Is manufacturing AI beginning to recede?

The manufacturing AI investment pie may really be about to "burn."
But don't rush to celebrate.
Because this time, what may get burned is not AI, but those projects that only tell AI stories.

Recently, Revalize, a supply chain and industrial software research firm, released a report with a very blunt title:
"The Manufacturing AI Boom Is Over."
What does this mean?
Does it mean manufacturing is no longer investing in AI?
Quite the opposite.
The report shows that in Q1 2026, only 36% of manufacturing enterprises had entered broad or advanced AI application; just half a year later, that number had reached 60%.
AI has not receded.
Rather, the first "give it a try" phase is over.

The question arises:
Since AI has already entered the factory, why is someone suddenly shouting "Boom is over"?
The answer may be hidden in one number:

Only 49% of manufacturing enterprises are strictly measuring AI ROI.The other half cannot even say clearly: how much money did this AI project actually make for me?

Revalize's survey also found that 50% of manufacturing enterprises still track AI results inconsistently and informally.

This gives rise to a very interesting phenomenon.
The boss says:
"We have fully embraced AI."
But when you ask on the shop floor:
"What problem did this AI actually solve?"
The answer may start to blur.
Ask further:
"How much efficiency improved?"
"How much cost dropped?"
"How much yield increased?"
"How much downtime was reduced?"
There is not even a financial measure.
This is the real danger signal for manufacturing AI.

01AI's Imaginative Space Is Too Large – and No One Can Calculate Its Value

Because in the past, digitalisation projects could often still talk about "infrastructure building."
But AI is different.
AI's imaginative space is too large.
AI can write code, read images, analyse data, make predictions, answer knowledge questions, and even make decisions for engineers.
So a situation easily arises:

The technical team thinks AI can do anything, management thinks AI is worth everything, but in the end no one has truly calculated how much it is worth.

What is more troublesome is that the deeper AI goes into the factory, the more obvious this problem becomes.

You think installing a large model is enough?
When you actually get to the site, you discover:
equipment data has no unified format;
MES, ERP, and PLM each have their own data;
equipment from decades ago is still in use;
master technicians' experience has not been digitalised;
many key processes are not even standardised.
So the biggest contradiction in manufacturing AI today may no longer be:

"Can AI do it?"But rather: "Does the factory have the conditions for AI to truly work?"

This is also why other manufacturing research this year has reached similar conclusions: AI has moved from pilots to scale, but data quality, system fragmentation, talent, and processes are increasingly becoming bottlenecks.

02The True Watershed for Manufacturing AI Has Only Just Begun

So the true watershed for manufacturing AI has only just begun.
In the past, the competition was:
who has more AI projects.
Next, the competition will be:
who can truly embed AI into production processes;
who can turn data into industrial knowledge that AI can invoke;
who can turn AI output directly into an action;
and more importantly –

who can ultimately convert that action into profit.

For example, an equipment diagnosis AI should not just tell you:
"This bearing may be abnormal."
It should continue further:
When should it be stopped?
What spare parts need to be replaced?
Who will perform the repair?
What is the estimated loss?
What consequences will occur if it is not handled?
How much money did it ultimately help the enterprise avoid losing?

Only at this stage
does AI transform from a "technical project" into an "operational capability."

03The Real AI Era May Have Only Just Begun

So, "Is the manufacturing AI pie about to burn?"
The answer is:
The first round of the AI pie may indeed be about to burn.
But this is not necessarily a bad thing.
Because after the foam is burned away,
what truly remains is no longer "how much AI we used,"
but a very simple question:

How much value has this AI actually created for the factory?

From now on,
manufacturing AI may no longer compete on whose story is bigger,
but on whose accounts can be clearly calculated.

After the AI frenzy ends, the real AI era may have only just begun.