Tesla just pushed Optimus production back. The factory already changed jobs

For fourteen years, Fremont built the Model S. Since early May 2026, it builds nothing at all — the line that produced more than 610,000 flagship Teslas was torn down in 46 days to make room for Optimus, the company's humanoid robot. In April, Elon Musk told investors production would start "late July or August." On July 22, Tesla's own shareholder update erased that date: the lines are still "being installed," and output is "anticipated later this year." The schedule moved; the industrial commitment did not.

The fact that can't be walked back

Announcements are cheap in robotics; factories are not. Tesla ended two profitable product lines — the Model S after 14 years, the Model X after 11 — and physically dismantled their assembly line. That is an irreversible industrial commitment, not a keynote slide. Whatever happens next, 2026 is the year a major carmaker bet real capacity on a humanoid robot.

The scale of the ambition is equally concrete: Musk's stated goal is a production run rate of one million robots per year at Fremont, with a second line at Giga Texas targeted for around summer 2027.

The date that has already moved three times

Start of production is a different matter. "Late July or August" was said on the April 22 earnings call — an intention, dated and public. Three months later, Tesla's own shareholder update replaced it with "later this year." Tesla corrected Tesla; no commentator was needed.

The same document says where the first units go: the Optimus Academy, an internal data-collection and training program. The first robots off the Fremont line will be neither sold nor placed on a production station — they will learn. That pushes the first proof of economic usefulness further out. In January, Musk himself admitted that zero Optimus robots were doing "useful work."

He also volunteered the reason to be cautious: Optimus has roughly 10,000 unique parts, and the line "will move as fast as the least lucky, slowest, dumbest part in the entire 10,000." Initial output, in his own words, will be "quite slow." Analysts following the program expect low thousands of units in 2026 at best — directed to Tesla's own operations for training, not to customers.

The context: Tesla is not first, it's betting biggest

While Tesla converts Fremont, Figure has passed 1,000 units of its Figure 03 at BotQ and runs logistics pilots with BMW. Unitree ships more humanoids than any Western maker at a fraction of the price. Chinese manufacturer AgiBot claims 15,000 cumulative units. Tesla enters series production later than its rivals — but no competitor has committed automotive-scale capacity. That is the actual differentiator: not the robot's demo reel, the factory behind it.

Three numbers that will decide the bet

Ignore the hype cycle and watch for three figures Tesla has never published:

1. Units produced at Fremont — a dated count, not a video. It is the only milestone that turns "anticipated output" into an established fact.
2. Hours of useful work per robot — Musk's January admission set the baseline at zero. Any verified internal deployment number changes the story.
3. Cost per unit — the "$20,000–30,000" figure often quoted is a target at mass scale, not a price. Nothing can be bought today, there is no order page and no waiting list, and any paid "Optimus pre-order" you find online is not from Tesla.

Public sale is claimed for 2027. Optimus milestones have now slipped three times; the honest reading of 2026 is not "deliveries next year" but "the money is committed, the date is not." For a program that spent years as a demo, tearing down a 610,000-car line remains the biggest verifiable step so far — and a few months' slip on 10,000 unique parts was the likeliest outcome, not the surprise.