Today I'm launching Dimetrics publicly, a research terminal for the AI-infrastructure trade: datacenters, neoclouds, and the bitcoin miners pivoting to HPC. I'm excited to be partnering with
@blockspace on distribution and usage of the platform!
Why I built it
The AI trade is driven by power and datacenters. To underwrite these names you need the data buried in SEC filings such as lease terms, individual site megawatt capacity, and the debt behind each site. I was having to spend hours going through filings and mapping out these metrics by hand every time. No tool did it, and the institutional ones aren't priced for an individual nor provide the level of granularity.
Inside the terminal
🧱 Thesis baskets: 17 AI-bottleneck themes (power shells, HBM, advanced packaging, grid equipment, photonics), each a live basket of public names
🏗️ HPC conversion tracker: a global site map, lease rates ($/kW), EV/MW multiples and the conversion spread as miners turn into HPC operators
⛏️ Bitcoin mining economics: network hashprice vs ASIC breakeven at your own power cost, difficulty and the next retarget
🖥️ GPU compute index: H100/H200/B200 spot prices across hyperscalers and neoclouds
🗺️ Datacenter political-risk map: county-level moratorium and zoning risk across every tracked site
📈 Markets: sub-index composites, correlations and Sharpe/Sortino across the space
Dimetrics MCP server
Dimetrics also ships as an MCP server. You can connect it to Claude, Claude for Excel, Codex or OpenClaw and pull any of this data into your own models, every figure sourced to the filing it came from:
🏗️ HPC lease economics: every disclosed lease: $/kW/month, total contract value, NOI, PUE, term
⚡ Power & sites: each operator's portfolio site by site: gross & critical-IT MW, energization pipeline, grid
operator / utility / county
💸 Capital structure: the debt financing each lease, tranche by tranche, tied to the deal it funds and cited to
EDGAR
🗺️ Datacenter political risk: county-level moratoriums and regulatory scores, so you see where the next permitting fight kills a buildout
Dimetrics is built by an individual analyst for the individual analyst trading neoclouds and datacenter names and the broader AI bottlenecks.
Live now →
dimetrics.ai