Researched by BEAST Library Curator | Verified by Harper | Quality: 8.3/10
BLUF: In 2026 the "horizontal building operating system" stopped being a bet and became consensus — Johnson Controls, Siemens, Honeywell, and Schneider are all shipping named GenAI agents on top of it, and the M&A machine is absorbing the independent vendor pool that fed it (JCI bought Nantum AI in April 2026; Trane bought BrainBox AI in January 2025). If you are specifying a building OS this year, the platform layer you pick will outlive your current facilities director. The one clause that determines whether you keep your leverage is data portability — and most FMs are still not writing it into the RFP.
The horizontal-platform thesis just went from bet to consensus
For a decade the smart-building pitch was a debate: does value accrue to the box (a better chiller, a smarter meter) or to the layer that aggregates every box into one pane of glass? In 2026 the four largest incumbents answered by putting their capital on the layer. Johnson Controls' OpenBlue, Siemens' Building X, Honeywell's Forge, and Schneider's EcoStruxure are four competing versions of the same thesis — a horizontal platform that ingests HVAC, fire, security, energy, and third-party data, then fronts it with generative AI agents.
The agents are now named products, not slideware. Per Memoori's 2026 incumbent-strategy analysis, Siemens shipped four named GenAI agents in March 2025 with a five-stage autonomy roadmap; JCI has shipped named OpenBlue agents since November 2024; Honeywell Forge runs Google Cloud Gemini agents; and Schneider announced its Foresight Operation layer in November 2025. This matters to a facility GM for one blunt reason: the marketing has caught up to a real capability shift, and the platform — not the equipment — is now where the decisions get made.
The market is scaling to match. The global smart-buildings market was valued at roughly USD 99.81 billion in 2025, reached USD 109.33 billion in 2026, and is projected to hit USD 248.26 billion by 2035 — a 9.54% CAGR (Global Growth Insights). That is the demand curve every one of these platforms is racing to sit on top of.
The consolidation squeeze: the independent vendor pool is being bought
Here is the structural story most facility teams are missing. Memoori tracks 360+ companies in the AI-in-buildings market across 69 use cases and 12 domains. That fragmented pool is exactly what the incumbents are consolidating — and the pace accelerated in 2026.
| Acquirer / Platform | Named AI agent(s) | Key 2025–26 acquisition | Proof-point scale |
|---|---|---|---|
| Trane Technologies | BrainBox AI Control | BrainBox AI (Jan 2025); Kieback&Peter 49% (Nov 2025) | 14,000+ buildings, up to 25% energy reduction |
| Johnson Controls / OpenBlue | OpenBlue agents (since Nov 2024) | Nantum AI (Apr 2026); FM:Systems ($455M) | Enterprise portfolio + workplace/IWMS layer |
| Siemens / Building X | 4 named GenAI agents (Mar 2025) | 5-stage autonomy roadmap (organic build) | Building X global rollout |
| Honeywell / Forge | Google Cloud Gemini agents | LenelS2 ($4.95B, access/security) | Dual-hyperscaler deployment |
| Carrier / Abound | Abound platform | — | 63,000+ buildings optimized |
BrainBox AI and Nantum AI — two of the most-cited independent building-AI names of the prior cycle — were both absorbed by incumbents inside a fifteen-month window. The lesson for a buyer is not "avoid startups." It is that the independent AI you sign today may be an OpenBlue or Building X feature by your next renewal, and your contract needs to survive that transition.
The independent alternative — and why it is not dead
Against the incumbents sits a class of vendor-agnostic platforms whose entire pitch is that they do not own the equipment. KODE Labs (KODE OS) and Facilio are the two clearest examples for a commercial portfolio.
KODE Labs raised a $30M Series B in April 2024 (led by Maverix Private Equity, with TELUS Ventures and I Squared Capital) and, critically, won a $14M General Services Administration contract announced 24 February 2026 to deploy a unified smart-building interface across 150 federal buildings in the U.S. National Capital Region. Cadillac Fairview has KODE OS live at Toronto-Dominion Centre and RBC Centre, and channel distributor OTI now resells the platform. KODE claims up to 30% average reduction in energy and carbon.
Facilio's differentiator is architectural: it is platform-first, requires no proprietary hardware, and overlays existing BMS and controllers regardless of protocol — unifying maintenance, compliance, energy, and tenant experience in one real-time interface. When a government buyer with 150 buildings and an owner-operator like Cadillac Fairview both pick an independent overlay over an incumbent stack, that is a signal the "open layer" thesis has real procurement traction, not just conference-stage enthusiasm.
The APAC read
Both independents have planted APAC flags — Facilio runs offices in Singapore and Chennai, and the region's regulatory forcing functions are getting sharper. For Taiwan operators specifically, the pressure is now structural: the 2026 hyperscale data-center PUE ≤ 1.3 / colo ≤ 1.4 mandate and Taipower's northern-Taoyuan >5MW connection constraints mean the buildings that most need a coordinating platform layer are the ones already under a hard efficiency gun. A horizontal OS that can prove portfolio-wide energy performance to a regulator — with exportable, auditable data — is worth more in Taipei and Singapore than in a market with no penalty wall. See our prior coverage of the NTT DATA / Daikin coordinated-control PoC for how the same coordination logic is playing out at the equipment layer.
Here's what I'd do if this were my building
If I were specifying a building OS for a portfolio in the next 90 days, I would treat the platform decision as a 10-year lock-in and refuse to sign without three clauses:
- Data portability, in writing. The single clause that survives consolidation: I own my normalized point data (in an open schema — Brick/Haystack tags), and I can export the full historian if I leave. Without this, an incumbent acquisition of my vendor becomes an incumbent acquisition of my leverage.
- Overlay before rip-and-replace. Independents like Facilio and KODE prove you can unify a mixed-vendor estate without discarding working controllers. I would demand a hardware-agnostic overlay path and pilot it on one messy building first, not the flagship.
- Agent accountability. Every incumbent now ships a "GenAI agent." I would require that any autonomous control action produces an M&V-grade, exportable audit trail — because an agent that saves 25% but can't prove it to a regulator or an insurer is a liability dressed as an asset.
The buyers winning in 2026 are not the ones who picked the "best" platform. They are the ones who kept their data portable while the market consolidated around them. For deeper vendor and funding context, see the proptech-capital briefings and the full AISB Library.
Have a question about this topic? Ask our CRE AI Agent →