Industry 4.0 in 2026: What Actually Got Built, and What Didn’t

When this article was first published in 2021, Industry 4.0 was mostly a forecast. Connected factories were something you read about in vendor case studies, the market projections were enormous, and the question every manufacturer was asking was whether any of it would actually arrive.

Five years later, the question has changed. It arrived. Just not evenly, and not in the shape most people expected.

This is a rewrite of the original piece, using what the data now shows rather than what the projections once promised.

Industry 4.0 is the point at which physical production becomes instrumented, connected, and partly self-directing. Sensors on machines generate continuous data, that data flows to systems that can analyse it in real time, and some decisions get made without waiting for a human to notice a problem.

One clarification worth making, because the terms get used interchangeably and shouldn’t be. Industry 4.0 is not a synonym for IoT. The Internet of Things is one technology inside it — the sensing layer. Industry 4.0 is the operating model that emerges when sensing, connectivity, analytics, and automation work together. You can buy IoT. You cannot buy Industry 4.0.

The third revolution gave machines instructions. The fourth gives them information, and increasingly, the authority to act on it.

Deloitte surveyed 600 manufacturing executives for its 2025 Smart Manufacturing and Operations research. The headline finding is about intent, and it is strong.

A second figure makes the same point from a different angle. Roughly 46% of manufacturers use industrial IoT at facility level — meaning most have it running somewhere, and the majority have not taken it plant-wide. The Manufacturing Leadership Council’s 2026 survey finds 72% of respondents place themselves at mid-level maturity, with nearly half now scaling beyond single-site pilots.

Pilots are no longer the problem. Scaling past them is.

The original version of this article quoted a market forecast. Most articles on this topic still do. It is worth explaining why we have dropped it.

Here is what four research firms currently say the IoT-in-manufacturing market will be worth in 2026:

Adoption data tells you something. Market forecasts, in this category, mostly tell you which report you read.

Three things arrived that were not part of the original conversation.

Two findings from the same Deloitte research sit uncomfortably together, and both are worth knowing.

More than a third of the 600 executives surveyed named their single biggest concern as equipping workers with the skills to make smart manufacturing work. Not funding it. Not selecting vendors. Staffing it.

This is the same pattern that shows up in every large digital transformation programme: the technical work is tractable, the organizational work is where it stalls. Manufacturing is not an exception. It may be the clearest example.

Manufacturers in the Deloitte survey self-report gains of up to 20% in production output and employee productivity, and up to 15% in unlocked capacity. These are executive self-assessments rather than independently audited results, so treat them as directional.

Predictive maintenance remains the single largest application, holding roughly 30% of deployments and growing faster than any other use case — largely because the value is easy to measure. Avoided downtime has a price everyone already agrees on.

The original version of this article listed six organizational profiles. They have held up well, so they are kept here with the reasoning updated.

The 57%-to-29% gap is a sequencing lesson. Manufacturers who bought AI before fixing their data are the ones sitting in the gap.

Step 1 — Instrument and connect. Sensors, connectivity, and a place for the data to land. Most manufacturers have done this. It is table stakes rather than advantage.

Step 2 — Fix the data foundation. Consistent formats, resolved identities across systems, pipelines that keep up with production, and governance over who sees what. This is unglamorous and it is where the 28-point gap lives. Our data engineering practice exists largely for this stage.

Step 3 — Deploy analytics and prediction. Predictive maintenance first, because the value is measurable. Then quality inspection, then throughput optimization. Analytics and BI is what turns sensor output into a decision.

Step 4 — Add autonomy selectively. Agentic and physical AI belong on top of a working data layer, not instead of one. ML engineering is the discipline that keeps these systems reliable once they are making decisions unsupervised.

The 2021 version of this article asked where Industry 4.0 was leading us. Five years of data gives a clearer answer than a forecast could.

It is leading toward plants where the sensing is done, the data foundation is half-built, and the intelligence layer is arriving unevenly — faster at manufacturers who fixed their data first, slower at those who bought models before pipelines. The technology stopped being the hard part somewhere around 2023.

What remains hard is the same thing that has always been hard: connecting systems that were never designed to talk to each other, and equipping people to work differently once they do.