Every AI dollar
ends up in a datacenter.
Models change quarterly. Chips change yearly. But every model, every chip, and every AI workload on Earth runs inside the same thing: a powered, cooled, connected building. The datacenter is where the entire AI capital cycle lands — and it is the layer of the stack that behaves like infrastructure rather than technology.
Hyperscalers — Microsoft, Alphabet, Amazon, Meta — have collectively guided to roughly $725 billion of capital expenditure in 2026, up ~77% from a record $410 billion in 2025. The vast majority is being directed at datacenters and what fills them: GPUs, networking, power and cooling.
Demand has outrun the buildings. According to CBRE, vacancy in primary North American datacenter markets fell to a record-low 1.4% at year-end 2025, with 74.3% of the 9,432 MW of new supply preleased before delivery. Whyte Consolidated invests in the layer of that stack where value is created and defended: the facility itself.
One facility,
four layers of capital.
A fully equipped AI facility costs upwards of $40 million per usable IT megawatt. But that capital is not one investment — it is a ladder of layers with very different economics, durability, and risk. The discipline is knowing which rungs to own.
Land & power
Entitled acreage with secured megawatts — substation position, interconnection queue, water and fiber. The scarcest input and the cheapest point of entry on the ladder.
Shell & site infrastructure
The building, switchgear, cooling plant and distribution — more than $10 million per megawatt of hyperscale capacity. Real estate economics with utility-grade durability.
Fit-out & compute
Racks, networking, liquid cooling and the GPUs themselves — upwards of $30 million per IT megawatt. The most capital-intensive layer, and the fastest-depreciating.
Contracts & cash flow
Multi-year leases and take-or-pay compute offtakes from creditworthy tenants. The layer that converts the physical stack into financeable, institutional-grade cash flow.
The lower rungs — land, power, shell — hold their value across hardware generations: the same powered building that houses this year's GPUs will house the next three refreshes. The upper rungs depreciate fast but rent the lower ones to exist. Whyte Consolidated concentrates where duration lives: power-secured land, development, and the operating facility.
Capital is abundant.
Buildable megawatts are not.
The binding constraint on AI buildout is no longer money — it is power, equipment, and permission. Each is measured in years, and none can be solved by writing a larger check.
Power
Grid interconnection queues run years, not quarters. A site with secured megawatts is worth multiples of the same land without them — capital cannot cure power scarcity on any useful timeline.
Equipment
Transformers, switchgear and generation equipment carry multi-year lead times. Positions in the delivery queue are assets in their own right.
Entitlements
Zoning, water, noise and tax treatment determine whether a site can ever be built. Jurisdictions are tightening even as demand accelerates — approved sites become scarcer, not cheaper.
“We are compute constrained in the near term.”
Binance founder CZ said he prefers “the shovels of AI” — data centers, power supply systems, and large-scale compute — over AI applications themselves, framing the current cycle as “infrastructure-first.”
Datacenters are becoming
an institutional asset class.
The financing market is catching up to the physical one. NVIDIA's partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR aim to mobilize more than $500 billion of third-party capital for AI compute — datacenter project finance applied to GPUs, with bankruptcy-remote vehicles, take-or-pay offtakes and controlled accounts. IREN's Microsoft-backed transaction already funded ~96% of GPU capex at investment grade.
Every one of those financings needs the same thing underneath it: a powered, entitled, operating facility with contracted tenants. As compute becomes financeable, the facility layer becomes the collateral everyone underwrites — and the owners of power-secured capacity sit at the base of every capital stack being formed.
Whyte Consolidated focuses on the specific work that creates value in this environment: securing power-ready land, structuring development partnerships with creditworthy tenants, and operating to institutional standards — the four levers detailed in our strategy, applied across 3.6 GW under development.
- NVIDIA's $500 billion bet: turning compute into an asset class — the financing architecture, the IREN precedent, and the risks
- The coming infrastructure economy — physical assets, programmable ownership, and the new capital stack
- Proof of Useful Work and the 2-for-1 GPU — how matrix-multiplication consensus turns AI compute into a second revenue stream