In 2011, IBM ran a Smarter Planet ad with a confident headline: “Let’s build a smarter city.” The body copy described an instrumented, interconnected, intelligent urban system that would coordinate traffic, water, energy, public safety, and waste in real time. The implied buyer was a mayor or a CIO of a municipal agency. The promised payoff was efficiency, resilience, and the kind of public-sector legibility consultants had been promising for decades.
Reread that ad in 2026 and the language sounds correct but the trade was wrong. Almost everything specific the campaign promised got delivered eventually, in different form, by different companies, on different timescales. Almost nothing was bought by the customers the campaign was aimed at. The smart-city pitch of the late 2000s and early 2010s is one of the cleanest case studies in modern enterprise marketing of a real capability sold to the wrong buyer with the wrong economics.
I’m starting a piece about what “smart” means in 2030 with this story because it would be a mistake to lead with optimism without acknowledging the wreckage. Institutional CRE has spent fifteen years being pitched smart-building software. Most of what was sold either didn’t work or didn’t scale. The skepticism in the room when the next vendor walks in is earned.
This piece is about what was wrong with the 2010 model, what’s actually different now, and what the next decade requires.
2010, in their own words.
The two anchor campaigns were IBM Smarter Planet, launched in 2008, and Cisco Smart+Connected Communities, launched alongside in 2009. Both pitched the same essential proposition. Instrumentation would generate data. Data would feed analytics. Analytics would surface insight. Insight would optimize the operation. The buyer was the city or the mass-scale operator. The check was meant to be eight or nine figures.
Songdo, Masdar, and the early NEOM master-plans were the showcase deployments. Songdo was billed as the world’s first ground-up smart city. Masdar was the carbon-neutral exemplar. NEOM is still being built. Each spent or is spending tens of billions of dollars on infrastructure premised on the smart-city marketing of the period.
The story they told sounded right. The story they delivered was harder.
Why most of it didn’t work.
The honest postmortem is in three books. Anthony Townsend’s Smart Cities (Townsend, 2013, W. W. Norton) is the most generous of them. Shannon Mattern’s A City Is Not a Computer (2021) is the most analytical. Adam Greenfield’s Against the Smart City (2013) is the most polemical. Read together they triangulate the failure modes.
Top-down architecture. The 2010 model presumed a single municipal or master-developer buyer with the authority to specify a system across an entire urban geography. Real cities don’t work that way. Real master-planned developments encounter twenty different agencies, twenty different procurement timelines, twenty different sub-contracts. The architecture didn’t survive contact with the procurement reality.
No tenant economics. The smart-city pitch never had a clean answer to the question of who, specifically, was paying the recurring software-and-services bill. The municipality couldn’t budget for it sustainably. The asset owner couldn’t underwrite it without a documented revenue line. The tenant didn’t know it was being sold. The economic model defaulted to capex-and-maintenance, which expired with the warranty.
Vendor lock-in. The proprietary protocols of the 2010 generation locked customers into a single integrator for the asset’s useful life. Replacement cost was punitive. Switching cost was punitive. The structural negotiating position of the customer worsened over time. Many of the early showcase deployments are still paying off the bad procurement decisions of 2012.
Misaligned procurement. Municipal RFPs of the period were structured to compare hardware specifications, not data outcomes. Vendors competed on box-count and feature-list. The integrator who won the procurement was rarely the one who could most credibly deliver the data outcomes the customer eventually needed.
The combined effect: a generation of buyers learned to be suspicious of smart-city marketing. Most institutional CRE owners I work with today were exposed to one or another version of this pitch in the 2012 to 2018 period. The scars are real.
The Sidewalk Labs reckoning.
The clearest single inflection point in the public credibility of the 2010 smart-city model was the collapse of Sidewalk Labs‘ Toronto Quayside project in 2020. Mattern’s analysis of the project, and the Toronto-specific reporting in the Globe and Mail and The Logic, are worth reading in full. The short version: a thoughtful, well-resourced developer with Google’s backing produced a master-plan that ran straight into the structural problems above — community-trust collapse on data governance, vendor lock-in concerns, and an economics model the public-sector partner couldn’t underwrite.
