A canary to detect the dangers in AI’s coal mine

By Alan Jacobson, AI Economics Strategist

A recent passage from The New York Times describes a structural failure in modern risk detection:

“Our models for detecting risk look at prices, volatility and correlations… They have no instruments for reading a grid failure, a drought or a severed supply chain. By the time warning signs show up in market data, the damage will already have been done.”

That is exactly where AI is today.

The industry is building at full speed—data centers, GPUs, models, agents—while relying on financial reporting frameworks that were never designed to see what’s coming.

We are measuring the wrong signals.

The blind spot

The current AI narrative is loud, but narrow:

• model benchmarks
• agentic workflows
• trillion-dollar valuations
• data center expansion
• environmental impact

All of it matters. None of it answers the only question that determines survival:

Does AI make money?

Right now, the people building AI—research scientists and product visionaries—are not operating inside a disciplined economic framework. They are optimizing for capability, not profitability.

That’s not sustainable.

And worse, the traditional tools used by analysts—revenue growth, EPS, forward guidance—are too slow and too blunt to detect what’s actually happening underneath.

By the time margin collapse shows up clearly in reported earnings, it’s already over.

The missing instrument

That’s the gap ACRD (AI Cost-Revenue Divergence) is designed to fill.

ACRD is not a forecasting model.
It’s not sentiment.
It’s not narrative.

It is a detection system.

Like the NYT describes, markets lack instruments to detect non-financial stress until it propagates into financial damage.

ACRD introduces that instrument for AI.

It measures, in real time:

• Revenue growth
• Cost of revenue growth
• The spread between them (in basis points)
• Gross margin compression

From these inputs, ACRD classifies companies into signals:

• Positive
• Neutral
• Emerging
• Active

This is not storytelling. It’s mechanical.

And what it reveals is uncomfortable:

In multiple AI-exposed companies, cost is rising faster than revenue.

That is the earliest detectable signal of structural margin pressure.

Why this matters now

Think of ACRD as the canary.

Not predicting collapse.
Not speculating.

Just detecting invisible stress before it becomes visible damage.

Because once it hits the income statement in a way everyone can see, three things happen fast:

1. Margins compress
2. Valuations reset
3. Narratives change overnight

The NYT warning applies directly:

By the time warning signs show up in market data, the damage will already have been done.

ACRD exists specifically to surface those warning signs earlier—inside the financials, before the market fully understands them.

The uncomfortable truth

There’s a deeper issue the industry keeps avoiding.

AI is being treated as inevitability.

It isn’t.

AI is not a scientific project.
It is not a philosophical milestone.
It is not a social movement.

It is a business.

And businesses live or die on one thing:

The relationship between cost and revenue.

Right now, that relationship is unstable.

• Inference costs are variable and unpredictable
• Pricing models (seat-based, token-based) are misaligned with compute reality
• Usage is increasing faster than monetization clarity

This creates a structural risk:

AI demand may scale faster than AI profitability.

If that happens, the entire industry hits a wall—regardless of how powerful the models become.

ACRD as early warning infrastructure

What the NYT describes for physical-world risk—grids, droughts, supply chains—ACRD applies to AI economics.

It is an early warning system for margin integrity.

It does not wait for:

• earnings misses
• guidance cuts
• layoffs

It detects the pressure while it is still forming.

That’s the difference between reacting to failure and seeing it coming.

The bottom line

The current AI conversation is focused on what AI can do.

ACRD forces a different question:

Can AI sustain itself economically?

Because if the answer is no, nothing else matters.

Not the models.
Not the benchmarks.
Not the hype.

The future of AI will not be decided by capability.

It will be decided by whether cost and revenue ever come into alignment.

ACRD is how you see that—before everyone else does.

– Published Tuesday, March 17, 2026

What’s the earliest signal you’re seeing that AI cost is becoming a problem?

Send signal: signal@revenuemodel.ai

I read every signal.