Economics of a Megawatt of AI Data Center

Key Takeaways

• Capex per MW totals $59 M, half for GPUs and servers.
• Pure infrastructure lease generates $15 M annual revenue per MW.
• Adding managed services can double revenue to $30 M+ per MW.
• Opex stays low at $1‑1.1 M, enabling >$1 M net profit.
• Payback period roughly four years on infrastructure alone.

Pulse Analysis

The rapid expansion of generative AI workloads has turned data‑center capacity into a strategic asset, but the sector’s growth hinges on clear financial metrics. By quantifying the cost of a megawatt of compute, Cruseo provides a template that investors can use to compare AI‑specific facilities against traditional colocation or hyperscale models. The $59 M upfront investment, with roughly half allocated to high‑density GPU racks, reflects the premium placed on raw processing power, while the remaining spend secures power‑dense infrastructure capable of handling the massive electricity draw.

A deeper look at the expense breakdown reveals why AI data centers differ from conventional sites. The IT component—GPUs, CPUs, networking, and storage—now commands a larger share of capital because performance per watt drives competitive advantage. Meanwhile, the physical plant, including on‑site power generation and advanced cooling, remains essential to sustain continuous high‑load operations. This split mirrors a broader industry shift toward vertically integrated facilities that own their power source, reducing reliance on external grids and improving reliability for latency‑sensitive AI services.

Revenue dynamics further tilt the equation in favor of rapid payback. Pure infrastructure leasing already delivers $15 M per MW annually, and adding managed services or cloud‑layer offerings can push total earnings beyond $30 M per MW. With operating expenses hovering around $1 M, profit margins exceed 90 %, translating to a four‑year return on the core build‑out. Such attractive economics are prompting venture capital and corporate investors to fund AI‑centric campuses, accelerating the deployment of next‑generation compute and reshaping the competitive landscape of cloud services.