We've just launched the Hyperscaler Deals page, a dedicated tracker that compiles every announced AI infrastructure contract between the public companies we cover and their hyperscaler clients (CoreWeave, Microsoft, AWS, and the rest), and surfaces the deal economics that actually matter to investors.
Why we built a tracker for hyperscaler deals
Hyperscaler deals are the clearest signal we have that a miner is becoming a serious AI infrastructure provider, but the announcements are scattered across press releases, 8-Ks, and earnings calls, and the disclosure formats are wildly inconsistent. One company reports critical IT load, another reports gross capacity. One quotes total contract value, another only annualised revenue. Some lock in fixed terms; others structure as joint ventures.
If you want to compare CORZ's CoreWeave contracts to APLD's, or stack TeraWulf and IREN side-by-side on revenue per megawatt, you've had to build the spreadsheet yourself. We've now done that work for you, normalised the numbers to critical IT load, and made it filterable.
What's on the page
The top of the page summarises the entire tracked dataset in four KPI cards: Total Contracted MW, Total Contracted Revenue, Avg Deal Duration, and Avg Revenue Per MW. These are the numbers we get asked about most often in DMs, and they update automatically as new deals are added.
Below the KPIs are three charts that benchmark the sector at a glance:
- Contracted MW by Company: who has the largest book of AI infrastructure capacity under contract
- Contracted Revenue by Company: the same view in dollars, so you can see whose deals are commanding premium pricing
- Deal Timeline: a bubble chart showing when deals were announced and how big they were, which makes the recent acceleration in deal activity obvious
The main event is the deals table, with one row per announced contract. The columns are designed so you can compare deal economics directly:
- Company and Client (e.g. CoreWeave, Microsoft, AWS)
- Service Type: colocation, AI Cloud / GPU-as-a-Service, or AI Compute JV
- Contract Size in MW
- Contracted Revenue (total)
- Duration in years
- Rev/MW and Annual Rev/MW
- Announced date
Three filters sit above the table (Company, Client, and Service Type) so you can isolate, say, all CoreWeave deals across the sector.
Click any row to expand it and reveal the full context behind that contract:
- Extension: renewal options and extension terms (e.g. "Three 5-year renewal options")
- Deal Source: a direct link to the source document (press release, 8-K, investor deck) that each number was drawn from, so you can verify the data yourself
- Note: methodology notes specific to that deal, including critical IT load caveats and any non-standard economics (for example, deals structured as a fixed infrastructure fee plus revenue share rather than a traditional hyperscaler lease)
The link to the data source, as usual, is a Professional feature for each deal.
How you can use it
A few ways we think this will be useful day to day:
- Benchmark a single company's deal economics against the rest of the sector. Is CORZ pricing its CoreWeave capacity at a premium or a discount to peers?
- Track hyperscaler exposure: how concentrated is each miner on a single client like CoreWeave or Microsoft?
- Spot the next mover. When a new deal lands on the page, you can immediately see how it compares to everything that came before.
- Model future revenue contribution by combining MW under contract with the per-MW numbers to size the AI revenue line in your own models.
A note on methodology: companies disclose deal terms differently, and not every deal is announced with full economics. We standardise everything to critical IT load rather than gross capacity (so a Cipher Mining deal can be compared directly to a Core Scientific deal), and we cite the source disclosure for every row. There's a full FAQ section at the bottom of the page covering the definitions in detail.
This is Version 1. If there a new feature you'd like to see added, hit reply on this post or message us. Most of the best ideas on the platform come straight from subscribers.
