Brookfield’s AI Investment Primer: Key Terms Shaping Capital Flows

AI investments in plain English...
This Brookfield primer offers one of the clearest overviews we've seen of AI investments.

Here’s a quick cheat sheet on the emerging AI vocabulary shaping the outlook:

Agentic AI:
AI systems that can plan, strategize, and execute tasks independently, adapting to new information without constant human oversight.

AI Factories:
Large-scale digital hubs purpose-built for AI workloads, with dense GPU clusters, advanced cooling, and high-speed networking.

Artificial General Intelligence (AGI):
AI with human-like ability to perform diverse cognitive tasks, including unfamiliar ones.

Artificial Narrow Intelligence (ANI):
AI that is highly capable in a specific domain but lacks general adaptability.

Artificial Superintelligence (ASI):
A hypothetical AI that surpasses human intelligence across all domains and can improve itself without human input.

Compute:
The processing capacity of hardware — especially GPUs — that powers AI training and inference.

Gigawatts (GW):
A measure of power capacity used to quantify the enormous energy demands of AI infrastructure.

GPU:
A graphics processing unit designed for highly parallel workloads; in AI, GPUs excel at training and running large-scale models efficiently.

GPU as a Service:
The rental of high-performance AI chips on-demand, avoiding the need for massive hardware investments.

Hyperscale:
Ultra-large data centers, often run by tech giants, built to serve massive computing needs at global scale.

Jevons Paradox:
The principle that efficiency improvements often increase total consumption rather than reduce it.

Moore’s Law:
The historical observation that computing power doubles roughly every two years with minimal cost increases.

Power Density:
The amount of electrical power a server rack consumes — AI racks often require 10× more than traditional ones.

Quantum Computing:
A next-generation computing approach using quantum mechanics to solve problems exponentially faster than classical computers.

S Curve:
The adoption pattern of new technology: slow start, rapid acceleration, then plateau.

Scaling Laws:
The finding that bigger AI models trained on more data tend to perform better.

Scratchpad Reasoning:
A technique where AI models “show their work” before giving a final answer, improving accuracy on complex tasks.

System 2 Architecture:
An AI approach optimized for deep reasoning and planning, as opposed to quick pattern recognition (“System 1”).

------

Think this won’t affect real estate?

It already has. Infrastructure allocations are rising, and they’re coming at the expense of real estate allocations.