Tesla's Q3 profit plunges 40%, Musk doubles down on AI and robotics investment

After reading Tesla's earnings report, I suddenly have some idea why Musk is in such a hurry to merge with SpaceX.

Over the past two trading days, the two trillion-dollar companies under Musk's name have successively released their earnings reports. First, SpaceX released its second quarterly report since going public, with revenue surging 161% year-over-year, achieving single-quarter profitability for the first time, and its stock price soaring 8% after hours.

But 24 hours later, the Q3 report card Tesla delivered presented a different picture:

Operating profit was $718 million, plunging 56% year-over-year, with the operating margin sliding from 5.8% in the same period last year all the way down to 2.6%. GAAP net profit was $787 million, down 43% year-over-year, making this the ugliest set of profit figures Tesla has posted in recent years.

However, the consolidated gross margin, which can reflect its profitability, reached 19.6%, with overall gross profit reaching $5.494 billion, up 9% year-over-year. The automotive business gross margin also returned to 18.5%, somewhat higher than market expectations.

So the reason for the profit decline cannot simply be attributed to "Tesla not making money selling cars." On the contrary, it is precisely "selling cars" that has kept up its current respectable appearance.

01. Cars make money, spent elsewhere

How did Tesla deliver such an extremely contradictory Q3 report card?

First, looking at the fundamentals: Tesla achieved revenue of $28.046 billion in the third quarter, essentially flat compared with $28.09 billion in the same period last year, and slightly below the market expectation of $28.13 billion.

Revenue fluctuated little; the problem lies in profit.

GAAP net profit was $787 million, plunging 43% year-over-year; adjusted net profit was $1.697 billion, down 4% year-over-year; adjusted earnings per share were $0.48, above the market expectation of $0.44. Operating profit was $718 million, shrinking by more than half year-over-year, with the operating margin sliding from 5.8% in the same period last year all the way down to 2.6%.

With profits plunging, many people's first reaction is that Tesla can't sell cars anymore, but that really isn't the case.

In the third quarter, the automotive segment, still the core pillar, posted revenue of $20.56 billion. Although this was down 3% year-over-year, it still delivered 486,500 vehicles globally, marking two consecutive quarters of deliveries exceeding 480,000. The automotive business gross margin even recovered further to 18.5%.

At the same time, vehicle inventory supply days fell from 15 days in the previous quarter to 11 days, far below the 27 days in the first quarter. In addition, the earnings report showed that the backlog of orders for passenger vehicles reached its highest level in four years.

This is enough to show that Tesla's cars were not hard to sell in the third quarter. Even Musk said on the earnings call that the company's current vehicle demand has exceeded production capacity and that it is working to increase output at its factories.

The problem lies in regulatory credits. In the past, such credits provided Tesla with a high-margin supplement, but this dividend is now rapidly diminishing. In Q3 of this year, Tesla's regulatory credit revenue from automobiles was only $113 million, plunging 73% from $417 million in the same period last year, offsetting part of the gains brought by improved production efficiency.

For this reason, after excluding regulatory credit revenue, the automotive business gross margin still stood at 18.1%, up 1.8 percentage points from 16.3% in the second quarter and also higher than 15.4% in the same period last year.

In other words, the money made from selling cars itself has not decreased; it has even become more profitable.

So why, when demand and gross margin are clearly growing, did operating profit and net profit instead fall sharply? The answer lies on the expense side.

In the third quarter, Tesla's capital expenditure surged 110% to $4.732 billion (about 31.590 billion yuan), which directly caused free cash flow to shrink sharply to $1.531 billion (about 10.221 billion yuan), a 62% plunge year-over-year.

In the third quarter, Tesla's operating expenses reached $4.776 billion, up 39% year-over-year. R&D expenses alone rose 69% to $2.76 billion.

Musk's "$1 trillion compensation plan," which everyone was watching last year, has also begun to enter the financial statements. In the third quarter, company-wide stock-based compensation reached $1.165 billion, up 76% year-over-year, which is also an important reason why selling, general and administrative expenses grew 29%.

Tesla did not shy away from where this money went in its earnings report: almost all of it was poured into AI, the Optimus humanoid robot and the Cybercab robotaxi.

Tesla said that increased investment in AI, superintelligence and other R&D projects, higher stock-based compensation expenses, and growth in selling and administrative expenses were important reasons for the decline in operating profit. At the same time, lower regulatory credit revenue and a decline in gross profit from the energy storage business also had a negative impact.

What can be seen is that cars are still the main business, but Tesla is already preparing for the next path, which also explains the seemingly contradictory performance in Q3.

And it is worth noting that Tesla has proactively lowered its full-year capital expenditure guidance from $25 billion to $20 billion, while also securing a $30 billion borrowing facility to prepare for future expansion.

