Is AI a Bubble… or the Accelerator of a System That Was Already Breaking?

A counter-analysis after watching Ed Zitron’s interview

I watched Ed Zitron talk about artificial intelligence as a gigantic machine built on promises, investment — and, above all, debt.

At first, his argument is difficult to dismiss.

Billions are being poured into AI models, GPUs, data centers, energy and infrastructure. Financing requirements keep growing. Companies are building infrastructure today for demand they expect to materialize tomorrow.

So yes:

There may absolutely be an AI bubble.

But something bothers me about framing the debate this way.

Because if we spend all our time looking at the debt, we may stop looking at what that debt is financing, who takes the risk, who captures the value — and, above all, what AI is actually accelerating.

---

Debt Is Not the Beginning of the Story

First, we need to clarify something about money.

When a commercial bank grants a loan, it does not necessarily take existing money from a vault and hand it to someone else.

Bank lending can create new deposits — effectively expanding the money supply.

But there is an essential distinction:

creating money is not the same thing as creating wealth.

A $10 billion debt can finance a $10 billion disaster.

Or it can finance infrastructure that generates economic value for the next thirty years.

That is why, to me, the existence of debt is less interesting than another question:

> What did we build with it?

Railways required enormous capital.

Oil required enormous capital.

Telecommunications required enormous capital.

The Internet absorbed enormous amounts of capital — and produced one of the most famous speculative bubbles in modern history.

The dot-com bubble burst.

The Internet didn't disappear.

That distinction matters.

“There is a bubble” and “this technology has no value” are two completely different statements.

---

AI Is Already Producing Something

This is probably where I differ most from the deeply pessimistic interpretation of AI.

Because I actually use this technology.

And what I observe is not theoretical.

One person can now write, code, translate, analyze documents, prototype an application, generate visual concepts, explore data, learn new skills and develop projects with capabilities that once required several different specialists.

AI does not automatically turn someone into a genius.

I see it more as a cognitive multiplier.

A good idea can move much faster.

A bad idea can move much faster too.

Someone capable of questioning, verifying, experimenting and correcting suddenly possesses an extraordinarily powerful lever.

And this is where even the expression “artificial intelligence” can distract us from the economic question.

I am less interested in asking whether the machine is truly “intelligent” than in asking:

> How much does it increase a human being’s capacity to act?

That is where much of its real economic value may lie.

---

AI May Not Be Creating a New System. It May Be Accelerating the Old One.

And this is where my thinking started moving away from Zitron’s argument.

Everyone talks about AI acceleration.

Fine.

But:

> Acceleration of what?

Productivity?

Capital?

Profits?

Automation?

Economic concentration?

Inequality?

Energy consumption?

Probably some combination of all of them.

And that is precisely the problem.

We are introducing an extraordinarily powerful acceleration technology into an economy that already had serious structural imbalances.

If a system distributes value efficiently, AI can accelerate wealth creation.

But if a system already concentrates capital heavily, AI can accelerate that concentration too.

The accelerator has no morality.

It simply accelerates the vehicle in which we install it.

---

Follow the Money

This is probably the question I found most absent from the discussion.

Let's assume Ed Zitron is right.

Let's assume a significant part of current AI investment is excessive.

Fine.

Who gets paid while the bubble is being built?

Chip manufacturers.

Energy companies.

Data-center operators.

Landowners.

Banks.

Bond markets.

Cloud providers.

Consultancies.

Developers.

Technology companies.

A bubble is not simply a mysterious black hole into which money disappears.

It is also an enormous mechanism for the allocation and transfer of capital.

And when the bubble bursts, another question appears:

> Who absorbs the losses?

That is where the debate becomes much more interesting.

Because if some of the infrastructure required by this new economy depends on electricity grids, land, tax incentives, public infrastructure or government guarantees, society has every right to ask what it receives in return.

Which leads to a brutally simple question:

> If part of the risk is socialized, why shouldn't part of the gains be socialized too?