Toronto Quayside was the canary. Within eighteen months several other high-profile smart-city programs quietly downsized. The market had absorbed the lesson.
The lesson, properly stated, was not that smart-city technology was wrong. It was that the procurement and economic model of the 2010 generation was wrong. The technology was, in fact, mostly capable of doing the things the marketing promised. The buyers were the wrong buyers. The contracts were the wrong contracts. The data ownership was the wrong shape.
What did work — and is now invisible.
Underneath the spectacle of the failed showcase deployments, the 2010 to 2020 period quietly delivered substantial infrastructure. ASHRAE‘s standardization of BACnet, the steady adoption of BAS protocols across the commercial inventory, the ten-times cost decline in commercial-grade IoT sensors, the maturation of cellular and Wi-Fi standards into something institutional-grade — none of this carried marketing. All of it now functions as infrastructure.
The commercial-grade IoT sensor that cost $400 in 2012 costs $25 in 2026. The PoE switch that anchored a Class A office’s data infrastructure in 2014 is a commodity. The deployment cost of a five-thousand-point instrumentation overlay on a 500,000-square-foot office in 2026 is a tenth of what it would have been in 2015 — and the data quality is meaningfully higher.
The infrastructure that didn’t have marketing got built. The platforms that had the marketing didn’t.
The 2026 reset.
Five things changed in the early-to-mid 2020s that make the 2026 conversation about smart real estate structurally different from the 2010 one.
One. Open data protocols matured. BACnet/SC. Matter. Brick Schema. Project Haystack. Model Context Protocol for AI-agent interoperability, now under Linux Foundation stewardship. The plumbing of an open, owner-controlled platform exists and is being adopted. The vendor-lock-in problem is structurally solvable.
Two. AI agent frameworks shipped. The same agent infrastructure now used by enterprise software runs facility-management operations meaningfully better than the rules-based BMS programs of 2015. Alarms triage, predictive maintenance, dynamic scheduling, work-order routing all run with markedly less human supervision than five years ago.
Three. Photoreal real-time engines became deployable. Unreal Engine runs the multi-stakeholder, photoreal twin of a 500,000-square-foot asset on a laptop in 2026. The technology is borrowed from film virtual production. The visual-fidelity gap that made the 2010-era smart-building UIs feel toy-like has closed.
Four. Capital markets started pricing ESG data. The CRREM pathway, GRESB scoring, the SEC climate-disclosure rule, and EU CSRD all reward sensor-grade data. The economic case for instrumentation no longer depends on capex savings alone. It depends on access to capital pools that are increasingly screening for compliant data.
Five. The buyer shifted from the municipality to the asset owner. The 2010 model required a single municipal customer with master-planning authority. The 2026 model installs at the asset level, with the institutional owner as the customer. The economics work because the buyer can underwrite the cash flow.
Why owner-controlled platforms now beat vendor-controlled ones.
The cleanest articulation of why the 2010 model failed is this: the data lived on the vendor. The 2026 model, when it works, puts the data on the asset.
When the data lives on the asset, the owner can extract it under standard schemas at any time. The owner can switch service providers without losing the historical performance record. The owner can present the data in offering memoranda at exit. The owner can underwrite the asset’s documented digital performance to LPs and lenders. The asset accretes value because the data does.
When the data lives on the vendor’s platform, none of those things are true. The owner rents access to their own asset’s performance record. The vendor controls the schema, the export format, the retention period, the migration cost. The exit value of the asset is bounded by the vendor’s willingness to release the data on the seller’s terms.
This is not a technical distinction. It’s a structural-economic one. It is the central architectural difference between the 2010 generation and the 2026 generation.
Five lessons to carry forward.
For practitioners building or buying real-estate technology in 2026, here are the five principles that earn the trust of the room.
One. Open data protocols, with the data on the asset, are non-negotiable. Vendors who can’t speak this language fluently in their first meeting are selling the wrong product.
Two. Lead with the asset owner’s underwrite. If the proposal doesn’t include an explicit answer to how the asset’s pro forma changes — at acquisition, through hold, at exit — it isn’t ready for institutional discussion.
Three. Treat instrumentation as the foundation, not the headline. The data infrastructure is the durable thing. The applications that run on top of it will change. The investment thesis runs on the data continuing to exist independent of the application.
