Guy waving an American flag in a tiny Tesla Cybertruck

A person waving a US flag drives a toy Tesla Cybertruck (Frederic J. Brown/Getty Images)

Tesla could sell a record number of vehicles in the US this quarter

That won’t be enough to keep full-year global sales from falling.

9/9/25 11:46AM

Tesla is widely expected to have a bad year, but it could have a very good third quarter.

A popular analyst who goes by the name Troy Teslike currently expects Tesla to sell a record 178,000 vehicles in the US this quarter, which is two-thirds of the way through, up from 156,000 a year ago. (Tesla doesn’t break out vehicle deliveries by region, so Teslike backs up those numbers using Vehicle Identification Number data.)

The surge, he says, is due in part to pulled-forward demand from subsequent quarters, as EV buyers generally try and purchase vehicles before the government’s $7,500 credit ends on September 30. Tesla is also trying to get ahead of the incentive deadline, since it’s expected to hurt both the company’s top and bottom lines, by offering steeper discounts than other EV makers. The effort appears to be working, as Tesla is running out of inventory in the US.

But while the US is Tesla’s biggest market, just one quarter of sales growth isn’t going to smooth out the declines from earlier this year, and presumably gains in Q3 will come at the expense of Q4 sales, after the incentive expires. Nor will it undo continued declining sales in Tesla’s other major markets, including China and Europe.

Globally, Teslike estimates Tesla sales will be 466,000 in the third quarter — less than a percentage point higher than Q3 2024 (though quite a bit higher than the FactSet analyst consensus of 433,000). For the full year, Teslike is estimating 1.6 million deliveries, a 9% drop in sales compared with last year, about the same as analysts.

“We’re in this weird transition period where we will lose a lot of incentives in the US,” CEO Elon Musk himself recently told investors, adding that Tesla "could have a few rough quarters” ahead.

$26B

Jon Keegan6h

Nvidia is planning on spending $26 billion to train its own AI open-weight models, according to a 2025 financial filing. Wired was first to report the information. Nvidia has released several of its own AI models, including the Nemotron reasoning model, as well as specialized ones for specific tasks.

Nvidia making its own large frontier models could allow the company to go head-to-head against some of its biggest AI customers.

Rani Molla8h

Musk blurs the boundaries of his companies even more with joint xAI-Tesla AI agent project

Tesla and SpaceX CEO Elon Musk said Wednesday that Tesla and xAI, which is part of SpaceX, would work on a joint AI agent project called “Macrohard,” also referred to as “Digital Optimus,” as part of Tesla’s $2 billion investment in xAI. The collaboration would pair Grok with what Musk described as a real-time computer-controlling AI agent running on Tesla hardware.

In his post, Musk said Grok would serve as the higher-level “System 2” reasoning layer directing “Digital Optimus,” a faster “System 1” layer that processes the last five seconds of screen video and keyboard/mouse inputs to take action. He said the system would run inexpensively on Tesla’s low-cost AI4 chip alongside more expensive Nvidia chips at xAI, and suggested it could, “in principle,” emulate the function of entire companies. “No other company can yet do this,” he said.

Macrohard or Digital Optimus is a joint xAI-Tesla project, coming as part of Tesla’s investment agreement with xAI.

Grok is the master conductor/navigator with deep understanding of the world to direct digital Optimus, which is processing and actioning the past 5 secs of…
— Elon Musk (@elonmusk) March 11, 2026

Business Insider reported earlier Wednesday that Tesla was taking up the AI agent mantle as xAI’s similar project stalled, but Musk’s post suggests the initiatives are more intertwined than previously understood.

The collaboration marks the latest example of Musk’s companies working closely together, further blurring the lines between Tesla and the recently merged SpaceX-xAI entity.

Rani Molla10h

Meta doubles down on custom inference chips after reportedly scrapping training chip

Meta said today that it’s expanding its custom silicon development to include four new generations of Meta Training and Inference Accelerator (MTIA) chips. The announcement comes just weeks after The Information reported that the social media company had scrapped its most advanced AI training chip, dubbed Olympus, after facing design challenges. In the meantime, it signed outside chip deals with Nvidiaand Advanced Micro Devices.

Early in its recent conference call, Broadcom CEO Hock Tan sought to reassure investors that the custom chip specialist’s relationship with the social media giant was only getting stronger.

“Now contrary to recent analyst reports, Meta’s custom accelerator MTIA road map is alive and well,” he said. “We’re shipping now.”

The new road map suggests Meta’s in-house chips will focus more on inference, which has more predictable workloads, over training — a technically more demanding area dominated by Nvidia:

“MTIA 300 will be used for ranking and recommendations training, and is already in production. MTIA 400, 450 and 500 will be capable of handling all workloads, but we will primarily use these chips to support GenAI inference production in the near future and into 2027.”

Meta CFO Susan Li told attendees at Morgan Stanley’s tech conference earlier this month that the company “eventually” plans to expand its custom chip design to include training models.

Expanding Meta’s Custom Silicon to Power Our AI Workloads

Early in its recent conference call, Broadcom CEO Hock Tan sought to reassure investors that the custom chip specialist’s relationship with the social media giant was only getting stronger.

“Now contrary to recent analyst reports, Meta’s custom accelerator MTIA road map is alive and well,” he said. “We’re shipping now.”

The new road map suggests Meta’s in-house chips will focus more on inference, which has more predictable workloads, over training — a technically more demanding area dominated by Nvidia:

“MTIA 300 will be used for ranking and recommendations training, and is already in production. MTIA 400, 450 and 500 will be capable of handling all workloads, but we will primarily use these chips to support GenAI inference production in the near future and into 2027.”

Meta CFO Susan Li told attendees at Morgan Stanley’s tech conference earlier this month that the company “eventually” plans to expand its custom chip design to include training models.

Rani Molla11h

Google completes acquisition of Wiz — its biggest ever

Today Google said it has completed its $32 billion acquisition of cybersecurity startup Wiz, the largest deal in the company’s history.

“This acquisition is an investment by Google Cloud to improve cloud security and enable organizations to build fast and securely across any cloud or AI platform,” the company wrote in the press release.

The companies agreed to the all-cash purchase last year, after quite a bit of back-and-forth.