Tesla's most significant strategic adjustment (from Chris zheng)

Reposting a great article by Zheng Xiaokang

Elon Musk is initiating the most significant strategic adjustment since the founding of Tesla, shifting the company's focus to intelligence-first and electric-second.

I first heard about this news in early April, with the gist being that Elon is going all in on FSD, and future products will be entirely built around FSD.

This refers to that famous meeting held for the first time in August 2021, attended by Tesla's core management team. At the time, Elon's stance was that he didn’t want to develop another smart electric vehicle below the Model 3 (temporarily referred to as the Model 2, though Tesla has stated it won’t be called that) and instead wanted to focus on Robotaxi.

Tesla's VP of Vehicle Engineering, Lars Moravy, and Chief Designer Franz von Holzhausen ultimately convinced Elon with a solution of co-production: removing the steering wheel would make it a Robotaxi, while adding it would make it a Model 2. This approach wouldn’t overly strain Tesla’s perpetually stretched engineering resources while allowing the company to pursue both the leadership’s and Elon’s desired products.

So why did the February 2024 meeting revisit this issue, escalating it to an either-or decision? Because Robotaxi is a To B product, while Model 2 is a To C product—fundamentally different.

Tesla has publicly disclosed that Robotaxi will feature Tesla’s first electric drive system abandoning high performance in favor of high efficiency. The body will adopt specifications similar to Tesla’s commercial semi-truck, and the battery pack is designed to last 1 million miles (1.6 million kilometers) over its full lifecycle.

But the most critical difference lies in "having a steering wheel, with humans as the ultimate system backup" versus "removing the steering wheel, requiring the system to operate flawlessly in all scenarios."

Although Robotaxi and Model 2 share the same platform, their designs diverge significantly. By February 2024, Elon’s stance was clear: halt Model 2 and go all out on Robotaxi.

This marks another "bet-the-company" moment in Tesla’s history: Q4 2008 with the Roadster, Q1 2013 with the Model S, and Q1 2018 with the Model 3—each requiring Tesla to pause all other projects and focus entirely on one high-priority initiative.

Designers have served as delivery specialists for the Model S, and Autopilot engineers have worked as assembly line workers for the Model 3.

The difference is that during a 2019 earnings call, Elon himself reflected on the "bet-the-company" approach: the Model 3, as Tesla’s last project to stake the entire company’s future, had become history. Today’s Tesla has grown too large to focus on just one project; some are in high-risk R&D phases, while others sustain cash flow.

But in 2024, the "bet-the-company" mindset has returned. On Friday, April 12, Tesla offices worldwide received layoff notices from headquarters, requesting lists of affected employees.

Most regions realized Tesla hadn’t conducted major layoffs in a long time and were waiting to gauge headquarters’ severity.

Soon, Tesla announced the resignation of Drew Baglino, a 18-year veteran and SVP of Powertrain and Energy Engineering, followed by layoffs of key members from multiple core projects.

All Drew -1 reports shifted to reporting directly to Elon, and regional sales VPs previously reporting to Tesla’s SVP of Automotive, Tom Zhu, also began reporting to Elon. Tesla then announced a global $2,000 price cut for its vehicles, the removal of Enhanced Autopilot in North America, and a reduction in FSD pricing from $12,000 to $8,000 for purchases and from $199/month to $99/month for subscriptions.

Internally, the saying goes: "If layoffs don’t hit the arteries, they don’t count as layoffs." This echoes Elon’s remark about product simplification: "If no components have to be added back due to over-simplification, it means simplification was insufficient." In an email, Elon stated layoffs would exceed 10%, with a second wave underway.

Around the same time, an interview with former Tesla Senior AI Director Andrej Karpathy about Elon’s management resurfaced:

"In a company, people naturally tend to expand teams and grow larger, but Elon always resists scaling. I had to fight to hire people—basically beg for approval.

Another thing: big companies struggle to remove low performers. Elon is very supportive of initiatives to weed out underperformers."

On the software front, FSD v12’s progress speaks for itself, but a more significant signal than user videos is Tesla’s Q1 purchase of 90,000 H100 GPUs (Elon’s AI startup xAI bought 50,000 H100s).

At market prices, this implies Tesla spent $3 billion on FSD training compute, surpassing its total investment from 2016–2023.

Elon also announced Tesla’s cumulative compute spending would reach $100 billion by 2024—while the company’s total 2023 capex was just $8.9 billion.

This scale of investment suggests Tesla has found the Scaling Law for autonomous driving models—a belated realization, as Elon noted in last year’s earnings call: 1 million video clips are barely enough; 2 million, slightly better; 3 million, and it’s ‘wow’; 10 million, and it’s unbelievable. This is a clear articulation of Scaling Law.

It also fundamentally reshapes Tesla’s business. A $100 billion investment means Tesla will have ~350,000 H100-equivalent compute by year-end, rivaling Microsoft and Meta as the world’s largest compute operators.

On regulation, Tesla is actively engaging with California’s DMV to secure permission—like Waymo—to operate Robotaxis in limited areas this year.

The first half was electrification; the second half is intelligence. The electrification narrative still has legs, but it’s time to take intelligence seriously.