Four. Build for the integrator-channel reality, not against it. The integrators who installed the 2010-era BMS are still the people who run the asset. The 2026 platform either composes with that workforce or it loses the deal. Most ambitious vendors lose the deal on this principle alone.
Five. Be honest about what the 2010 generation taught us. The market has a long memory. The vendor who walks in and pretends the prior generation didn’t exist is communicating, by omission, that they haven’t done the homework.
What “smart” should mean in 2030.
A working definition: a “smart” commercial real-estate asset in 2030 is one whose performance is legible — to its operator, its owner, its tenants, its lenders, its LPs, and its regulators — through standardized, owner-controlled, sensor-grade data, on which AI agents and other software run to optimize the asset’s recurring digital yield and to document its operational and embodied performance against capital-markets requirements.
That sentence is dense on purpose. Each clause is doing work. Legible — not just “instrumented.” Standardized — open protocols, not proprietary stacks. Owner-controlled — the data lives on the asset. Sensor-grade — telemetry, not estimates. AI agents — meaningful automation, not reactive dashboards. Recurring digital yield — a real revenue line, not a savings claim. Documented — capital-markets-ready, not marketing-ready.
This is the definition we’d like the industry to standardize on. The marketing-grade definition will be looser. The market will eventually price the difference.
The companies that get it.
A handful of vendors and integrators are doing serious work along the 2026 lines. They tend to share certain features: they are owner-aligned, not vendor-aligned. They publish their data schemas. They name their academic and standards-body affiliations. They show their work on cap-rate compression rather than savings claims. They were not the loudest companies at the 2018 smart-cities conferences.
A larger set of vendors are still selling the 2014 product with new logos. They tend to share certain features too: proprietary protocols, vendor-controlled data, marketing-grade metrics, and a tendency to talk about “operating systems for the built world” rather than about basis points of cap-rate compression. The market will sort them in the next 36 months.
This is not a category in which the cleanest brands are always the best products. Diligence on each potential platform partner remains the work.
The next decade.
Institutional CRE has a clean shot at the next decade if it learns from the last one. The trapped value is real. The infrastructure is mature. The capital-markets pricing is starting to move. The owners who instrument their portfolios early — under open protocols, on owner-controlled platforms, with academic provenance behind the work — will own the documented performance record that prices through 2030.
The owners who don’t will spend the next ten years watching their cost of capital widen against their better-documented peers.
The 2010 pitch was wrong because the procurement model was wrong. The 2026 opportunity is real because the procurement model is finally aligned with the asset owner’s economics. The lesson the prior generation paid for in dollars and reputation now serves as the guardrail.
Sources cited
- Townsend, A. (2013). Smart Cities: Big Data, Civic Hackers, and the Quest for a New Utopia. W. W. Norton.
- Mattern, S. (2021). A City Is Not a Computer: Other Urban Intelligences. Princeton University Press.
- Greenfield, A. (2013). Against the Smart City. Do Projects.
- Sidewalk Labs. (2019). Toronto Quayside Master Innovation and Development Plan. (Project closed May 2020.)
- Carvalho, L. (2015). “Smart cities from scratch? A socio-technical perspective.” Cambridge Journal of Regions, Economy and Society.
- Cugurullo, F. (2018). “Exposing smart cities and eco-cities: Frankenstein urbanism and the sustainability challenges of the experimental city.” Environment and Planning A.
- IBM. Smarter Planet / Smarter Cities marketing collateral, 2008–2014.
- Cisco Systems. Smart+Connected Communities marketing collateral, 2009–2014.
- ASHRAE. ANSI/ASHRAE Standard 135 (BACnet); BACnet/SC, Standard 135-2020.
- Connectivity Standards Alliance. Matter v1.2+ Specification.
- Anthropic. (2024). Model Context Protocol Specification.
- Brick Schema Consortium. Brick Ontology v1.3. (UC Berkeley initial development.)
- Project Haystack. Tagging Conventions. project-haystack.org.
Logan Grooms leads real-estate operations at Smart City Labs and has spent two decades in CRE, operations, and technology. Reach the SCL team here.