A useful discipline: measure how much space a topic gets in the filing versus the keynote. Run that test on humanoids using Tesla's FY2025 10-K, surfaced through EdgarBeast and filed at sec.gov. Optimus appears there as a place the company applies its self-driving AI learnings and custom inference chips — a downstream application, described compactly.
What it is not, in that filing, is a separately reported operating segment with its own revenue, margin, and unit disclosures. That absence is not a scandal; it is a status report. Segment reporting follows materiality and the way management runs the business. A topic that gets a sentence as an AI application, rather than a segment table, is being told to you as promising-but-not-yet-financially-material.
Hold that next to the keynote, where the same robot can carry the company's entire future-value narrative. The gap between 'central to the pitch' and 'one application clause in the 10-K' is the single most useful hype filter available, and it requires no opinion — only reading where the filing spends its words.
This is not unique to Tesla. The pattern repeats across the sector: a capability dominates the marketing and occupies a clause in the filing. The filing is the constrained document — written under liability, allocating space by materiality — so its proportions are the more honest map of what is shipping versus what is being sold as a story.
None of this says humanoids will not matter; the sec.gov framing of Optimus as an AI-stack application is genuinely informative about the strategy. It says: size the excitement to the disclosure. When the keynote gives a topic an hour and the 10-K gives it a sentence, believe the sentence about today and reserve judgment on the hour. Reality check sourced via EdgarBeast.