AR in Construction: What It Costs, Which Hardware Survived, and What to Build vs. Buy

Augmented reality in construction puts the Building Information Model (BIM) onto the physical site, viewed through a tablet, a phone or a headset, so a crew sees the ductwork exactly where the ductwork is going. The abbreviation is overloaded on a jobsite, so to be clear: this article means augmented reality, not an Action Request or an Architect’s Response.

One fact should frame the rest. Microsoft stopped producing HoloLens 2 in October 2024, exited AR headset hardware altogether in February 2025, and set support to end on 31 December 2027. That device anchored the category for years, and plenty of published guidance still recommends it for new deployments.

The technology itself works. What has changed is which hardware you can safely specify, what a deployment costs once model preparation is counted, and whether you should license a platform instead of building one.

What Does AR Mean in Construction?

Augmented reality places computer-generated content into your view of the real world. On a construction site that content is almost always the BIM model — the 3D design plus the data attached to it, anchored to the physical location where the work is happening.

You look through a tablet, a phone or a head-mounted display, and the ductwork appears where the ductwork is meant to go. The model moves with you as you walk.

Three related terms get used interchangeably and mean different things. Virtual reality replaces your view entirely. Mixed reality lets digital objects interact with the physical space around you. Extended reality (XR) is the umbrella covering all three.

The taxonomy of AR itself, marker-based, markerless, location-based, projection, matters more when you are specifying a build than when you are choosing a use case. I have covered that separately in our guide to how AR applications are built.

AR vs VR vs Mixed Reality on a Construction Site

The practical difference for a contractor is where each one is used. VR lives in the office and the training room. AR and MR go to the site, which is why they carry the hardware and connectivity constraints covered further down.

How AR Is Used in Construction: Seven Applications That Are Actually Deployed

Design visualization and constructability review

Teams walk a model at full scale on the site itself and catch problems that read fine in 2D. This is the most common use by a wide margin, 77% of surveyed AEC professionals named design visualization and evaluation as their primary AR/VR application.

Clash detection in the field

Office-based clash detection finds model-to-model conflicts. On-site AR finds model-to-reality conflicts, which is a different and often more expensive category of problem, the duct that clashes with a beam that was poured 40 mm out.

Progress capture

Some platforms use the device’s AR positioning to recognise where you are in the floorplan and photograph from the same point every visit. Consistent capture points make progress genuinely comparable week to week.

Remote assistance

A specialist in the office sees the crew’s camera feed and annotates directly onto it. The annotation stays anchored to the object even when the camera moves.

Setting out and layout

Engineering-grade systems project the model onto the slab so crews can position elements without a surveyor on standby for every step. This tier is covered in more detail below.

Safety training and hazard recognition

Research on VR-based safety training for construction trades reported hazard recognition improving by 39% and hazard management performance by 44%. Training mostly runs in VR rather than AR, and it is one of the strongest evidenced applications in the field.

The more interesting safety work is moving in a different direction. Dr. Omidreza Shoghli and colleagues at the William States Lee College of Engineering, University of North Carolina at Charlotte, have been building AR systems that predict vehicle intrusions into highway work zones and warn the worker through the headset in real time. That points at where AR may earn its keep first — not showing people a model, but telling them something they cannot see coming.

Client presentations and bid support

Showing an owner the finished building standing on the site changes the conversation. Architecture firms and owners are consistently ranked as the parties benefiting most from these technologies, ahead of contractors.

How AR Connects to BIM, Revit and Autodesk Forma

This is where most of the work actually sits, and it is the part that surprises people.

A Revit model carries far more geometry and metadata than a handheld device can render at a usable frame rate. Loading it raw onto a tablet produces something slow enough that crews stop using it by the second week.

The normal path is an export to IFC — the open file format the industry uses to move models between tools, or to a platform-specific format, followed by an optimization pass that strips detail nobody needs on site. Fastener geometry and manufacturer metadata can go. Structural grid, MEP runs and setting-out points stay.

Budget for it. A pilot that allocates nothing to model preparation is a pilot that stalls before the first site walk.

If your models sit behind a common data environment or an ERP that also holds cost and schedule data, the pipeline needs to be built once and maintained. That work is closer to connecting model data to existing systems than to app development.

AR Hardware for Construction in 2026: What Survived

The hardware market reset and a lot of guidance has not been updated to reflect it.

Microsoft ended production of HoloLens 2 in October 2024, confirmed in February 2025 that it was exiting HoloLens hardware development entirely, and set support to end on 31 December 2027. No successor has been announced. Devices already deployed keep working and keep receiving security updates until that date.

That matters because HoloLens anchored the enterprise AR category for years and sits underneath other products. If a vendor recommends it for a fresh rollout in 2026, they are working from an old script.

Two things to take from that table. First, the accuracy column separates the market more meaningfully than price does. Second, the bottom row is where the majority of firms should begin, because it costs nothing in hardware and tests whether the workflow holds before anyone buys a headset.

One more constraint that rarely appears in vendor material: site PPE. A practitioner interviewed for the Cal State Long Beach study described headsets running hot enough to be unpleasant over a shift, and noted that some devices simply cannot be worn with a hard hat on. Confirm PPE compatibility before you buy anything, not after.

AR in Construction Cost Breakdown by Build Tier

Hardware is the visible cost and usually the smaller one. The spend that decides whether a deployment works sits in model preparation, integration and the pipeline that keeps models current.

The costs buyers routinely miss are the recurring ones. Model preparation repeats every time the design changes. Someone has to own device provisioning and app updates. Platform licenses are usually per-seat per-month and scale with crew size rather than project value.

