By the 5Tech engineering team · Published May 2021 · Written for operations and plant leaders and CTOs planning multi-year IIoT investment. Every market figure here is cited to its source or labelled illustrative; every forecast is stated with its horizon, its assumptions, and the reasons it could run slow. Examples are typed (industry pattern / illustrative). 5Tech has no published customer case studies, and none are implied.
The forecasts for the Industrial Internet of Things (IIoT) are directionally right and remarkably consistent: more connected devices, more value pooling in the factory, double-digit market growth for the rest of the decade. What they are not is a schedule. The single plant-wide adoption curve you could budget against does not exist — real adoption runs at four or five different speeds at once, gated by legacy equipment, interoperability, skills, security, and data quality. This piece is about planning against what the data actually supports, segment by segment, rather than against a hockey-stick slide.
The forecasts are directionally right — and consistently early
Start with the most useful fact in the whole dataset: the analysts who build these forecasts keep having to lower them. IoT Analytics trimmed its 2025 device outlook by roughly 300 million connections against its own prior forecast. McKinsey's estimate of value captured by 2020 and 2025 landed below what the same firm projected in 2015. That is not a knock on the forecasters — it is a signal about how to read them.
The operational reason for the lag is well documented. In McKinsey's manufacturing surveys, a large majority of companies have not scaled their Industry 4.0 use cases beyond pilots — the widely cited “pilot purgatory,” where roughly 70% of initiatives fail to move company-wide and most spend more than a year stuck there (McKinsey). The technology curve keeps climbing; the adoption curve is throttled by integration work, change management, and the last IT/OT mile.
So treat every IIoT forecast as a ceiling that holds only if a stack of assumptions holds — not as a delivery date. The rest of this article makes those assumptions explicit.
The device counts are real. The single, plant-wide timeline you could budget against is the fiction.
Read every forecast as four separate claims
A market slide blurs together four very different kinds of statement. Pulling them apart is the single most valuable discipline for anyone signing a multi-year cheque.
- Proven — measured, repeatable, cited. Example: connected-device growth and IIoT market size are real and independently tracked; predictive maintenance has delivered material gains across programs.
- Expected — a credible projection with a stated horizon and assumptions. Example: private cellular continuing to grow off a small base through the late 2020s.
- Possible — plausible but unproven at scale, or dependent on unfinished standards. Example: autonomous AI agents taking bounded actions on the line.
- Promotional — a vendor's roadmap presented as an industry timeline. Treat as marketing until a primary source or your own pilot says otherwise.
Most disappointment in IIoT programs comes from budgeting a possible or promotional claim as if it were proven. Below, each forecast is tagged with a horizon, the assumptions behind it, and the reasons it could run slow.
A connected discrete-manufacturing line — the segment where IIoT value pools first and fastest, because much of it is new-build and capital-rich. Most of the world's plants do not look like this yet, which is the whole point.
The real dividing line: greenfield versus brownfield
Before segmenting by industry, segment by plant age — because it drives the timeline more than any technology choice. A new plant designs connectivity, data models, and security in from day one. A plant running twenty- to forty-year-old assets has to retrofit around equipment that predates the internet, and every safety-related change has to be re-validated.
Greenfield (new build)
Connectivity, open data models, and IT/OT security specified up front. IIoT forecasts roughly apply here — this is where the fast CAGR is real. The constraint is capital and integration skill, not physics.
Brownfield (retrofit)
Long-lived assets, proprietary fieldbuses, no data model, air gaps being reluctantly opened. Adoption is staged and years-long. This is most of installed industry — and why aggregate curves overstate any single plant's pace.
The trap is applying a greenfield forecast to a brownfield reality. When a vendor curve says a technology is “going mainstream,” ask which of these two worlds it was measured in.
Segment by industry, not by one timeline
IIoT value is heavily concentrated in the factory (~26% of the total pool, per McKinsey), but “the factory” is not one buyer. Discrete manufacturers move at a different pace than process industries, which move differently again than logistics or utilities. The table below is an illustrative synthesis of the cited market research and public adoption data — a way to place your own operation, not a scoreboard.
| Segment | Dominant pattern | Realistic adoption pace | Primary brake |
|---|
| Automotive & electronics (discrete) | Greenfield-leaning, capital-rich | Fastest; leads IIoT spend | Integration & change management |
| Process (oil & gas, chemicals, metals) | Brownfield, 20–40 yr assets, safety-critical | Slow, staged retrofit | Legacy equipment; functional-safety re-validation |
| Logistics & warehousing | Mixed; mobile assets | Fastest-growing CAGR | Standards for mobile fleets; labour |
| Utilities & energy | Regulated, geographically dispersed | Steady, grid-modernisation-paced | Regulation; remote-site security |
| Food, pharma, regulated discrete | Validation-heavy | Moderate | Compliance & validation overhead |
Grand View attributes the fastest forward CAGR to logistics and transport, and the largest current share to manufacturing — consistent with the pattern above. The planning lesson: your segment, not the aggregate market, sets your realistic pace.
