AI’s Reality Check

It started with great excitement, the kind we have seen before whenever something new promises to change our lives. In tea shops, offices and online discussions, people spoke in awe about machines that could diagnose diseases, drive cars, analyse mountains of data, create art, write computer code and even talk back like humans. Companies rushed to show their Artificial Intelligence (AI) plans, investors poured in money, and share prices climbed rapidly, almost as if they could only go up.

Wall Street mirrored this optimism. US indices marched upward, powered by heavyweight technology names. Amazon, Microsoft, Nvidia, Meta and Tesla became shorthand for the future itself, while financial giants such as Visa and JP Morgan highlighted how deeply AI was penetrating payments, banking and risk management. The so-called ‘Magnificent Seven’ - Alphabet, Apple, Amazon, Meta, Microsoft, Nvidia and Tesla command a combined market capitalisation larger than the entire Chinese economy.

Then came the pause, stock prices corrected, funding became cautious. Soon, people started using a familiar word – ‘bubble.’  But before we rush to declare a crash and enjoy saying “we told you so,” it is worth pausing for a calmer look. What we may be seeing is not a collapse, but a sensible pause. In simple terms, it is the market taking a breath, separating big promises from practical progress.

Early euphoria

Every technological shift arrives wearing the borrowed clothes of history. The dot-com boom of the late 1990s promised a new economy and briefly delivered inflated valuations before crashing spectacularly. The housing bubble of the mid-2000s had wrapped excess in the comforting language of bricks and safety, only to expose the dangers of easy money.

Artificial intelligence, however, is a slightly different guest at the party. Unlike many dot.com firms that had websites but no revenues, AI already works. It translates languages, spots tumours, predicts supply chains, flags fraud and writes serviceable emails.

Markets, being emotional creatures, tend to price the distant future into the impatient present. In the last two years, expectations raced ahead of deployment. Every company presentation suddenly included an AI slide, often placed strategically between ‘vision’ and ‘growth.’ Investors rewarded ambition generously.

The recent cooling in US indices has been driven less by disappointment and more by arithmetic. Training large AI models is expensive. Chips are scarce and monetisation takes time. When quarterly numbers from even admired leaders such as Amazon, Microsoft or Tesla did not immediately match long-term storytelling, markets adjusted their spectacles.

This adjustment is being interpreted by some as a bubble deflating. Yet, corrections are the market’s way of asking better questions. Who will pay, how much, and for what exact value? These are not hostile queries. They are relevant ones.

History suggests bubbles burst when the core assumption proves false. The assumption behind AI - that intelligence can be automated in useful ways - has already been demonstrated. The uncertainty lies elsewhere: scale, costs and returns. How widely can AI be deployed? How quickly can expenses fall? Which sectors benefit first, and which resist longest?

The dot.com crash did not kill the internet. It killed weak business models. Amazon survived, pets.com did not. The housing crisis did not end home ownership. It exposed reckless lending. In hindsight, these episodes look less like endings and more like filters.

AI appears to be passing through a similar filter. Capital is becoming selective. Grand claims are being replaced by specific use cases. Instead of “AI will change everything,” the pitch is quietly shifting to “AI will reduce processing time by 25 pc”.

There is also a geographic angle. Much of the AI exuberance was priced in global markets, while adoption is unfolding unevenly. In countries like India, AI is less a luxury toy and more a productivity tool. Banks use it to detect fraud, farmers to forecast weather, startups to scale customer support.

Regulators, meanwhile, have entered the discussion - another sign of maturity. Debates around data use, bias and accountability are gaining momentum. Regulation is often dismissed as a drag on innovation, yet it can function as a steering wheel rather than a brake.

The real irony lies in our impatience. We demand revolutions to justify quarterly earnings and expect general intelligence to arrive by next Tuesday. When that fails to materialise, disappointment sets in. History tells a different story. Every transformative technology -electricity, automobiles, smartphones - passed through phases when investors doubted its economics and timing.

What we are seeing now is not a loud burst but a quiet recalibration. AI is shifting from promise to process.  Labelling this phase an ‘AI bubble’ makes for catchy headlines but ignores nuance.

(The writer is a retired Bengaluru-based banker. Views personal.)