About AlphaPod
The evidence layer
for the AI trade.
AlphaPod.ai was built on a simple premise: knowing who actually wins the AI buildout requires evidence, not confidence. We track the datacenter projects, the labor filings, and the developer-adoption data that prove it — and map every number back to a public ticker and a source row.
Our Mission
Who wins the AI buildout — proven, not asserted
Demand proven by permits and grid filings. Cost proven by labor filings. Durability proven by adoption data. Three intelligence surfaces, one entity graph, one covered-ticker universe — built so every claim has a source row behind it.
The map is free because evidence-grade tracking of the AI buildout shouldn’t require an institutional data budget to see. Per-ticker exposure, labor exposure, and developer signals — joined into a single evidence dossier per ticker — are what AlphaPod is built to sell.
Our Values
What We Stand For
Provenance Over Opinion
Every figure on AlphaPod traces back to source rows and a versioned derivation — a permit, a grid filing, a WARN notice, a package-download series. No black boxes, no trust-me outputs.
Discipline Over Hype
Confirmed (T1+T2) capacity always leads; pipeline (T3/T4) is always labeled as pipeline. We don’t aggregate speculative capacity into headline dollar figures, and we don’t claim capacity-forecast accuracy we haven’t earned.
Integrity Over Growth
We will never optimize for engagement at the cost of accuracy. AlphaPod does not provide investment advice — it provides evidence-graded intelligence investors use to make their own decisions.
Institutional Rigor for Everyone
The same evidence-tier discipline, the same entity-resolution graph, the same corrections ledger — whether you’re tracking one ticker or running a fund’s AI-thesis book. The Buildout Map is free; nothing about the evidence standard is watered down for it.
Our Technology
A Deterministic Spine, Not a Chatbot
Every number displayed on AlphaPod is computed by versioned SQL and Python derivations from source data — never written by an LLM. Language models write prose only: Brief sections and dossier narratives, constrained to numbers already computed and validated before publish.
Deterministic Derivation Spine
Capacity math, evidence-tier gating, the labor EPS bridge, and the developer-commoditization triangle all run on versioned constants and zero-LLM code — reproducible and backtested against subsequently reported outcomes.
Models Are Swappable
Every LLM call is a typed task with a schema, grounding, and entity-validation gate before publish. The cheapest model that passes eval wins the slot — models are a line item, not the moat.
Entity Graph + Corrections Ledger
Project→parent→ticker, employer→ticker, and package→ticker mappings with human-review provenance. Every correction — restatement, dedupe, tier reclassification — is recorded append-only, not silently overwritten.
See the Evidence Yourself
Start with the AI Buildout Map — free, no credit card required — and see what evidence-graded intelligence looks like across 600+ covered tickers.