---

Public vs. Private: Maybe This Is the Real Debate

A public hospital does not exist to maximize EBITDA.

A public school should not have to prove that teaching a child to read produces an acceptable quarterly return.

A library can operate at an accounting loss while generating enormous social value.

We already understand, therefore, that there are different forms of value.

Yet increasingly, we apply private-sector accounting logic to institutions whose primary purpose is collective.

At the same time, whenever an industry considered strategically important suddenly requires enormous infrastructure, public intervention becomes perfectly acceptable again.

That contradiction deserves examination.

I am not arguing that public debt is magically free.

Nor am I arguing that governments could simply erase every number from every balance sheet without consequences.

Debt represents claims, assets, future income, obligations and ultimately a distribution of losses.

Deleting the number does not delete the economic consequences.

But the opposite simplification is equally dangerous:

treating every public deficit as evidence of economic failure.

The better question is:

> What collective value did we create in exchange for that debt?

---

And What About the Knowledge AI Was Built On?

Here the question becomes even more uncomfortable.

AI did not emerge from a vacuum.

It rests upon decades — arguably centuries — of accumulated human knowledge.

Science.

Software.

Books.

Code.

Research.

Art.

Language.

Culture.

Publicly funded universities.

Open-source communities.

And countless contributions from ordinary people.

AI transforms this enormous information heritage into productive capacity.

So another question appears:

> Who owns the value generated from collective human knowledge?

Only the company operating the servers?

The investors who financed the GPUs?

The people who produced the knowledge?

The users?

Society?

Probably not just one of them.

And that is precisely why the question of distribution may ultimately matter more than the simple question of debt.

---

Does the Cake Even Exist?

I can already imagine Zitron's response:

> “You're asking for your share of the cake while I'm telling you the cake was bought on credit.”

Fair enough.

But even then, somebody built the oven.

Someone sold the electricity.

Someone supplied the ingredients.

Someone owns the bakery.

Someone financed it.

And millions of people are already buying slices.

So let's follow the entire chain.

Who invests?

Who borrows?

Who produces?

Who gets paid?

Who owns the infrastructure?

Who absorbs the losses?

And who captures the gains if the bet succeeds?

That is the accounting I am interested in.

---

The Universal Basic Income Question Will Eventually Arrive

And ultimately, this brings us to a debate that AI may simply be accelerating.

Imagine that a task that previously required ten people can eventually be performed by three people assisted by AI.

Productivity increases.

Excellent.

But what happens to the other seven?

More importantly:

> Where does the economic value of the labor that is no longer required go?

To the remaining workers?

To the company?

To the owner of the AI model?

To shareholders?

To consumers through lower prices?

To governments through taxation?

To society as a whole?

This is where Universal Basic Income becomes much more than a philosophical curiosity.

It becomes one possible answer to a structural problem:

How do we distribute purchasing power in an economy where production may increasingly require less human labor?

UBI may not ultimately be the right mechanism.

Perhaps the answer will involve public services, taxation, social dividends, shorter working hours, broader ownership of capital, sovereign investment funds — or something we have not invented yet.

But the underlying question cannot simply be avoided.

If automation dramatically increases production while reducing the amount of human labor required to produce it, the way income is distributed will eventually have to evolve.

---

Maybe AI Isn't the Real Thing We Should Fear

Ed Zitron asks an important question:

> What if we have massively overinvested in AI?

I would add another one.

> What if the greater danger isn't that AI fails — but that it succeeds inside an economic system incapable of distributing the gains of that success?

Because inefficient AI can burn money.

But extraordinarily efficient AI, controlled by a small number of actors and introduced into an already concentrated economy, creates a much bigger question.

It can accelerate productivity.

Accelerate capital.

Accelerate automation.

Accelerate profits.

Accelerate inequality.

And perhaps accelerate crises that existed long before AI arrived.

AI may be neither the monster nor the savior.

Maybe it is simply the accelerator.

And before we push the pedal any further, perhaps we should decide collectively:

> Where exactly are we going — and who gets to ride in the car?