One trend is worth weighing against that spend. Between 2018 and 2023, change orders and cost management climbed from near the bottom of the list of emerging AR/VR applications to third place among surveyed professionals. Firms are starting to aim this technology at commercial disputes rather than design presentation, which is a more durable place for it to sit and a much easier one to build a business case around.

For a rough sense of what a comparable custom build involves before you scope AR specifically, our indicative cost calculator covers the standard variables, and typical build timelines sets expectations on duration.

Buy a Platform or Build Custom? The Four-Question AR Decision Test

I use four questions with clients scoping this. Three or more “standard” answers means licence. Two or more “specific” means build.

Licensing is the right answer more often than people expect. A mature platform has already solved model optimisation, device management and offline behaviour, and those are the parts that take longest to get right.

The same Cal State Long Beach research picked up a related shift: firms are increasingly choosing to outsource AR/VR work rather than build the capability in-house, and leaning on lower-cost virtual design and construction tooling to reduce how much internal AR expertise they need at all. That matches the staffing numbers, the share of firms with no AR/VR specialists doubled over the same period. Very few contractors want a permanent AR team.

Building earns its cost when the workflow is the differentiator. If AR is feeding an ordering pipeline, a proprietary approval chain or a data model nobody sells off the shelf, a platform will fight you. That is the point at which commissioning a custom build becomes the cheaper option over a three-year horizon.

What a Real AR Build Looks Like: Spatial Scanning and Render Accuracy

We built CPTNS, an AR and AI-driven design platform in the construction space, and the engineering problem in it is the same one a site-based AR tool faces.

Users scan their surroundings on a phone, the app stitches multiple images into a continuous view of the space, and an AI coping plotter positions material selections onto that geometry using geometric algorithms. The output has to be accurate enough that someone will place an order against it.

Three decisions carried that build. The scan-and-stitch step had to tolerate real-world lighting rather than studio conditions. Rendering ran natively on iOS and Android, because cross-platform rendering could not hold the frame rate the interaction needed. The platform maintained 99.95% uptime through peak load.

The transferable lesson is that the AR layer is the visible part and the smaller engineering problem. Capture quality, geometry handling and render performance are what determine whether the tool survives contact with real users.

Why AR Pilots Fail on Real Sites

The honest position on this technology is that adoption has gone backwards, and that is worth understanding before you commit.

Balali’s surveys, run through the Construction Management Association of America, reached more than 200 AEC professionals across 2018, 2020 and 2023. The trend line is not the one the category’s marketing implies.

AR/VR use among respondents rose from 56% to 65%, then fell to 44% in 2023. The share of firms with no AR/VR specialists on staff went from 14% to 29%. The proportion of respondents with under a year of experience with the technology reached 42%, which points to churn rather than growth.

His team concluded that early optimism gave way to tempered expectations as costs, limitations and implementation difficulty became clear. That is the context a demo will not give you.

Here is what actually breaks, and what to do about each.

1. Model preparation is under-budgeted. Named by 61% of surveyed practitioners as the top obstacle. Mitigation: scope the optimisation pipeline as a deliverable with an owner, not as a setup task.
2. Site conditions defeat the device. Jobsite usability problems including poor lighting were the second most-cited limitation at 43%. Dust, glare and glove operation all degrade touch interfaces. Mitigation: run the pilot in the worst conditions on the project, not the best.
3. Connectivity is assumed. Many platforms expect a live connection for collaboration and sync. Mitigation: require genuine offline working and a reliable sync-on-reconnect path in the evaluation, and test it with the radio off.
4. The accuracy tier is wrong for the job. A tablet gives a visual impression; it will not hold position well enough for setting out. Engineering-grade systems reach 3–5 mm and cost accordingly. Mitigation: decide the tier from the task, then let it set the budget.
5. PPE compatibility is checked last. A headset that cannot be worn with a hard hat is not a site device. Mitigation: put PPE compatibility in the procurement criteria.
6. Nobody defined the problem first. Dr. Steven K. Ayer, Associate Professor of Construction Engineering at Arizona State University, came to AR by hunting for problems the technology could solve, and he puts the awkward question directly: what happens when a technology gets implemented without a problem to solve? Pilots launched because the technology looked impressive end when the champion moves on. Mitigation: one metric, agreed before the pilot starts, RFIs avoided, rework caught, site visits eliminated.

How to Run a First AR Pilot Without Wasting the Budget

Keep it small enough that failure is cheap and specific enough that success is legible.

Pick one use case and one crew. Choose the accuracy tier from the task rather than the budget. Use devices you already own unless the task genuinely requires survey accuracy.

Scope model preparation as its own deliverable with a named owner and a real allocation. Run the pilot on the hardest site conditions you have, not the easiest. Agree one success metric before anyone starts.

Then give it a full project phase. Two site walks will tell you whether the technology works; only a phase will tell you whether the crew keeps using it.

Conclusion: Match the Tier, Budget the Pipeline, Pilot Small

AR in construction has moved past the demo stage, and it delivers inside a narrower band than the category’s marketing implies. The technology is capable. The failures are almost entirely about specification.

Choose the accuracy tier from the work rather than the brochure. Budget for getting models onto devices, because that is what practitioners report as their biggest obstacle. Check the hardware still has a support runway, since the best-known device in this category does not.

Then start with one crew and one measurable outcome. A contractor who runs a disciplined pilot on a tablet will learn more in one project phase than a firm that buys twenty headsets and hopes.

Not sure which tier your use case needs? Run through it with a solutions architect. Five questions, one call.