Scaling back the pace of investment on one hand while stocking up ammunition in advance on the other also shows that the money-guzzling beasts of Robotaxi, Optimus and the chip plant are about to enter a protracted battle.

02. Musk keeps quiet on "merger"

Perhaps because the earnings report itself was lackluster, or perhaps to soothe shareholders' doubts about spending heavily without seeing results, Musk rarely gave a clearer commercialization timetable for each business line on this earnings call.

First, FSD is starting to become a business. In management's view, the most important change in the third quarter was not how many more vehicles were delivered, but that more and more vehicles were already tied to FSD subscriptions at the time of delivery.

As of the third quarter, Tesla FSD had 1.74 million active paying users worldwide, half of whom were subscribers, and in North America more than 75% of new vehicle deliveries included paid FSD, up nearly 20 percentage points from the second quarter.

This means Tesla is trying to extend automotive profit from a one-time sale into sustained software revenue.

On the conference call, CFO Taneja specifically explained that each new vehicle comes with a one-month free FSD trial period, and that this proportion refers to customers still choosing to pay for a subscription despite having the free trial, and also hinted that more metrics on FSD revenue, renewal rates and churn rates would be disclosed in the future.

It is worth mentioning that an analyst estimated on the call that FSD annual recurring revenue was already close to $1 billion, and Taneja acknowledged that estimate.

But the regulatory issue remains the biggest constraint. Musk said FSD has now been approved in nine EU countries. He expects EU-wide approval could come within a few months, and said he would be surprised if it came later than the first quarter of 2027.

As for the Chinese market, there was also new discussion on the call. Tesla believes its technology still has an advantage and is actively pushing for broader FSD approval in the Chinese market.

Compared with FSD's growth, Musk described Robotaxi's progress as "like a rocket entering orbit." By the end of the third quarter, Tesla Robotaxi's cumulative unsupervised driving mileage was close to 2 million miles, 8 times the level at the end of the second quarter.

Musk said the Cybercab fleet in Austin has grown about 10 times in size since launching a month ago, and Tesla plans to launch Cybercab in Dallas in November and expand it to other cities before the end of the year.

Not only is the operating area expanding, operating hours are also set to increase, with 24-hour all-weather operation and expansion to highways before the end of the year, while ordinary owners' vehicles are expected to be able to join the Robotaxi fleet around the second quarter of 2027, provided conditions such as geofencing, vehicle management and regulatory approval are met.

After a round of praise in the middle, he still did not forget to add a blunt truth at the end. Musk said Robotaxi's actual contribution to the financial statements will not begin to show until after the second quarter of 2027, and that it will not truly become an important source of profit until the third and fourth quarters.

In other words, investors will have to wait another three or four quarters before they can see this story delivered in the numbers.

And the robot, on which even greater hopes are placed, still needs a longer investment cycle. Although in the third quarter Optimus had already begun early production on a high-volume production line at the Fremont factory, and even the first batch of robots meeting mass-production quality standards had rolled off the line.

But reaching the goal of 1 million units a year, or 20,000 units a week, will take "at least a year, perhaps longer," Musk admitted, saying the hand remains the main hardware difficulty. Musk said the robot's fingers must be soft enough to pick up fragile objects while also withstanding repeated motions and long-term wear, while protecting internal sensors, and these requirements increase design difficulty.

AI capabilities and reliability also need to improve in tandem. According to Musk's expectations, Optimus is scheduled to integrate the new-generation AI5 chip around April 2027 to further enhance the robot's computing power.

He also said a demo video of Optimus would be released in about a month, and specifically stressed that he "doesn't want the movements to be pre-programmed." That statement is noticeably more restrained than his earlier boast of "mass production of a million units next year."

Moreover, the delivery timetable has also been pushed by Musk to around the end of 2027 before deliveries to external customers can begin.

The chip manufacturing plan supporting the automotive and robotics businesses is also proceeding in parallel. Musk expects equipment commissioning to begin in the first half of 2027, with an attempt to manufacture the first chip in May or June, while the real mass-production target points to the end of 2028.

The most unusual moment of the day was that, at SpaceX's earnings call the previous day, Musk had also said that a merger between SpaceX and Tesla was already under discussion and that a special committee had been set up.

Yet a day later, on Tesla's earnings call, he would not say a word. "Regarding a potential merger with SpaceX, I think I said everything I needed to say yesterday (Tuesday)," Musk said.

He also left a meaningful line—tap dancing in a minefield, that's it. Then he did not elaborate.

This article is from the WeChat public account "SuperEV-Lab" (ID: SuperEV-Lab), author: Wang Lei, published by 36Kr with authorization.