reliability, making its value proposition to certain enterprise customers clearer and more direct.
As investor patience wears thin, the companies built on pure hype will fail or be acquired for their talent. The companies with solid technology, clear use cases, and a real plan for integrating with the messy reality of legacy systems will survive and thrive.
The Foundations That Will Remain
When the dotcom bubble popped, it was devastating for investors. But it wasn't the end of the internet. It was a painful but necessary correction that cleared away the noise. The companies that survived, like Amazon and Google, went on to define the next two decades of technology. The infrastructure that was built laid the foundation for everything we use today.
I believe the same will be true for AI. This period of intense hype, for all its irrationality, is driving incredible progress in foundation models and tooling. When the dust settles, we will be left with a powerful new set of building blocks. As a developer, my focus remains on using these blocks to solve real problems and deliver tangible value. That's the core philosophy behind my AI Engineering Playbook, and it’s the only sustainable path forward in a landscape prone to bubbles.