Where AI is heading — and what it could mean for construction.
AI can hold roughly 5x more project information in one request.
Today
2M tokens
Future
—
Why it matters for construction
More specs, contracts, RFIs, emails, meeting minutes, and project records can be reviewed together instead of split into smaller pieces.
Tracked metric: 2M tokens today → 10M-token frontier
More context helps the model see more at once; it does not guarantee perfect judgment.
Many AI chips can be linked together to work like one large machine.
Today
Current large AI systems
Why it matters for construction
Larger AI systems can handle heavier workloads: large drawing sets, document review, image reasoning, and multi-step agents.
Tracked metric: Google Ironwood pod: up to 9,216 chips
Specialized AI chips are getting more work done with less power and less cost pressure.
Why it matters for construction
This can help AI tools become cheaper to run and more practical to deploy across projects.
Tracked metric: Trainium3 UltraServers: up to 4.4x compute, 4x energy efficiency, ~4x memory bandwidth vs Trainium2
AI and digital demand are pushing data centers toward much larger electricity needs.
2024
415 TWh
2030
—
Global data-center electricity (TWh)
Why it matters for construction
This creates demand for power, cooling, electrical, mechanical, civil, substations, backup power, and mission-critical delivery.
Tracked metric: 415 TWh in 2024 → 945 TWh by 2030
Timeline
Viable usage window
Controlled environments: 2026–2030; open jobsites: 2030+
Current efforts
Boston Dynamics Atlas industrial deployments; Figure AI factory and logistics pilots; Tesla Optimus development.
Projected construction uses
Prefab shops, warehouse staging, material handling, laydown yards, cleanup, and controlled logistics.
Timeline
Viable usage window
Practical pilots now; broader adoption: 2027–2030
Current efforts
Trimble SiteVision, BIM coordination workflows, Apple Vision Pro enterprise pilots, Autodesk-style model coordination.
Projected construction uses
Field overlays, installation guidance, punch walks, QA/QC, safety training, owner walkthroughs.
Timeline
Viable usage window
Available now in narrow use cases; broader adoption: 2026–2030
Current efforts
Smart helmets, proximity sensors, fatigue monitoring, connected PPE, voice-enabled field assistants.
Projected construction uses
Heat and fatigue alerts, worker location awareness, voice notes, incident capture, documentation, site access workflows.
Timeline
Viable usage window
Available now; broader trade adoption: 2025–2028
Current efforts
Dusty Robotics FieldPrinter, HP SitePrint, Trimble robotic total stations, Hilti drilling and layout automation.
Projected construction uses
Layout, drilling, scanning, progress capture, QA/QC documentation, repeatable layout in corridors, slabs, and interiors.
Timeline
Viable usage window
Controlled logistics routes now; broader supply-chain impact: 2027–2035
Current efforts
Waymo autonomous driving, Aurora autonomous trucking, Kodiak Robotics trucking pilots, Torc Robotics, autonomous yard trucks.
Projected construction uses
More predictable material deliveries, lower freight volatility, automated yard movement, material tracking, site logistics, and just-in-time planning.