Construction · Deep dive

OpenSpace

360° reality capture for construction — a camera on a hardhat, a walk someone was doing anyway, and Spatial AI that pins every frame to the floor plan; 275,000+ users and 43B+ square feet captured, now pushing from documentation into progress tracking (Disperse, acquired Nov 2025) and field task management (OpenSpace Field, Feb 2026).

emerging

The question that decides it: OpenSpace won reality capture by being the lightest layer to adopt — passive hardhat capture, unlimited users, pricing tied to construction volume rather than seats — and is now trying to convert that installed base (275,000+ users, 43B+ sq ft captured as of Nov 2025) into buyers of interpretation: Track's Disperse-powered progress tracking and Field's task management, all without new capital since the March 2022 Series D at $902M. Does the capture wedge actually convert — Track and Field attaching at rates that grow revenue per customer faster than Procore and Autodesk bundle free capture into the systems of record that already own the schedule — or does documentation commoditize into a feature, leaving a 2021-priced company whose intelligence layer still depends on Disperse's human analysts to verify what the AI sees?

Snapshot

OpenSpace is the volume leader in construction reality capture: strap a consumer 360° camera to a hardhat, walk the site as you already do, and its Spatial AI pins every frame to the floor plan — a Google Street View of the jobsite, rebuilt on every walk. Founded in San Francisco in 2017 by three MIT Media Lab alumni, it has raised roughly $199M — last priced at $902M in its March 2022 PSP Growth-led Series D — and by November 2025 counted 275,000+ users and 43B+ square feet captured across 94 countries. In the past year it has made its decisive strategic turn: from passively recording sites to interpreting them — acquiring London progress-tracking firm Disperse (November 2025), launching the task-management product OpenSpace Field (February 2026), and signing group-wide enterprise deals with the likes of BAM and Comfort Systems USA. It matters because it is the company that proved documentation could be nearly free to adopt — and it is now testing whether the wedge converts into intelligence revenue before the platforms above it give capture away.

Founding story

The three founders were Media Lab colleagues a decade before they were co-founders. Jeevan Kalanithi co-created Siftables — sensor-laden game cubes — with David Merrill at MIT, spun them into Sifteo, and sold the company to 3D Robotics in 2014, where he became president and pushed the drone maker toward construction software as consumer drones collapsed into DJI’s price umbrella. That detour is the origin: at 3DR he watched builders pay for aerial site data while the inside of buildings — where most of the money and most of the disputes live — went undocumented or was documented by a field engineer with a phone camera and a Friday afternoon.

Kalanithi left 3DR in mid-2016 planning to decompress, and instead reconnected with two MIT friends: Philip DeCamp, a computer-vision research scientist at the Media Lab, and Michael Fleischman, who had turned his MIT research into Bluefin Labs and sold it to Twitter in 2013. The two were exploring what modern machine vision could do with cheap 360° cameras; Kalanithi supplied the industry. Builders everywhere voiced the same complaint — documentation was manual, sampled, and stale — and the trio’s prototype was almost naive in its simplicity: put the camera on a hardhat and let software do everything else. They founded OpenSpace in 2017 with seed money from Lux Capital. The founder-market fit is the inverse of Buildots’ (three Israeli officers who chose construction cold): these are imaging and vision people who had already sold two companies, and whose bet was that zero-effort capture would beat better analysis, because superintendents do not adopt tools that add work.

How it works

The physical loop: a superintendent or engineer clips a 360° camera (Insta360 and Ricoh models are supported) to a hardhat, taps record in the OpenSpace mobile app, and walks the site — a walk they were doing anyway. The camera takes roughly two 360° frames per second. Nobody stops, aims, or tags anything; that passivity is the entire adoption thesis.

The hard part is the Vision Engine. Indoors there is no GPS, so OpenSpace localizes the footage against the uploaded floor plan using visual-inertial odometry — in effect, SLAM tuned for construction sites, which are adversarial environments for vision (repetitive corridors, surfaces that change week to week, dust, poor light). After upload, the engine stitches the walk into a navigable, plan-pinned visual record, typically processed within minutes to hours, with each new walk layered onto the same map so anyone can stand in a spot and scrub backward through time. On top of that record sit comparison tools: BIM Compare puts the model and the as-built side by side in the browser; Split View compares two dates.

