BIMnopoly
Does the rise of physical AI mean the end of format monopolies in AEC? A structural shift is coming, away from controlling formats and towards controlling the intelligence layer that interprets the built environment
For more than two decades, BIM has been hailed as the industry’s path to a shared, information-rich representation of the built asset.
This coupling has produced a vicious lock-in: not simply a preference for a tool, but instead a dependency on the tool’s complete ecosystem to gain the ability to interpret the data it stores.
Until now, this hindered the ability to create new AEC tools, as they required ’permission’ from the platforms to integrate with them and could at best serve as a plug-in. Natural data migration between competing platforms? A complete no-go!
But all of that is about to change in the age of Physical AI, where matter becomes code.
Discover what’s new in technology for architecture, engineering and construction — read the latest edition of AEC Magazine
Subscribe FREE here
Power to the users!
While the first generation of AI was mainly conceptual we are beginning to see a rise of complexity leading to industry-grade outputs. Fields like coding, web design and other text-based professions are being completely disrupted, while technical fields remain mainly untouched.
Yet, a new class of AI is beginning to emerge, that of matter. Physical AI presents a completely new paradigm. Through understanding space, material and geometry, it promises to add reason to fields like construction, engineering and manufacturing – fields that have been rarely disrupted since the Industrial Revolution.
Not only does physical AI threaten to wipe out the planners, it also threatens to wipe out the planning platforms. Large multimodal models, geometric reasoning, graph learning and domain-specific ‘physical AI’ linking digital representations to construction logic are changing the game.
The critical shift is not that AI can ‘generate models faster’, but that it can decouple meaning from format by learning the syntax and semantics of building information across heterogeneous sources.
If meaning is no longer confined to a single platform’s internal representation, the strategic value of closed formats and gated ecosystems declines.
This article examines how BIM gatekeeping emerged, why BIM data has historically been difficult to standardise and reuse across project phases, and how AI-driven semantic interpretation may reconfigure interoperability, platform competition and the practical realisation of true BIM as a single source of truth.
From gated community to life on the cloud
Formats are not exclusive to AEC, of course. But nowhere are they more evident. While coding platforms produce universal code, BIM is yet to produce a truly standardised open format, and most information is still locked inside planning software, rather than code.
Attempts to use IFC have been unsuccessful as they only allow reading, where writing leads right back to the original software. So how did this happen? Lock-in for CAD and BIM was first developed through grounded mechanisms: proprietary file formats, object libraries, interior attributes and application-specific parametric behaviours. In the beginning, control was enforced by desktop licensing, specialised plug-ins, and data structures optimised for authoring rather than exchange.
Cloud platforms repositioned control from ‘who can open the file’ to ‘who can access the data’. The gate no longer sits at the file boundary; it sits at the subscription boundary
In these days, format was the moat. DWG, Doc, PDF, PPT and more supported the creation of gated communities and very successful companies around company-owned formats.
In practice, however, mainstream BIM platforms have evolved into highly controlled ecosystems in which data, semantics and workflows are tightly coupled to proprietary platforms and now-centralised cloud services.
Interoperability was technically possible (and frequently demanded), even though it caused conflicts of interest between users and vendors. As the years went on, these formats started opening up and their makers advanced the service scope to maintain their leadership. BIM presented a new challenge as it contained much more than linear data. It contained truth.
As BIM migrated to cloud-hosted environments, gatekeeping intensified. Cloud platforms introduced new integration layers: centralised model coordination, permissions, API access, automated change tracking and managed data environments.
While these features can improve collaboration and governance, they also repositioned control from ‘who can open the file’ to ‘who can access the data’. The gate no longer sits at the file boundary; it sits at the subscription boundary. Data portability becomes a product choice, rather than an inherent property of the workflow. As the mergers and acquisitions race intensified, more and more software was added to siloed clouds, encompassing the whole value chain of the building. The message was: you’re either in or out!
In other words, the move to the cloud did not eliminate lock-in. It modernised it. BIM as an ‘empty canvas’ The informational quality of a BIM model depends on manual data input: naming conventions, parameter definitions, object classification, assembly logic, and discipline-specific attributes.
This creates a systemic vulnerability. If the modelling team does not invest heavily in data governance (‘datasitting’) and constant validation, the model may contain untrustworthy semantics. The result aligns with a famous phrase in the coding world: garbage in, garbage out.
AEC project delivery is segmented by contract and liability. Each stakeholder is responsible for a different scale, creating large gaps.
• Geometry: Architects who produce the initial BIM model typically model design intent and code compliance coordination.
• Numbers: Contractors require constructability logic, sequencing assumptions and procurement-ready quantities. These often remodel a BIM model for VDC.
• Details: Fabricators require fabrication-level tolerances, connections and manufacturing constraints, ideally working with DfMA models.
