Honest fence (read first): This Optimus Bot Monday flagship is built on Elon Musk’s own posts on X. On October 1, 2026 he wrote that Tesla “cut our RAM in half for the Tesla AI5 chip (now 72GB of LP5) and 1/3 for AI6 (now 144GB of LP6),” that “this was the only way to get enough volume for Optimus production and greatly reduces cost,” and that Tesla expects “a negligible effect on Optimus performance, as memory bandwidth is a bigger limiting factor than total memory storage (bandwidth was held constant).” Within about a day he revised AI5: Tesla “decided to nudge AI5 up a little to 96GB, as Tesla would otherwise be the only company using the min RAM version of LP5” (as quoted by Not a Tesla App, TeslaNorth and Wccftech). AI6 stays at 144GB. These are executive statements, not a Tesla datasheet, not an investor filing, and not a benchmark. “Negligible effect” is a forecast — no Optimus performance data was published with it. A memory decision made for volume is not proof of any Optimus unit count. Soft-re-cover ban: this is not the Sep 28 “hundreds per week” ramp story and not the earlier AI5 tape-out brief — the new fact is the memory trade. Embed: Dirty Tesla tbVAns7eem4 (Oct 2; chapters 00:36 and 05:17 cover the cut and the 96GB update).

Dirty Tesla — “Ai5, Pothole Avoidance, & Wireless Charging” (published October 2, 2026; chapters 00:36 and 05:17). Memory figures cite Musk’s X posts and outlets quoting him, not the YouTube title.

Editorial Thesis

The useful file for practical readers is not “AI5 changes everything.” It is that Tesla’s top executive just said, in public, that memory supply — not robot design — was the constraint he had to engineer around for Optimus volume. That is a rare, specific, falsifiable supply-chain claim: smaller DRAM pools per chip, bandwidth held flat, and a bet that the robot’s AI is bandwidth-bound rather than capacity-bound. Claim vs proof: Musk’s posts are the receipt for the decision; nothing yet is a receipt for the outcome.

Honest Verdict

Believe that Tesla changed the AI5 and AI6 memory plan: the CEO said so twice, with numbers, and the 96GB revision shows the plan is still moving. Treat “the only way to get enough volume” as Musk’s framing of a supply trade-off, and treat “negligible effect on Optimus performance” as an unverified expectation until Tesla shows robot results on the shipping configuration. Do not read the cut as a sign Optimus volume has arrived — it is a sign Tesla is planning for volume it has not yet shown.

What the primary actually says

Elon Musk on X — Oct 1, 2026

  • AI5 RAM cut in half, to 72GB of LP5 (LPDDR5).
  • AI6 RAM cut by a third, to 144GB of LP6 (LPDDR6).
  • “This was the only way to get enough volume for Optimus production and greatly reduces cost.”
  • Expects “a negligible effect on Optimus performance” because memory bandwidth matters more than total capacity; bandwidth was held constant.

Musk reply — Oct 2, 2026 (as quoted by Not a Tesla App / TeslaNorth / Wccftech)

  • Tesla “decided to nudge AI5 up a little to 96GB, as Tesla would otherwise be the only company using the min RAM version of LP5.”
  • Net effect (Not a Tesla App): AI5 ends about one-third below the earlier plan rather than half; AI6 unchanged at 144GB.

Context reported by Not a Tesla App (Oct 3) — secondary, attributed

  • Current AI4 computers ship with 32GB of GDDR6 (16GB per chip across two chips).
  • Micron has estimated a humanoid robot could need around 200GB of RAM — the outlet notes Tesla is now aiming below that.
  • Samsung recently began trial production of AI5 at its Taylor, Texas fab, with mass production aimed at 2027.
  • Musk has said AI5 goes to Optimus and Tesla’s training clusters before cars.

Why this is today’s Optimus Bot Monday flagship

AI Shift News has already covered the Fremont ramp reports, the hands bottleneck, and the AI5 design milestone. What is new this week is a first-party, numbered statement that Tesla re-specced its robot brain chips around DRAM supply — and then adjusted the spec again a day later. That is a real decision readers can track against Tesla’s Q3 call on October 21.

Human review risks

  • Writing 72GB as the final AI5 figure — the latest Musk statement is 96GB.
  • Presenting “negligible effect” as a measured result.
  • Implying AI5 is in mass production now (sources say trial production; mass production aimed 2027).
  • Using the cut as evidence of current Optimus unit volume.

Bottom Line

Believe that Tesla changed the AI5 and AI6 memory plan: the CEO said so twice, with numbers, and the 96GB revision shows the plan is still moving. Treat “the only way to get enough volume” as Musk’s framing of a supply trade-off, and treat “negligible effect on Optimus performance” as an unverified expectation until Tesla shows robot results on the shipping configuration. Do not read the cut as a sign Optimus volume has arrived — it is a sign Tesla is planning for volume it has not yet shown.

Sources