Artificial intelligence has had a busy few years.
It’s written code, drafted contracts, generated artwork, eaten entire industries for breakfast — and now it’s coming for the physical world, too. The next frontier in the great AI arms race isn’t another flashy chatbot or a neural network that paints sunsets; it’s infrastructure. Land, power, copper, concrete, water, fiber. Very old-school things for a very new-school technology.
And Brookfield Asset Management’s new $10 billion Artificial Intelligence Infrastructure Fund, launched with Nvidia and the Kuwait Investment Authority, is the latest — and perhaps most aggressive — indication that the built world is about to become the center of the AI universe.
The fund aims to develop and acquire up to $100 billion in AI infrastructure assets, including data centers, AI factories, gigawatts of power generation, and the specialized hardware-ready environments that AI and ML workloads now demand. It’s one of many signals that the global economy is shifting into an era where real estate, not just algorithms, is the deciding factor in who wins the next wave of innovation.
For corporate occupiers, especially large-scale tenants operating across multiple markets, the implications are nothing short of transformative. This is not another hype cycle — it’s a structural shift in the CRE landscape.
Let’s break down what’s happening, why it matters, and what the C-suite needs to know to survive (and ideally, benefit from) the biggest infrastructure buildout since the modern power grid.
AI Development: $7 Trillion, Gigawatts, and Land- Lots of Land
Brookfield projects that the global AI buildout over the next decade will require $7 trillion in capital across:
- Data centers
- Power generation
- Advanced compute
- Distributed file systems
- Scalable storage solutions
- Cloud-based AI infrastructure
That’s not a typo. Seven. Trillion. Dollars.To put that into perspective:
- The entire U.S. commercial real estate market is valued at ~$20T.
- Global telecom networks cost ~$14T to build over 40 years.
- AI wants half of that — in 10 years.
This is why Brookfield, already managing more than $115B in digital infrastructure, renewables, and semiconductor manufacturing, is tripling down. The company has committed:
- SEK 95B (~$10B) to an AI data center campus in Sweden
- €20B in AI projects across France
- $5B with Bloom Energy for 1 gigawatt of behind-the-meter power for AI factories
And they’re not alone. CBRE just spent $1B buying Pearce Services to scale data-center-aligned infrastructure services. Every major cloud provider is hoarding capacity. Sovereign wealth funds are piling in. And corporate tenants? Many have no idea they’re about to get pulled into the blast radius.
AI Workloads Need Physical Infrastructure
The public thinks AI is virtual. Something floating in the cloud, abstract and sleek. The truth is far messier — and far more physical.
Modern AI systems depend on a coordinated ecosystem of hardware and software components that must operate at extreme throughput, low latency, and massive scale. Consider the stack:
1. Compute: GPUs, TPUs, and Specialized Hardware
Traditional central processing units (CPUs) can’t keep up with the parallel processing demands of:
- Deep learning
- Generative AI
- Matrix and vector computations
- Large-scale model training
This is why Nvidia’s graphics processing units (GPUs) and tensor processing units (TPUs) have become the gold standard. These chips require:
- Dense electrical capacity
- Advanced cooling
- Ultra-high-throughput fiber
- Physical space for expansion
All things data centers didn’t traditionally have.
2. Data Storage: The Rise of Scalable and Distributed Systems
Model training isn’t just compute-intensive — it’s storage-hungry.
Datasets for ML models are growing exponentially, requiring:
- Distributed file systems
- Scalable storage systems
- High-performance data lakes
- Version control systems for model development
This is not your standard enterprise NAS closet.
3. Software: ML Frameworks and Data Processing Libraries
AI workloads require a different software ecosystem:
- Machine learning frameworks (TensorFlow, PyTorch, JAX)
- Data processing frameworks (Spark, Ray, Dask)
- Data ingestion and analysis pipelines
- Infrastructure for model evaluation, deployment, and ongoing monitoring
These systems are deeply integrated with the physical environment. Your building’s HVAC and mechanical systems suddenly matter to your CIO’s algorithms. Welcome to 2025.
4. Energy: The New Kingmaking Constraint
The AI race is not about code — it’s about power.
Training advanced AI models consumes 10x–100x the energy of traditional IT environments.
This is why:
- Brookfield is backing nuclear reactors
- Data center operators are signing 10–20-year PPAs
- Energy-rich markets (Nordics, Texas, MENA) are becoming AI magnets
- Tenants are facing power-scarcity-driven rental spikes
Your next office location decision might hinge on grid capacity, not commute time.