• As-made: Facility operators require asset registers, maintainability attributes and operational metadata, requiring as-made digital twins.
Because responsibility is partitioned, the model is repeatedly rebuilt, reinterpreted, or ‘up-modelled’ across phases. Even when geometry is reused, semantics often are not. This leads to duplication of effort, loss of intent and gaps between disciplines – especially at the critical transitions from design to pre-construction, and from construction to fabrication.
Standardising the gap
The industry has open standards and classification systems, but in practice, object semantics vary widely. The same physical element may be represented differently across platforms, companies, or even teams within a single project. Custom parameters proliferate, internal object definitions diverge and data dictionaries become project-specific.
BIM platforms have historically achieved defensibility through proprietary representations of exactly the things that were meant to be open:
• Parametric behaviours: Rules for how objects interact and update.
• Metadata schemas: Parameter sets, classifications and property groupings tied to platform logic.
• Authoring-specific intent: Design history and modelling operations that are not easily exported.
• Project graphs: Internal relationships between elements, systems, spaces and constraints.
When this information is encoded in platform-specific schemas, portability is inherently ‘lossy’. If we can encapsulate this data in open language, we can create true BIM.
AI’s disruptive potential in this context is frequently misunderstood. The primary mechanism is not content generation; it is semantic interpretations of meaning. Modern AI systems can learn to extract and normalise meaning from heterogeneous building information, including partially structured BIM exports, drawings, schedules, specifications, point clouds and construction documentation.
In AEC, ‘syntax’ can be read as structured patterns – geometric relationships, typological configurations, adjacency networks, system hierarchies, and recurring assemblies – rather than merely text grammar.
This ability to infer meaning makes it possible to interpret building data even when parameters are inconsistent, incomplete or platform-specific. It is precisely the scenario where traditional BIM interoperability struggles.
Let architects plan like architects. Let contractors work like contractors. AI removes the need for data structure. Whether its LLMs, computer vision or even OCR, we are already at the tipping point of the end of format
Attempts have been made in the past to build a universal language of building representations as well as IFC, though they failed for the same reason formats thrive- inconsistency.
What I propose is not to teach new standardisation methods, but the exact opposite.
Let architects plan like architects. Let contractors work like contractors. AI removes the need for data structure. Whether its LLMs, computer vision or even OCR, we are already at the tipping point of the end of format.
Physical AI and new world models go a step deeper and allow us to not just read data, but to understand its meaning, and reason.
In effect, AI becomes an interpreter above the platform level. This comes with implications in terms of interoperability, platform competition and digital twins. If semantic interpretation is no longer exclusive to incumbent platforms, three changes follow:
1) Data standardisation becomes enforceable, not aspirational
Standards typically fail in practice when enforcement is manual. AI enables automated checking, mapping and validation of schema compliance, classification correctness and data completeness. This can shift standardisation to support higher-quality digital twins because model data becomes more consistent and auditable.
2) From design aid to knowledge source
A BIM model begins to function as a queryable knowledge representation of an asset rather than a static design layer. With AI-driven semantics, BIM can better support lifecycle use cases: quantity/ cost analytics, constructability assessment, fabrication readiness and operational asset intelligence.
This repositions BIM closer to its original intent: a persistent source of truth supporting a living digital twin.
3) Blue versus red pill: “Do you really want to know?”
This question comes at a pivotal moment in The Matrix, occurring just before Morpheus explains that the Matrix is a “prison for your mind”. Just like the Matrix represented a closed garden that ‘protects’ one from the real world, BIM provides a closed garden of protected data.
Like any opacity, democratising data presents a power shift in hierarchy, increasing liability and equalising responsibility. This is not always favourable for stakeholders who wish to keep their cards close to their chest or to remove liability as much as possible. We should not underestimate the nature of the industry when talking about tech integration. There are no Platonic solutions in the real world.
Conclusion
BIM gatekeeping has been sustained by a tight coupling between meaning and proprietary representation. Platforms have acted as the primary silos of building data, while output was flattened 2D data. AI disrupts this advantage by enabling meaning extraction, semantic mapping, and normalisation across various sources – effectively reducing format exclusivity as a barrier.
As this capability matures, closed ecosystems lose strategic leverage, interoperability becomes less about file exchange and more about semantic continuity, and the industry has a renewed pathway toward BIM’s original promise: a reliable, lifecycle-spanning source of truth that can serve as the backbone for the industry.
In that sense, Physical AI is not simply the next feature cycle in BIM software. It is a structural shift from controlling the format toward controlling the intelligence layer that interprets the built world.
Tal Friedman is an architect and designer working in the field of parametric design with an emphasis on structural systems and design for fabrication
The post BIMnopoly appeared first on AEC Magazine.
Powered by WPeMatico