Every day, millions of us open ChatGPT, Claude, Gemini, Grok and Perplexity.

We ask questions, write emails, generate images, analyse documents, create software and prepare business plans.

Most of us never think about what happens after we press Enter.

We see a simple text box.

Behind it is one of the most expensive technological systems humanity has ever attempted to build: massive data centres, specialised chips, cooling systems, power infrastructure, global networks and some of the world’s most expensive engineering talent.

And one thought keeps coming back to me:

What if the AI we enjoy today is cheap only because someone else is paying the real bill?

Investors are paying. Technology companies are paying. Shareholders are paying. Venture funds are paying. Some companies are borrowing billions, while others are making infrastructure commitments worth extraordinary amounts of money.

Everyone is betting that AI will eventually earn enough to justify today’s spending.

But what if it does not happen quickly enough?

The next great AI crisis may not come because machines become too intelligent.

It may come because intelligence becomes too expensive.

We Are Using a Ferrari and Paying Bus Fare

Think about what we receive today.

For a small monthly subscription—or sometimes completely free—we can access technology created using billions of dollars of investment.

Does that business model sound permanent?

Imagine someone offering you a private jet journey for the price of a bus ticket. Your first question should be:

Who is paying the difference?

The AI industry is currently fighting for territory. Every company wants users, developers and businesses to become dependent on its ecosystem.

The strategy is familiar:

First, make it exciting.

Then make it convenient.

Then make it essential.

Then change the price.

The real question is what happens when AI becomes so deeply connected to our work that we cannot easily walk away.

The AI Race Is Becoming a Financial Survival Test

The public sees a competition between ChatGPT, Claude, Gemini, Grok and Perplexity.

But the real competition is happening behind the screen.

Who can buy enough chips?

Who can build data centres?

Who can secure electricity?

Who can finance the next generation of models?

Who can serve hundreds of millions of users while continuing expensive research?

Who can survive if profits arrive five years later than investors expect?

AI investment by major technology companies is already measured in hundreds of billions of dollars annually, while longer-term infrastructure ambitions run into trillions.

The dangerous part is simple:

Every company is spending like it will be a winner. They cannot all win.

It is also important to distinguish debt from investment and operating losses. OpenAI faces enormous cash requirements and infrastructure ambitions. Anthropic depends on huge investor funding and strategic partnerships. xAI has raised both equity and billions in debt financing. Perplexity AI is backed mainly through venture investment.

Different financial structures. Same uncomfortable question:

How much money must be burned before artificial intelligence becomes sustainably profitable?

AI Does Not Live in the Cloud

The word “cloud” is misleading.

The cloud is a building filled with machines consuming enormous amounts of electricity and producing enormous amounts of heat.

AI does not live in the sky.

AI lives in data centres.

The future AI race may be decided not only by researchers and algorithms but by power generation, grid connections, transformers, cooling systems and semiconductor factories.

The strange truth is that the future of artificial intelligence may depend on extremely old-fashioned infrastructure.

A company may create the smartest model in the world, but if it cannot secure affordable power and computing capacity, intelligence alone will not save it.

Soon, AI companies may compete for megawatts as aggressively as they compete for engineers.

AI Agents Could Explode the Cost

Today, we ask AI:

“Write an email.”

One request. One answer.

Tomorrow, we may say:

“Manage my sales activity today.”

An AI agent could then research hundreds of companies, read thousands of pages, identify decision-makers, write personalised messages, update databases, analyse responses and schedule meetings.

I gave one instruction.

The AI performed 10,000 actions.

Now imagine billions of people and millions of companies using autonomous agents.

Legal agents.

Coding agents.

Financial agents.

Sales agents.

Research agents.

Cybersecurity agents.

Management agents.

The cost per AI action may fall dramatically, but the total number of AI actions could explode.

The winner may not be the AI with the highest intelligence.

It may be the AI providing the most useful intelligence per dollar, per rupee, per watt and per chip.

Will Ordinary People Be Priced Out?

Today, AI feels democratic.

A student and a multinational corporation can ask questions to the same system.

But I doubt that equality will last.

We may soon have different classes of intelligence:

Free AI for basic questions.

Affordable AI for everyday productivity.

