Hey {{first_name | there}},

That question would have sounded ridiculous a year ago.

Today, financial analysts are openly asking whether OpenAI could run out of cash by mid-2027.

What happens when the most powerful AI company in the world cannot make the economics work?

And what that tells us about where AI is actually heading.

OpenAI is reportedly burning billions a year, with infrastructure costs accelerating faster than revenue.

Datacenters. Training runs. Inference at scale.

This stuff is not expensive. It is brutally expensive.

Sam Altman once called ads a “last resort.” But last week, OpenAI announced ads.

That alone tells you the pressure is real.

When the company defining the AI era starts experimenting with advertising, the signal is clear. The current model is not sustainable.

And that pressure is already reshaping how AI behaves, how it is deployed, and how much we trust it.

Here is where it gets strange.

Stanford recently found that AI is raising average wages by 21% while reducing wage inequality.

The reason is simplification.

AI takes tasks that once required deep expertise and makes them accessible to almost anyone. A junior developer can now do senior-level work. An entry-level analyst can produce outputs that once took years of experience.

On paper, this is incredible.

But it creates a new problem.

When everyone can perform expert-level tasks with AI, expertise itself starts to erode. We get outputs without understanding. Decisions without depth. Confidence without comprehension.

This matters because AI companies are betting that volume usage will save them financially.

But volume usage without trust is fragile.

On January 12, Anthropic released Claude Cowork.

Within days, SaaS stocks dropped sharply.

Why?

Because Cowork lets non-technical users automate workflows, manage documents, organize files, and build systems without subscribing to half a dozen tools.

Chegg is down 99% since ChatGPT launched.

Not because humans stopped learning.

Because the middleman software stopped being necessary.

This is the same pressure OpenAI is facing.

AI is eating the software layer that once generated predictable subscription revenue. And the companies selling the AI are discovering that replacing software does not automatically replace the business model.

You cannot burn billions while charging eight dollars a month.

Something has to give.

As AI becomes more capable, it is also becoming more dangerous in a quiet way.

Early AI failed loudly. Errors. Crashes. Obvious mistakes.

New models fail silently.

They produce outputs that look correct, pass surface checks, and feel confident, while being fundamentally wrong.

This is already happening in code, analysis, research, and decision-making.

The reason is subtle.

Models are increasingly trained on what humans approve of, not what is objectively correct. They optimize for acceptance, not truth.

This is disastrous for trust.

And trust is the one thing AI companies cannot afford to lose while already struggling financially.

Now connect the dots.

OpenAI needs revenue fast. Ads promise revenue.

But ads change incentives.

A tool that once optimized for helpfulness will slowly optimize for engagement. For retention. For advertiser safety and satisfaction.

Every platform that ever introduced ads said the same things first. Clearly labeled. No influence. User trust comes first.

History says otherwise.

When AI becomes ad-supported, neutrality becomes a liability.

And when trust erodes, users churn. When users churn, the economics get worse. Not better.

AI is doing all of the following at once:

The real risk is not AI replacing jobs.

It is AI being deployed at scale without sustainable incentives, without verification, and without business models that reward correctness.

The biggest danger is not superintelligence.

It is AI becoming just good enough that we stop questioning it, while the companies behind it scramble to stay solvent.

That is how real damage happens.

What are you seeing in your own work?

Have you noticed these silent failures yet?

Hit reply. I read every response.

– AP

P.S. If you are building something in AI right now, I would genuinely love to hear about it. What are you creating that AI cannot easily replicate? What is your moat?

Connect your email, and you’ll instantly get a CRM with enriched customer insights and a platform that grows with your business.

With AI at the core, Attio lets you:

Join industry leaders like Granola, Taskrabbit, Flatfile and more.

👉 Try Attio Pro for free