Over-investment in AI is one thing, its impact on the world quite another
Sceptics see an AI bubble, but we must tell signals apart from the noise. Profits lag at the consumer-facing end while value flows to players in control of AI building blocks—chips, data centres, etc. Focus on the transformation underway, not the investment binge.
Every major technological boom attracts sceptics. Today, that chorus is growing louder around artificial intelligence (AI). Critics are right about one thing: there is excess. Accounting tricks, circular investments and heroic revenue projections all deserve scrutiny. But the mistake is to assume that financial excess equals technological emptiness.
To confuse the investor story with the innovation story is to misunderstand how value is now created and where it truly resides.
Noise versus the signal: The financial system around AI contains a rising number of excesses that would have been unacceptable to regulators or markets in bygone eras.
Capital moves in loops: a large company invests billions in a partner, at times with warrants, and the partner spends those billions purchasing the wares of the investor. The same transaction gets recorded as an investment in a company with revenue growth for one, and as revenue for the other. Business carries on, even if sticklers pen doomsday scenarios. It can go on for years.
Yes, this is froth and will likely cause a deep cyclical reversal. But dismissing AI because of financial froth is like shrugging off the internet in 1999 because an electronics chain went bankrupt. The signal is not in the hype. It lies in the quiet, irreversible shift beneath it.
That shift concerns where value now resides. Many still measure the success of AI by the profitability of its application layer of chatbots and work assistants, and declare the sector unviable when those companies fail to show profits.
Yet, AI’s value creation no longer mirrors the old digital model where front-end firms captured most of the value. In this new era, the application layer is a conduit, not a reservoir. Value flows upstream to players that provide AI enablers such as chips, memory, networking systems and data infrastructure.
Where value really flows: The misunderstanding runs deep. People see consumer apps and measure their worth by outdated notions of margins and returns. What they miss is a system-level transformation, the steady reallocation of value towards those providing scarce capability. Compute capacity, precision manufacturing and data engineering have become the substratum for digital skyscrapers. To judge AI’s worth by the profit statements of its chatbots is to miss the tectonic shift in the global value chain.
For decades, software captured most of the value generated, while hardware was treated as a commodity. That has changed. The strongest pricing power lies with those in control of high bandwidth memory, advanced packaging, optical interconnects and large-scale GPU clusters. These are the new choke points of the modern economy.
The rise of builders: None of this means AI is beyond critique. Many firms will fail, as they should. But the argument that AI cannot be profitable is both simplistic and wrong. The question is not whether one model company justifies its valuation today, but whether the system as a whole is rewriting the logic of productivity.
It already is. AI is transforming drug discovery, materials science and logistics, fields where complexity once defied human reasoning. We’re not just building faster tools, we are introducing new forms of cognition into the process of discovery.
Critics say AI cannot do your job but can convince your boss to fire you. A part of it is true: jobs of all types are under pressure. However, those societies and political systems that resist new-era innovations would cede ground to others that adopt AI without hesitation. Societal problems caused by new technologies cannot be wished away with simplistic assumptions that somehow new types of work will be created for all. But they cannot be eliminated by opting out either.
AI’s firm foundation: Every cycle of hype leaves behind productive infrastructure. The railway bubble left tracks. The dotcom bubble left fibre optics. The AI boom will leave behind a vast base of compute power, apart from data and foundational models. These assets will not vanish if valuations fall. They will be repurposed.
The greater misunderstanding lies in calling AI just another technology. It is not. What is unfolding is a mathematical revolution, a new way of representing and reasoning drawn from unstructured information. The leap from structured to unstructured data processing is as profound as the shift from analog to digital. It changes not only how we compute, but how we perceive and design. Let’s not trivialize the biggest human-capability shift since electricity.
To sceptics, every massive capex plan looks like a delusion because they assume that value must appear as profit in the same way as they used to experience previously. But in AI, value creation and value capture no longer move in sync. Consumers gain the benefits of AI advances immediately, while profits accrue slowly to those who build the enabling layers of scale and stay in the race even as weaker players drop out, leaving outsized gains for survivors. Here is a corporate ‘squid game,’ where nobody is playing fair and the police are missing.
There will be market corrections, no doubt, and they will hurt. Worry-bugs could opt out, hoping to catch falling knives in times of distress. But in perpetually reading a story of impending doom in the AI tea leaves, they shift the focus of attention away from a tech revolution that generations may discuss for centuries to come.
The author is a Singapore-based innovation investor for GenInnov Pte Ltd
topics