Professional AI for programmers, doctors, lawyers and researchers.

Enterprise AI agents costing lakhs or crores per year.

The free AI may answer questions.

The expensive AI may run businesses.

This could create something more dangerous than the digital divide.

An intelligence divide.

Rich companies may not simply have more employees. They may have better artificial intelligence.

Powerful countries may have better models.

Wealthy students may have advanced personal tutors.

Rich investors may have AI systems continuously analysing opportunities.

AI may democratise basic intelligence while concentrating the most powerful intelligence in the hands of those who can afford it.

Maybe Only a Few Will Control the Real AI

I do not think thousands of AI companies will disappear.

But perhaps only three or four organisations will be capable of building the world’s most advanced frontier models.

The rest may build applications on top of them.

This is similar to cloud computing. Millions of businesses use cloud services, but only a handful operate global hyperscale infrastructure.

The same may happen with intelligence.

Thousands of AI products.

Hundreds of brands.

But underneath them, perhaps only a few real intelligence suppliers.

That concentration would create power unlike anything we have seen before.

The Wild Card: Someone May Make AI 90% Cheaper

There is one possibility that could destroy today’s assumptions.

What if someone discovers a radically simpler architecture?

What if small specialised models become good enough for most tasks?

What if new chips dramatically reduce energy consumption?

What if advanced AI runs locally on personal devices?

Then some of today’s massive data-centre investments could become technological monuments to an expensive transitional era.

The greatest threat to a company spending $100 billion may not be another company spending $200 billion.

It may be a small team discovering how to do the same thing for $10 billion.

The Part Nobody Wants to Discuss

The AI industry tells us a beautiful story.

AI will empower everyone.

AI will democratise knowledge.

AI will give every child a tutor, every worker an assistant and every entrepreneur a team of intelligent agents.

I hope that future arrives.

But history suggests another possibility.

First, AI will become essential.

Then the market will consolidate.

Then the strongest companies will control the infrastructure.

Then advanced intelligence will be divided into pricing tiers.

And finally, society may discover that we did not democratise intelligence.

We privatised it.

The frightening future is not necessarily a robot war.

It may be much quieter.

A child with free AI competing against a child with a ₹1 lakh-per-month AI tutor.

A small business with a chatbot competing against a corporation operating 50,000 autonomous agents.

A developing country depending on intelligence infrastructure controlled elsewhere.

A worker being evaluated by an AI system more powerful than any AI they can personally afford.

And governments becoming dependent on private companies that control models they do not fully understand and infrastructure they cannot easily replace.

That future would not look like science fiction.

It would look completely normal.

There would be subscription plans.

Corporate contracts.

Terms and conditions.

Monthly invoices.

And perhaps that is what should worry us most.

The most powerful technology in human history may not rebel against humanity.

It may simply become unaffordable to most of it.

We are worried that AI will become smarter than us.

Perhaps we should be more worried about who will own that intelligence when it does.

Because the final AI war may not be humans versus machines.

It may be something far more familiar:

Those who can afford intelligence versus those who cannot.

And by the time we realise that the world’s most powerful intelligence has become concentrated in the hands of a few companies and governments, we may already be too dependent on it to say no.

Perhaps the greatest danger in the AI revolution is not that Artificial Intelligence will fail, but that it will succeed after destroying the economics that created it. The world is pouring hundreds of billions into data centres, chips, memory, electricity and infrastructure while AI revenues still remain far behind the scale of investment, and this race is already capable of affecting the price of everyday technology.

History has seen this before: the internet changed the world, yet the dot-com bubble still destroyed companies and fortunes because money was invested faster than sustainable demand could arrive. AI could follow an even larger version of that pattern—the technology survives, but investors lose billions, workers lose jobs, weaker companies collapse, and the strongest survivors buy the infrastructure and competitors left behind.

Then comes the most disturbing possibility: after ordinary consumers pay higher prices, investors absorb losses and society carries the disruption, the world’s most powerful intelligence may end up controlled by only two or three corporations. We were warned for years that one day machines might control humanity. Perhaps we misunderstood the real danger.

The machines may never need to control us. It may be enough for a few companies to control the machines—and charge the rest of humanity for access to intelligence itself.

Comments

comments