So What Does This Mean for Corporate Tenants?
Here’s where the story stops being abstract and starts getting deeply relevant — and a little uncomfortable — for large-scale occupiers.
1. Competition for Power and Space Will Reshape Pricing
Data centers are absorbing enormous chunks of local grid capacity. In some metros:
- Power is being rationed
- Costs are rising
- Timelines for utility upgrades are extending from 12 months to 5–7 years
Corporate campuses, manufacturing, R&D facilities, and even office towers will feel the squeeze.
Expect energy scarcity to become a core CRE variable.
2. Zoning and Land Availability Are Tightening
Municipalities are fast-tracking land use approvals for AI facilities because they bring jobs, investment, and prestige. That means:
- Some submarkets will rezone for digital infrastructure
- Certain parcels will become uncompetitive for non-AI uses
- Corporate tenants may find themselves “priced out” by AI factories
Put simply: the coolest new neighbor on your block might be a 1-gigawatt hyperscale campus.
3. Artificial Intelligence Infrastructure Is Becoming a Tenant Obligation
Most large tenants already rely on:
- Cloud-based AI infrastructure
- ML training workloads
- Generative AI applications
- Complex data ingestion and processing pipelines
- Machine learning infrastructure embedded in operations
But as AI initiatives scale, enterprises are discovering a new pain point: AI needs space. Real space. Not just server closets — but:
- Local compute rooms
- High-density racks
- Model deployment nodes
- On-prem systems to protect sensitive data
- Redundancy systems to safeguard uptime
“Traditional IT infrastructure” was easy. AI infrastructure? Not so much.
4. CRE Negotiations Will Soon Require AI Literacy
Five years ago, no corporate tenant ever asked:
- What’s the building’s power-to-floor-plate ratio?
- Can the site support parallel processing capabilities?
- How compatible is the building with distributed storage systems?
- Will our ML models achieve low latency in this metro?
Today? These are becoming standard RFP questions for advanced AI enterprises. AI requirements are seeping into site selection. If your lease doesn’t reflect this yet, it will.
5. Existing Systems May Become Obsolete Faster Than Expected
Mechanical systems, electrical capacity, fiber backbones, and cooling infrastructure all age at double speed when servicing AI workloads.
Companies relying on:
- Traditional IT environments
- Legacy data storage
- Non-redundant power systems
…are discovering that AI doesn’t politely fit inside existing constraints.The question becomes: Do you retrofit? Or relocate?
How Large Tenants Should Respond Now
Here’s the uncomfortable truth: tenants who wait will lose leverage. Those who move early will gain an advantage.
1. Conduct an AI-Infrastructure Readiness Assessment
Evaluate:
- Power redundancy
- Cooling capacity
- Data throughput
- Security requirements for sensitive data
- Space for specialized hardware
- Capacity for AI and ML workflows
This is the new due diligence.
2. Add AI Infrastructure Requirements to All RFPs
Site selection must now account for:
- Latency to cloud regions
- Proximity to data centers
- Available substation capacity
- Local permitting environment
- Renewable energy access
- Scalability for future ML applications
You are no longer choosing a building. You are choosing an ecosystem.
3. Renegotiate Leases With AI-Specific Clauses
Savvy tenants are now asking for:
- Guaranteed access to additional electrical capacity
- Ability to deploy AI hardware and software systems onsite
- Rights to expand mechanical needs
- Power-cost predictability clauses
- Fiber-upgrade allowances
AI infrastructure is a moving target — your lease should flex with it.
4. Create a Cross-Functional AI + CRE Strategy Team
Facilities, IT, data science, and procurement must now collaborate. Why? Because CRE decisions directly impact:
- Model performance
- Efficient model training
- Data analysis and storage
- AI task throughput
- Feature engineering
- Machine learning algorithms
Your building is now part of your algorithm.
The Bottom Line: AI Infrastructure Is Not Someone Else’s Problem
AI may be the most powerful digital transformation in history, but its future is profoundly physical. Land, power, cooling, storage, and compute are becoming the backbone of AI development — and therefore the backbone of enterprise competitiveness.Brookfield’s $100B AI infrastructure program isn’t just a fund. It’s a flare gun fired into the sky. It signals:
- Where capital is going
- Where global competition is heading
- And where every large tenant must evolve
The winners in the next decade will not just be the companies with the best AI models — but the companies with the best infrastructure strategy.If your CRE planning doesn’t already account for AI workloads, ML models, scalable storage systems, specialized hardware, and the rising cost of power…you’re not behind the curve.You’re not even on the field.