Name the delaying factors — every forecast has the same five
Whatever the technology, the same five brakes explain the gap between the curve and the calendar. A forecast that does not account for these is a wish.
- Legacy & brownfield — the installed base of un-networked, proprietary equipment is the dominant constraint, and it does not get replaced on a forecast's schedule.
- Interoperability — McKinsey estimates roughly 40% of potential IoT value depends on systems talking to each other; walled-garden protocols strand the rest (McKinsey).
- Skills — the scarce resource is people who understand both OT and IT; talent is a repeatedly cited barrier, and it caps how fast anything scales.
- Security & IT/OT convergence — connecting an asset enlarges its attack surface and erodes the old air gap; security done wrong can itself disrupt control timing.
- Data quality — every twin, model, and dashboard is only as good as the sensing and labelling beneath it; poor data quietly caps the value ceiling.
Specific forecasts, stated with discipline
Here is where the topic earns its keep. Each row below is a real forward-looking signal with a source, a horizon, the assumptions it rests on, and the concrete reasons it could arrive late. Note what is deliberately absent: no “standard within N months,” no single year when “IIoT arrives.”
| Signal (source) | Horizon & scope | What it assumes | Why it could run slow |
|---|
| Private cellular becomes a default option to evaluate (GSA: >2,000 organisations across 84 countries by end-2025; manufacturing leading) | Growing off a small base through the late 2020s; greenfield & hard-to-cable sites, high-income markets first | Spectrum access, available integrators, a clear ROI case, used to complement wired control | Uncertain economic value (a recognised adoption barrier), brownfield already cabled, LTE-vs-5G maturity, skills |
| Field-level, deterministic OPC UA over TSN (OPC Foundation UAFX / Field Level Communications) | Specs published now; broad multi-vendor field deployment over the second half of the decade | Vendors ship and interop-certify products; plants adopt TSN-capable infrastructure | Huge proprietary fieldbus installed base, certification cycles, vendor product roadmaps |
| Digital twins as a standard engineering tool (Grand View tracks a ~30%+ CAGR market) | Scaling through 2030+ where data is good; strongest in discrete & asset-intensive sectors | High-quality, well-modelled sensor data and an integration budget | Data quality, integration and model-maintenance cost, over-scoped “whole-plant” ambitions |
| AI agents that take bounded actions on the line (emerging; possible, not proven at scale) | Pilots now; supervised, guard-railed use over a multi-year horizon | OT-safe guardrails, human-on-the-loop, validated fallback, deterministic control kept authoritative | Safety, liability, operator trust, model drift, immature validation practice |
Two of these are safety-relevant, so the distinction matters: IIoT connectivity and analytics belong to monitoring and decision-support, and at most supervisory, roles. The most timing-critical, safety-related control loops stay on deterministic wired links (TSN, EtherCAT) and validated controllers. Private 5G is excellent for mobile and hard-to-cable assets — it is not a replacement for a safety interlock. An AI agent acts only within guardrails, behind a validated fallback, with the deterministic control layer remaining authoritative.
What the multi-year decision actually costs
Because forecasts describe a market and not your plant, the investment case has to be a number. The value of an IIoT program over its life is not the top-line CAGR — it is benefits net of a long tail of operational costs, discounted by the real probability that a pilot scales at all.
Program value = ( downtime avoided + scrap/quality gains + energy saved + labour redeployed + compliance/insurance benefit ) − ( sensors + connectivity + integration & interoperability + platform + OT security + data governance + skills/training + model & twin maintenance + hardware refresh/end-of-life ) × P(scales beyond pilot)
Two terms are routinely underestimated. The interoperability tax — the integration work to make proprietary systems talk — is where a large share of value is won or lost. And the lifecycle costs are real and recurring: sensors and gateways age out, twins and models drift and need re-validation, security patches never stop, standards evolve under you, and vendor lock-in raises the cost of every later change. A five-year IIoT plan that budgets only for installation is a one-year plan with a four-year surprise.
