Microsoft Fairwater
FJI source-qualified: four-document gate met, including project-specific government evidence. Qualification evidence and decisions →
Project status: First facility operational; second facility targeted for 2028. Capacity scope: First facility operational; current operating MW not established. How we verify evidence.
← Back to project indexExpanded analysis · Original disclosures linked below · Amounts in USD · Company claims attributed; analyst interpretations labeled.
The one-minute investment takeaway
Fairwater’s first facility has crossed from construction into operation. That changes the investment question: the key issue is now how effectively Microsoft monetizes an operating AI system, rather than whether a large building can be completed. The adjacent second facility remains a future development.
Readiness basis: Classification of evidence in the reviewed profile; strongest documented phase, not full-campus completion. First facility operational; second facility targeted for 2028 Readiness source →
What the sources establish
- Operator / company disclosure Operating milestone. Microsoft announced the first facility fully operational on June 23, 2026, after startup activities in April. Its second Mount Pleasant facility was under construction with completion scheduled for 2028. Source ↗
- Operator / company disclosure Investment boundary. Microsoft’s estimated USD 4.7 billion Wisconsin hyperscale construction spending spans 2024–2028. It is not a disclosed GPU budget, current-year expense or standalone cost of the first facility. Source ↗
- Operator / company disclosure Technical connection. The September 2025 design disclosure identifies a system interconnecting hundreds of thousands of NVIDIA GPUs for training and inference. Treat that as a disclosed architecture and scale description, not an independently measured October 2026 installed inventory. Source ↗
Who receives the money?
Our analysis of the disclosed company roles. Revenue routes below are business-model interpretations, not estimates of booked project revenue.
| Company / exposure | Economic role | What investors need to establish |
|---|---|---|
| Microsoft / MSFT | Owns and operates the AI infrastructure; monetization can occur through cloud services and products. | A facility creates capacity. Revenue requires customer consumption or product adoption; neither is reported here by site. |
| NVIDIA / NVDA | Confirmed accelerator connection in Microsoft’s design disclosure. | Supplier revenue depends on equipment contracts and delivery timing, not Microsoft’s total construction spending. |
| Other infrastructure vendors | Potential electrical, cooling and networking demand. | No project-specific vendor award is inferred solely from a company’s product category. |
How financially meaningful is it? Analyst interpretation
Our analysis: Fairwater is strategically relevant to Microsoft’s AI capacity, but site-level revenue and earnings are not disclosed. Training may support future products rather than immediately billable third-party consumption. That creates a timing mismatch between upfront hardware and construction commitments and eventual revenue. A completed facility can improve capacity while initially lowering cash flow or increasing depreciation.
For sold compute: available accelerator-hours × paid utilization × realized price. For internal training: assess resulting product revenue or cost savings separately; do not count training expenditure as sales.
Illustration: if paid utilization improves but price per compute-hour falls by the same proportion, revenue need not improve. For internal workloads, higher utilization can still create value, but it requires evidence of better products, cheaper delivery or durable customer adoption.
Milestones that would change the assessment
| Milestone / disclosed target | Evidence to watch |
|---|---|
| First facility: operating | Look for deployment, reliability and utilization updates that go beyond the completion announcement. |
| Second facility: 2028 target | Watch steel, utilities, equipment readiness and commissioning separately; construction completion is not proof of paid workloads. |
| Financial follow-through | Compare cloud and AI monetization with depreciation, capital spending and operating margins; avoid attributing company trends entirely to this site. |
What could break the investment case?
Our analysis: Large fixed commitments amplify the cost of underutilization. Hardware generations may change before assets earn back their costs. Operating constraints, customer concentration, power costs and local concerns can affect expansion or ongoing operation.
What would change our view?
Our analysis: The thesis weakens if new capacity does not relieve demand constraints or product revenue fails to cover increasing depreciation and operating costs. Verified paid utilization, efficient delivery and stable margins would provide stronger support than another capacity headline.
What remains unknown Not established in this review
No campus revenue, profit, paid utilization or verified current GPU inventory is established. The operator’s fully operational statement describes the first facility; it does not imply all future facilities are complete.
This is project research, not a valuation model or a buy recommendation. Financial illustrations are assumptions, not management guidance. Review current company filings and valuation before making an investment decision.
Company investor information & quarterly statements
Microsoft
Owner/operator
Official investor & shareholder information · Quarterly financial statements: FY2026 Q4, quarter ended June 30, 2026
NVIDIA
Disclosedacceleratorplatform
Official investor & shareholder information · Quarterly financial statements: Q2 FY2027 ended2026-07-26
Corporate disclosures provide company-wide context. They do not establish this project’s revenue, spending, profit, ownership share or operating capacity. Disclosure verification records →