Twelve frontier-lab markets plus nine audience-level unlocks. Compounding, not additive. If a competitor Tier-1 lab locks Teleox first, the defender loses optionality across all twelve simultaneously, and needs 24–36 months at $200–400M to rebuild.
These markets share the same two pillars. A lab that ships the substrate in one category pays close to zero marginal cost to ship it in the next: the same embedders, the same constellation guard, the same LoRAs. The total pool is $1T–$1.7T by 2030–2032, not the arithmetic sum of the individual lines.
If a competitor Tier-1 lab licenses Teleox first, you lose optionality on all twelve markets simultaneously and need 24–36 months at $200–400M to rebuild the substrate. The strategic-acquirer offer band — a 15% capture of a $600B–$1T labs-side pool at a 10× revenue multiple — sits at $15B–$40B. Scope and structure belong in the diligence conversation.
Every strategic acquirer will read the same stack through a different lens. The math lands differently for each — but the pool is large enough under every lens that the offer conversation is serious in all five. The full acquisition-math breakdown sits on /strategic.
15% × $800B addressable pool × 10× revenue multiple = $200B EV uplift attributable to the stack. Rational offer band is a fraction of that.
Post-settlement, licensing more text is priced as liability. Meaning extraction turns the existing corpus into 100x+ labeled training signal without new acquisition — the only path that keeps the safety-and-interpretability narrative intact.
At that capex cadence, whoever extracts the most labeled signal per training dollar wins. TCT is the multiplier on top of the compute already committed; the constellation guard is what makes the output regulator-grade at deployment.
The compression-as-intelligence literature originates inside the house. Teleox operationalises it across 13 frozen embedders and ships it through the Google Cloud channel into regulated enterprise — extending the Gemini surface into markets the current stack cannot enter.
Musk has already framed the exhaustion of human text as the binding constraint. Extracting 100x+ labeled signal from the corpus xAI already owns is the stated plan of record — Teleox is the shipped version, with the deterministic-output stack that unlocks regulated verticals Grok cannot currently address.