What OpenSpace historically did not do is interpret — say what percent of drywall is hung, or whether the schedule is slipping. That is what Disperse adds. OpenSpace Track (launched June 2025, powered by Disperse, whose team OpenSpace acquired outright in November 2025) combines the imagery with computer vision and expert human verification — Disperse ran an analysis team heavy on trained architects and engineers — to report milestone-level progress against plan, validate work-in-place for billing, and flag schedule risk. OpenSpace Field (GA February 2026) pushes the other direction, into workflows: smartphone-based task and punch-list management with AI Autolocation, which infers where on the site an issue was logged from imagery alone — real-time indoor positioning with no beacons or hardware — plus AI voice notes, with bi-directional Procore and Autodesk Construction Cloud sync.

Product and business overview

The platform, rebranded in 2025 as “Visual Intelligence,” has five named components. Capture is the core 360° documentation product and the wedge. BIM+ (BIM Compare and related tools) handles model-versus-reality comparison and measurement. Air (May 2025) folds drone imagery — orthomosaics, point clouds, meshes — into the same record, a direct response to DroneDeploy’s air-plus-ground bundle. Track (June 2025) is the Disperse-powered progress-tracking layer. Field (February 2026) is task management. Around the products: integrations with Procore, Autodesk, Egnyte and others, and a FedRAMP Moderate authorization (July 2025) — the only 360° capture platform with one, opening federal and defense-adjacent projects. Buyers span general contractors (Suffolk, Turner-scale GCs), specialty trades (Comfort Systems USA signed an enterprise agreement covering its subsidiaries), owners, and European majors — BAM signed a three-year group-wide contract across the Netherlands, UK and Ireland in February 2026.

Business model and pricing

Revenue is subscription SaaS, and the pricing model is the strategy. OpenSpace prices on annual construction volume (ACV) — a percentage of what the customer builds, scaled by product tier (Core or Enterprise) — with a published floor of $10,000 a year, unlimited users, and unlimited captures (company pricing page, 2026). That inversion of per-seat and per-project models removes every marginal reason not to capture, maximizing the data flywheel — and it means revenue grows with customer construction volume, not usage, which is precisely what the Track and Field launches give the ACV percentage to expand into.

No revenue figure is disclosed. Latka pegged ARR at $56.7M for 2023, a number to hold loosely — the same page mislabels the company bootstrapped. What is verifiable: no priced equity round since the $9M August 2022 extension, no announced down round, and layoffs in November 2023 (TrueUp/Sunset trackers; echoed in Glassdoor reviews). The generous read is that the company cut to near-breakeven on Series D cash and can fund the intelligence pivot internally — it had the cash to buy Disperse in late 2025. The ungenerous read is that a $902M, ZIRP-era mark has been unclearable for four years, and the acquisition was partly a way to buy growth into a story the next round needs.

Traction over time

Two things stand out. First, the capture curve is genuinely steep — square footage roughly tripled from 2021 to 2024 and kept compounding without new capital. Second, the company’s own disclosures wobble: the November 2025 Disperse press release cites 275,000 users and 43B square feet while the year-end 2025 review claims ~350,000 users and 50B+ — inconsistencies of the kind that suggest marketing numbers rather than audited metrics. Every figure above is company-published; none is independently verified, and no revenue, NRR, or logo-count series exists publicly.

Market analysis

Construction is a ~$13T global output industry with chronically flat productivity; the directly relevant slice — reality capture — was estimated at ~$5.7B in 2024, projected to ~$13.2B by 2030 at a ~14% CAGR (Growth Market Reports and comparable analyses, 2025). The structural forces are favorable: data centers and advanced manufacturing are the fastest-growing, most documentation-obsessed construction categories (Kalanithi explicitly pitched data-center owners in 2026 press); OpenSpace claims customers report 41% fewer insurance claims; disputes make an objective visual record self-justifying; and labor scarcity raises the value of remote oversight. The counterforce is structural too: documentation is the most commoditizable layer of contech. The capability — stitch 360° video to a floor plan — has migrated from breakthrough (2018) to feature (2026), available in some form from Procore, Autodesk, DroneDeploy, Cupix, and cheaper regional clones. TAM is expanding while the defensible layer inside it moves up-stack, which is exactly the migration OpenSpace is attempting.

