In the late ’90s, India fell in love with the internet. Neon-lit cybercafés were everywhere; companies were adding “.com” to their names to double valuations with little more than a URL and a pitch deck; investors believed they were sitting on a gold mine, dreaming of a future where clicks would replace commerce. Valuations soared. Startups with no revenue—and often no product—were going public. Then came 2000: the music stopped, and the mirage cleared. Trillions of dollars and inflated dreams evaporated.
The dot-com bust didn’t kill the internet—it killed the promise of an instant digital future. Instead, it cleared the runway for sustainable adoption. Out of that rubble rose Google, Amazon, and the foundations of the modern digital economy. Reality reset, driven by those who deliberately built for the future. The lesson is clear: most technology revolutions don’t arrive with a bang. They arrive disguised as slow, frustrating, incremental change.
Hype cycles around technology are not new. Our amnesia around them seems to be. I first noticed AI entering enterprise use in deployable forms around 2013—embedded in chatbots, predictive analytics, and fraud detection systems. Years later, we have arrived at generative AI, a tremendous leap: systems that can create content, design workflows, write code, and even offer decision-making guidance. ChatGPT took the world by storm not just because it was powerful, but because it was accessible. Now, everyone can play. That level of unfettered access changed the conversation almost overnight.
Artificial intelligence offers alluring dreams—promising to write better code, replace customer support, and even create unicorns run by a single founder and a dozen agents. But the enterprise has yet to catch up with the dazzling headlines. Business-wide AI adoption is slower and more complex. It is often uneven and fragmented, with AI labs operating in isolated silos. This is not about plugging in a new tool; it is about rewiring an entire system—shifting how we work, learn, and even think.
The question is whether AI will crash like the dot-com bubble. The probability, I would argue, is around 33 per cent—not inevitable, but very possible. This is not alarmist. The ingredients are familiar: soaring valuations, inexperienced startups, overhyped demos, bold headlines, and investors chasing the next miracle without understanding the technology or its use cases. Many organisations remain stuck in pilot mode, with prototypes failing to translate into scaled impact—hobbled by incomplete data, a missing trust layer, and largely analog processes.
We are already seeing startups valued at hundreds of millions, even billions of dollars, without proprietary models, differentiated data, or clear customer traction—experiments riding an open-source wave. A crash, should it come, may not be such a bad thing.
What would it look like? AI budgets quietly shrinking; overfunded startups beginning layoffs or pivoting to “enterprise SaaS”; LLM enthusiasm tempered by cost, quality, and compliance constraints; boards starting to question the return on investment. What it might leave behind are the builders—teams solving real problems, quietly laying the foundations for AI’s true revolution. The conversation could shift from what AI could do to what it should do. By clearing distractions, we may emerge stronger, setting the stage for durable innovation.
The way forward is not panic, but preparation. The iceberg may be real, but that doesn’t mean the ship must sink. This phase does not signal failure; it signals pruning—replacing vanity projects with meaningful applications that address real challenges, whether making society more equitable, improving access to quality education and healthcare, or strengthening governance.
Many promising AI initiatives are suffocated by procurement delays, data-security concerns, and turf wars between functions. Those that succeed share rare traits: executive sponsorship, shared ownership, and a relentless focus on solving genuine user problems.
The real challenge lies in the process. We must move beyond trying to fit AI into legacy workflows and instead redesign those workflows entirely. Success comes from letting go of processes that no longer serve us and using AI as the enabler, not the ornament.
What is missing? Today, AI largely helps companies do what they already do—just faster: summarising meetings, writing emails, boosting productivity. It rarely creates new markets. These tools remain siloed, and no one has yet cracked the full stack—from data ingestion and modelling to decisioning, compliance, and user experience. With AI infrastructure still evolving, achieving scale remains difficult.
It is time for companies to unlearn, reimagine, solve one real problem better than anyone else, focus on workflows, prove retention, and understand precisely which system they are replacing.
The internet didn’t fail. We misunderstood and misused it. I doubt AI will fail either. The opportunity ahead is enormous—but so is the temptation to get carried away. The lesson from 2000 is simple: technology survives. Hype doesn’t.
Vineet Nayar is former CEO of HCL Technologies and founder-chairman of Sampark Foundation.