On business value, the upside is genuine but conditional. Across programs, predictive maintenance has been associated with meaningful reductions in unplanned downtime and maintenance cost — McKinsey reports downtime cuts commonly in the tens of percent, with maintenance-cost reductions in a comparable range (McKinsey). Read those as industry ranges to test against on your assets, not a guaranteed return — they depend on data quality, honestly labelled failures, and asset criticality.
The interoperability bet is the one safe long-horizon call
If any architectural decision deserves to be made against the ten-year view, it is this one. Proprietary, single-vendor stacks are exactly what stranded ~40% of IoT value in McKinsey's analysis, and they make every future addition — the next machine, the next analytics tool, the next agent — more expensive. Designing around open, vendor-neutral interfaces protects the investment as needs grow.
The established, deployable-today layer is OPC UA for secure, vendor-neutral information modelling, with MQTT and the Sparkplug specification for lightweight, well-structured plant messaging. The emerging layer is field-level determinism — OPC UA Field eXchange (UAFX) mapping onto Ethernet TSN, with SPE/APL physical layers — whose specifications are published but whose broad, multi-vendor field deployment is a second-half-of-the-decade proposition, gated by the proprietary fieldbus installed base and certification. Adopt the proven layer now; track the emerging one; assume neither is turnkey plant-wide.
Underneath all of it sits data quality. A twin, a predictive model, or an AI agent inherits the honesty of the sensing beneath it, which is why disciplined embedded sensing and IoT is the foundation every later layer is built on.
What to remember
- The IIoT forecasts are real and directionally right, but consistently early — read each as a ceiling under assumptions, not a schedule.
- Segment by plant age first (greenfield vs brownfield) and by industry second; there is no single timeline to plan against.
- Every forecast is throttled by the same five brakes: legacy equipment, interoperability, skills, security, and data quality — name them in your plan.
- Keep IIoT in its lane: monitoring and decision-support, not the safety-critical control loop; wireless complements deterministic wired control, it does not replace it.
- Bet long-horizon on open standards (OPC UA, MQTT/Sparkplug), and budget the interoperability tax and full lifecycle — not just installation.
Where to start
Skip the plant-wide platform decision and the five-year roadmap. Pick your segment's realistic pace from the table above, choose one high-consequence problem — a failure that keeps surprising you, scrap you can measure, an energy sink you can see — and instrument only that. Write down the acceptance criteria, the interoperability standard you will design to, the security review, and the fallback, then prove a measurable improvement on your own data before you scale. A program that survives contact with brownfield reality beats any forecast slide.
If you want help turning these trends into a staged, evidence-led plan for your operation — sequenced to your segment and your existing equipment — book a free engineering consultation with 5Tech.
References & further reading
- IoT Analytics — Number of connected IoT devices (State of IoT): 18.5 billion in 2024, ~21.1 billion in 2025, 39 billion by 2030 at 13.2% CAGR, and the reasons for the downward revision.
- McKinsey & Company — IoT value set to accelerate through 2030: US$5.5–12.6 trillion by 2030, factory settings ~26% of value, and the shortfall against 2015 projections.
- McKinsey & Company — The Internet of Things: mapping the value beyond the hype: interoperability required for ~40% of potential IoT value.
- McKinsey & Company — Avoiding Industry 4.0's pilot purgatory: the majority of manufacturers not yet scaled beyond pilots.
- McKinsey & Company — Prediction at scale: getting more value out of maintenance: predictive-maintenance downtime and cost ranges across programs.
- Grand View Research — Industrial IoT market: US$483 billion (2024) to US$1,693 billion (2030) at 23.3% CAGR; segment leadership.
- Grand View Research — Digital twin market: market size and ~30%+ CAGR outlook.
- GSA — Private Mobile Networks (December 2025): >2,000 organisations across 84 countries; LTE/5G split; manufacturing leading.
- 5G-ACIA: industrial 5G use cases, capabilities, and the recognised uncertainty about economic value that slows adoption.
- OPC Foundation — Field Level Communications (OPC UA Field eXchange / UAFX): OPC UA at the field level, TSN mapping, and SPE/APL physical layers.
Market figures are cited to their sources; the industry-segment and forecast-discipline tables are an illustrative synthesis of those cited sources, not a specific installation or a customer result.
Send 5Tech one workflow, inspection task, sensor problem, robotics challenge, or prototype idea. We will review it and suggest a practical next step. Start My Free First Phase — a free first review of one idea. If it is not practical, we will tell you.