Competitive intel

The category sorts by depth of interpretation versus breadth of adoption, and OpenSpace anchors the breadth end. Buildots ($166M raised; $300M estimated valuation, Calcalist, May 2025) is the depth extreme: element-level tracking aligned to BIM and schedule, delay forecasting, deployed on BIM-mature megaprojects for Intel, Turner, and VINCI. Buildots frames OpenSpace as the good-enough down-market rival; the counter is that OpenSpace has perhaps an order of magnitude more jobsites, needs no BIM model to land, and — with Disperse — now sells a progress layer into a base Buildots would have to win site by site. Buildots still wins on forecast-grade accuracy and owner-side credibility on complex projects. DroneDeploy ($142M+ raised) bought StructionSite in November 2022 and bundles aerial with interior capture; OpenSpace Air was the direct answer. Matterport, now inside CoStar (~$1.6B acquisition closed February 2025), owns photorealistic digital twins but not the weekly jobsite cadence. Reconstruct and Cupix compete on measurement depth and price respectively, both far smaller. The existential adjacency is Procore and Autodesk: both host the schedule and budget that make imagery meaningful, both ship free or bundled capture, and both are layering AI across their platforms. OpenSpace’s Field product syncing bi-directionally with Procore is simultaneously an integration and an admission of where power sits.

History and evolution

• 2014-2016 — Kalanithi sells Sifteo to 3D Robotics, becomes president, steers 3DR toward construction; leaves mid-2016.
• 2017 — OpenSpace founded in San Francisco; $3.5M seed (Lux Capital).
• 2018 — Product launches; early GC deployments.
• 2019 — $14M Series A; Suffolk Construction invests as customer.
• 2020 — $15.9M Series B (Menlo); COVID travel restrictions make remote site visibility mandatory; 3B sq ft captured by November.
• Feb 2021 — $55M Series C (Alkeon); ClearSight analysis tools (object search, progress views) begin the first interpretation push.
• Mar 2, 2022 — $102M Series D led by PSP Growth at $902M (Bloomberg); total $190M.
• Aug 2022 — ~$9M extension for Asia expansion. The last disclosed equity.
• Nov 2023 — Layoffs (TrueUp/Sunset); the post-ZIRP contraction hits contech broadly.
• Apr 2024 — 30B sq ft, ~50,000 projects, 93 countries.
• May-Jul 2025 — OpenSpace Air (drones), OpenSpace Track with Disperse (progress tracking), FedRAMP Moderate authorization — three strategic launches in ninety days.
• Oct 28-Nov 2025 — Acquires Disperse outright (terms undisclosed), citing strong Track adoption; Construction Dive frames it within a contech M&A wave.
• Feb 2026 — OpenSpace Field GA (Feb 3); BAM signs a three-year, group-wide contract across Netherlands, UK and Ireland.
• Jul 2026 — Kalanithi publicly targets data-center construction as the growth vector (Commercial Observer).

The visible stumbles: the 2023 layoffs, four years without a mark, and a half-decade in which the company’s interpretation ambitions (ClearSight, 2021) had to be fulfilled by buying someone else’s team in 2025.

What people say

The case for. Customer reviews on G2 and Capterra (2024-2026) are consistently strong and consistently about the same three things: it takes near-zero effort to capture, the plan-pinned time machine settles disputes and saves site visits, and coordination between trades improves when everyone sees the same current imagery. Reviewers specifically praise seeing every phase from earthwork to closeout and pulling up the site during owner meetings without traveling. The enterprise signals agree: Suffolk was a customer-investor by 2019, Comfort Systems USA and BAM signed group-wide agreements (2025-2026), and FedRAMP authorization implies security review depth most contech startups never attempt. Glassdoor runs 4.2/5 across 64 reviews with 94% recommending — high for a company that has done layoffs — with praise for the product team’s shipping pace.

