Reports dated September 25 describe a sharp increase in Tesla’s Optimus humanoid-robot output at Fremont, according to Electrek and CryptoBriefing, both citing The Information. Weekly production is said to have risen from dozens of units in the second quarter to hundreds in August, about a tenfold step-up. Most of those machines, the same accounts stress, are still for internal testing and data collection rather than unsupervised factory labour.

Tesla’s manufacturing path, as relayed by those outlets, is to reuse unused Model S and Model X capacity for a more continuous Optimus line. The near-term figure cited is more than 1,000 units a week by the end of 2026, with longer-term talk of capacity on the order of 20,000 a week. Those targets are forward-looking. They have not been independently confirmed in an official Tesla filing or earnings transcript in the materials reviewed here.

Volume is not the same as usefulness. Electrek’s framing—that the robots still struggle to generalize—matches the core technical limit in the reporting. Optimus can perform trained, supervised tasks in controlled areas. Transferring skills to new environments and novel objects remains difficult. Hands are a particular bottleneck: assembly involves more than 100 components, and Tesla is still building a supply chain from a near-standing start. CryptoBriefing described those logistics as extremely challenging.

Ars Technica reported a parallel workforce issue: some Tesla employees have balked at training Optimus units that could eventually replace their own roles. That matters for the investment narrative because the robotics thesis depends on factory data and human demonstration. If training is slower or more contested than management assumes, the generalization problem does not get easier.

Early commercial plans, according to the same secondary reporting, lean toward leasing a limited number of units to selected businesses rather than broad outright sales. That would keep Tesla closer to the hardware, software, and liability surface. It would also delay a clean robotics revenue line for equity analysts.

Reported picture from those accounts:

  • August run-rate: hundreds of units a week at Fremont, about ten times the second-quarter pace
  • Year-end 2026 target: more than 1,000 units a week on a more automated line
  • Longer-term capacity talk: on the order of 20,000 units a week
  • Use today: mostly internal testing and data collection, with limited controlled factory tasks
  • Early go-to-market: leasing to select businesses, not mass sales

For Canadian investors, this is primarily a Tesla (NASDAQ: TSLA) narrative item, not a near-term cash-flow event. Optimus is a long-duration option inside a company whose valuation already embeds ambitious autonomy and robotics assumptions. A ramp that is mostly internal does not, by itself, change 2026 revenue mix. It also does not create a distinct Canadian tax or regulatory catalyst, or alter how U.S.-listed shares sit in a typical registered or non-registered account. The relevance is execution: whether Tesla can industrialize a humanoid, whether Fremont’s S/X lines are a credible bridge, and whether supply-chain and AI limits cap the story through year-end 2026.

What is reasonably established is only that multiple secondary outlets, citing The Information, describe a tenfold August ramp to hundreds of units per week, heavy internal use, and acknowledged technical and supply hurdles. What remains speculative is the year-end 1,000-plus weekly run-rate, the 20,000-week capacity figure, the leasing timetable, and any implication that Optimus is close to general-purpose factory labour. Official Tesla confirmation, or commentary from Elon Musk on the same metrics, would be the next evidence step.

A useful distinction is unit counts versus unit usefulness. Hundreds of supervised, internally deployed robots are an operational milestone. They are not proof that Optimus can be leased at scale, that hands and generalization are solved, or that a labour-replacement story is investable in the near term. The opposite reading—that Tesla is spending real factory time and supplier effort on a product still limited to controlled tasks—is equally compatible with the same sources.

Canadian portfolios that hold TSLA directly or through U.S.-listed funds will feel this mainly through sentiment and valuation multiples, plus the usual USD/CAD translation of results, not through a Canada-specific product or policy change. The September 25 reporting is a manufacturing and AI execution update, not a finished robotics business. This is not tax, legal, or investment advice.