---
type: "post"
id: "musk-puts-first-external-optimus-deliveries-around-end-f82a5d"
slug: "musk-puts-first-external-optimus-deliveries-around-end-f82a5d"
title: "Musk puts first external Optimus deliveries around end of 2027, says hands, durability and AI still need work"
description: "Tesla now expects to begin delivering Optimus humanoid robots to external customers around the end of 2027."
retrieval_nugget: "Tesla now expects to begin delivering Optimus humanoid robots to external customers around the end of 2027. The update matters because it makes the bottleneck explicit: useful physical AI requires durable hardware, task data, and reliable behavior beyond a scripted demonstration—not merely a production line or an impressive video."
published_at: "2026-10-28"
updated_at: "2026-10-28"
record_date: "2026-10-28"
date_kind: "published_at"
topics: []
entities: ["Tesla","Optimus","Elon Musk","AI5","XPeng","Boston Dynamics","Atlas"]
related_patterns: []
novelty_label: "new_fact"
what_changed: "Musk now targets first deliveries to outside customers around late 2027 and identifies unresolved hands, durability, and AI challenges beyond the early-production milestone."
prior_refs: ["memory:2fb39e55-4300-8700-a8a1-fa78c7e7fe94","memory:46ba91bb-7a32-8446-a856-4e2125a75935","memory:3d4c463f-8b2d-85c6-ad89-730fc2c7e237"]
source_url: "https://electrek.co/2026/10/22/tesla-optimus-delivery-late-2027-hands-ai-problems"
source_urls: ["https://electrek.co/2026/10/22/tesla-optimus-delivery-late-2027-hands-ai-problems"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:musk-puts-first-external-optimus-deliveries-around-end-f82a5d"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/musk-puts-first-external-optimus-deliveries-around-end-f82a5d/","markdown":"https://newruntime.com/posts/musk-puts-first-external-optimus-deliveries-around-end-f82a5d.md","json":"https://newruntime.com/posts/musk-puts-first-external-optimus-deliveries-around-end-f82a5d.json"}
---

# Musk puts first external Optimus deliveries around end of 2027, says hands, durability and AI still need work

## Retrieval answer

Tesla now expects to begin delivering Optimus humanoid robots to external customers around the end of 2027. The update matters because it makes the bottleneck explicit: useful physical AI requires durable hardware, task data, and reliable behavior beyond a scripted demonstration—not merely a production line or an impressive video.

## Summary

Tesla now expects to begin delivering Optimus humanoid robots to external customers around the end of 2027. The update matters because it makes the bottleneck explicit: useful physical AI requires durable hardware, task data, and reliable behavior beyond a scripted demonstration—not merely a production line or an impressive video.

## What happened

According to [Electrek's report](https://electrek.co/2026/10/22/tesla-optimus-delivery-late-2027-hands-ai-problems) on Tesla's Q3 2026 earnings call, Elon Musk said Tesla expects its first external Optimus deliveries around the end of 2027. The initial target is industrial work that is dangerous, repetitive, or needs to operate around the clock. This is a stated target, not a record of customer deployments or completed commercial deliveries.

Tesla has started early Optimus production at its Fremont facility, on a high-volume line in space previously used for Model S and Model X. Musk said most of a line designed for one million robots annually is installed. But he also said reaching that rate—about 20,000 robots a week—will take at least a year, and possibly longer, because portions of the design are still changing. The reported production capacity is therefore a design objective rather than a demonstrated sustained output rate.

On the call, Tesla identified three broad workstreams: AI, electromechanical design, and reliability. Musk said the robot must not wear out after a month of use. A Tesla Optimus engineer singled out the hand and custom gears as difficult areas. Musk described the hand as the principal issue: fingers need enough compliance to handle delicate objects while surviving repeated wear, without the self-repair properties of human skin. Those requirements make dexterity a systems problem, involving mechanical design and reliability as well as control.

Tesla also described a data and training plan. Unlike Tesla's driver-assistance effort, Optimus does not have a large deployed fleet collecting task data. The company plans to build 10,000 to 20,000 robots for an "Optimus Academy" where they can practice tasks. Tesla expects to put its AI5 chip into Optimus around April 2027. Musk also promised a video within one or two months showing tasks that are not pre-programmed, arguing that scripted humanoid demonstrations do not establish performance in real settings.

The reported delivery date is later than Tesla's earlier public ambitions. Electrek notes that Musk had previously discussed internal use in 2025 and high-volume availability to other companies in 2026. The report also contrasts Tesla's claims with public progress at XPeng and Boston Dynamics, but it does not provide independently comparable task-performance, uptime, cost, or customer-value data across those systems. Its core news is Tesla's revised target and its description of remaining constraints.

## What the archive adds

This is a new fact on a known Tesla Optimus story. A recent Field Note documented Tesla's claim that it had begun producing its first commercial-grade Optimus robots at Fremont amid rising AI and robotics spending: [Tesla has begun producing its first commercial-grade Optimus robots at its Fremont factory](https://newruntime.com/posts/tesla-has-begun-producing-its-first-commercial-grade-9d0d9e/). The new reporting adds a specific external-delivery target and a more detailed account of the remaining technical constraints.

## Trend signal

No supplied New Runtime pattern directly fits this update. The evidence is primarily about the unresolved reliability and learning constraints of a humanoid robot, rather than a demonstrated shift in agent architecture, economics, protocols, or organizational practice.

## Why now

The timing is tied to Tesla's Q3 2026 earnings call and its early Fremont production ramp. Tesla is presenting an external delivery target while acknowledging that both the machine's design and its task-learning system remain under development. The promised unscripted-task video in the next one to two months is the nearest stated public checkpoint in the report.

## Why it matters

For teams building with AI, Optimus is a useful corrective to software-only intuitions about agent progress. In a physical system, model behavior is inseparable from sensor reliability, actuators, wear, safety margins, task setup, and the cost of collecting examples of work. A robot that performs a short, prepared sequence may still fail the product test if it cannot recover from variation, operate for long periods, or handle objects without damage.

Tesla's proposed Optimus Academy also highlights a familiar challenge in a different form: deploying a capable model is not enough when the organization lacks representative data from real work. The meaningful measure will be verified task completion over time in customer environments, with evidence about reliability and intervention rates—not robot count, factory capacity, or an isolated demo. A delivery date becomes meaningful only when customers can evaluate that operational evidence.

## Limits and uncertainty

This account reports statements made on Tesla's earnings call through Electrek; it does not include a Tesla product specification, customer contract, pricing, safety case, or independent performance test. The late-2027 date is an expectation, not a committed delivery record. The report says Tesla plans an Optimus Academy and an AI5 integration, but does not establish how the robots will collect data, what tasks they will train on, or whether results will transfer to customer sites. It also offers no standardized comparison with XPeng or Boston Dynamics. A promised video, even if released, would be limited evidence unless it documents task variation, duration, intervention, and repeatability.

## Skill to build now

Define the operational contract for one physical or embodied workflow before evaluating a demo. Specify the task boundary, acceptable failure modes, human intervention path, duration of operation, objects and environmental variation, and the evidence required to call the task complete. This turns "can the robot do it?" into a testable question that can distinguish scripted behavior from useful work.

## Reader takeaway

Tesla's late-2027 target is a schedule claim, not proof that general-purpose humanoid labor is ready. The more informative disclosure is the remaining work: durable hands and gears, an AI architecture that generalizes, and a way to collect enough task experience. In physical AI, reliability and evidence of completed work are the product.

## Open question

Will Tesla publish evidence that Optimus can perform varied industrial tasks repeatedly without scripted setup or frequent human intervention?

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)