The complaints. Customer gripes cluster on the interpretation edges, not the core: BIM Compare rendering is slow on heavy models, desktop sync lags, batteries die on long walks (reviewers ask for powered mounts), and new users need real training despite the “just walk” pitch. The deeper operational complaint — visible in construction forums and implicit in the product’s own design — is capture fatigue: the walk is passive but still someone’s job, cadence decays when a project gets busy, and stale captures quietly gut the value. Employee-side, Glassdoor’s negative minority recurs on layoffs and transparency (“lots of layoffs,” “lack of true transparency,” 2023-2025 reviews). And the structural critiques: every traction number is company-published and internally inconsistent between releases; the pricing floor of $10K keeps genuinely small builders out while ACV-based pricing gives large customers an incentive to audit what they pay as volume grows; and Track’s accuracy still leans on Disperse’s human analysis team — trained architects and engineers reviewing imagery — which is a margin question the company has never addressed publicly.

Outlook: the open question

OpenSpace works if the capture wedge converts: if Track and Field attach across the existing base fast enough that revenue per customer compounds — the ACV percentage expanding as customers add interpretation products — proving the installed base is a distribution moat and not just a feature audience. The bull case is genuinely strong: the largest jobsite footprint in the category, adoption physics (passive capture, unlimited users) that rivals cannot match without torching their own pricing, the Disperse team giving it a credible answer to Buildots at a fraction of Buildots’ deployment burden, FedRAMP and enterprise deals opening markets, and — if the Latka-scale revenue is even directionally right — a business that survived four years without raising, which no ZIRP tourist did. It stalls if documentation finishes commoditizing before the intelligence layer earns its premium: Procore and Autodesk bundling good-enough capture into the systems of record, Buildots holding the high-value forecasting tier, and OpenSpace squeezed into the middle — too shallow for megaproject owners, too paid for what platforms give away — with a $902M 2022 mark it cannot clear and human-verified margins it cannot disclose. Watch three tells: whether OpenSpace ever publishes Track/Field attach rates or any revenue figure (confidence signal); whether the next capital event is a priced up-round, a flat insider extension, or a sale to Procore/Autodesk/CoStar — for whom it is an obvious target (verdict signal); and whether Disperse’s human-verification layer gets automated away or quietly grows headcount (margin signal). The company built the habit 275,000 people have on site every week. The open question is whether a habit is a moat.

Sources and further reading

• Bringing construction projects to the digital world (MIT News, October 2020 — founding story, Sifteo/3DR/Bluefin background)
• Mike Fleischman, Philip DeCamp and I are starting OpenSpace.ai (Jeevan Kalanithi, Medium, 2017)
• Proptech Startup OpenSpace Reaches Valuation of $902 Million (Bloomberg, March 2, 2022)
• OpenSpace Follows Recent $102M Raise With $9M More in Funding (Commercial Observer, August 2022)
• OpenSpace Layoffs: What Happened & Why? (Sunset layoff tracker, November 2023)
• OpenSpace Launches Comprehensive Progress Tracking Solution (PR Newswire, June 4, 2025 — Track powered by Disperse)
• OpenSpace Becomes First and Only FedRAMP Authorized 360° Reality Capture Platform (PR Newswire, July 8, 2025)
• Amid M&A uptick, OpenSpace acquires progress tracking firm (Construction Dive, October-November 2025)
• OpenSpace Acquires Construction Progress Tracking Leader Disperse (PR Newswire, November 2025 — 275,000 users / 43B sq ft figures)
• OpenSpace Announces General Availability of OpenSpace Field (PR Newswire, February 3, 2026)
• BAM Expands Collaboration with OpenSpace Across the Group (OpenSpace, February 2026)
• OpenSpace CEO Jeevan Kalanithi: 5 Questions (Commercial Observer, July 2026)
• OpenSpace Reviews (G2, accessed July 2026) and OpenSpace Reviews (Glassdoor, accessed July 2026)