Walter Isaacson's biography records that Elon Musk decided to shut down the Sacramento data center to cut costs after taking over Twitter
Walter Isaacson's biography records that Elon Musk decided to shut down the Sacramento data center to cut costs after taking over Twitter, concentrating the equipment in Portland. Employees said they needed to hire people to lift the floor, dismantle cables, and seismic rods, as the original services were still running. He believed it would take a week to move the computers and another week to plug them in, totaling two weeks.
During the process, he complained about the situation. His cousin, who was traveling with him, asked why he didn't go see it now. The plane was diverted to Sacramento. On-site, he said the cabinets didn't look difficult to move. A security knife was borrowed to pry open the floor vent, and he crawled under to open the electrical cabinet and unplug the cables. There was no explosion. He said it didn't seem that difficult.
The next day was Christmas Eve. He called people from Tesla, SpaceX, and Boring Company, and Steve Davis brought in semi-trucks and moving vans. By the weekend, using all the trucks available in Sacramento, they moved over 700 cabinets in three days, with the rest handled by the infrastructure team in January. The data center had previously tried to stop them, but the equipment was already loaded onto the trucks.
Many programs were still written to retrieve data from Sacramento. The correct order was to clarify connections first, switch to other data centers, confirm they could run, and then dismantle. He dismantled directly. In March 2023, he admitted that the entire shutdown of Sacramento was a mistake. He was told the data center had redundancy but was not informed that there were 70,000 hard-coded references to Sacramento in the code, causing some things to remain broken.
In May, Ron DeSantis announced his campaign on Twitter Spaces, which was affected by delays and crashes. The biography connects this incident partially to the data center shutdown. Isaacson wrote that this was an example of recklessness, impatience with dissent, and pressure. He also noted that Musk's record of ignoring dissenters was good but not perfect.
This was cost-cutting ahead of system migration. The buyer was the new owner looking to save on data center bills, while the seller was the old data center still being referenced by the code. The event was driven by the debt pressure after the acquisition. The equipment arrived in Portland in three days, but the calls did not transfer over. The benefit was the immediate cessation of site costs, while the pressure was from having to return during the holidays to fix references and the public events colliding with platform failures.
Source: Public information
ABAB AI Insight
After the acquisition was completed in October 2022, Twitter was burdened with acquisition debt, and the Sacramento data center was chosen as a cost-cutting measure. Employees requested lifting floors, dismantling cables, seismic rods, and professional movers. Musk pried open the floor with a security knife and called in people from Tesla, SpaceX, and Boring Company on Christmas Eve, moving over 700 cabinets in three days. In March 2023, he admitted the shutdown was a mistake due to 70,000 hard-coded references. In May, DeSantis's Spaces event experienced lag, which the biography linked to this relocation.
The capital path was to stop site bills first and then have people fix the code. Redundancy was verbally confirmed, but references were not counted. Trucks and cabinets were completed in three days, but service calls still pointed to the old address. What was saved was data center operations, but added were holiday callbacks and public failures. He later described the mistake as incomplete information, not the act of moving itself.
The analogy is to immediately merge data centers after a corporate acquisition without first mapping application dependencies. Bank system migrations typically switch traffic over months. Twitter compressed this switch into the Christmas week. The industry position is cost control after acquisition, not capacity expansion.
Structurally, this is a transfer of pricing power. The migration time was originally quoted by the infrastructure team based on dependencies but was changed on-site to three days. The mechanism was that no explosion was taken as a sign to continue. The 70,000 references in the code were not on the plugs, so while the equipment arrived, the service did not.
ABAB News · Cognitive Law
Just because the plug was pulled and there was no explosion doesn't mean the references have been fixed.
What can be moved in three days are cabinets, not hard-coded addresses.
The saved site fees will come back as failures and rework.