Elon Musk is using a brutal reset at Tesla to sketch out an even more audacious future, one that fuses custom AI chips, robotaxis, Cybertruck autonomy and vast new supercomputers. Instead of retreating after a sharp profit slump, he is doubling down on artificial intelligence, betting that software and data centers will matter more than sheet metal.
Those “insane next moves” are starting to come into focus: a multibillion‑dollar push into his xAI startup, a revived Dojo program, a new Cortex 2 supercomputer in Texas, and a plan to turn the Cybertruck into a fully autonomous platform that feeds a global robotaxi and robotics business.
Tesla’s $2 billion xAI bet and a $20 billion AI war chest
The clearest signal that Musk is reorienting Tesla around AI rather than cars is the decision by Tesla Inc to pour $2 billion into his separate startup xAI. In a filing described by Kara Carlson, the company detailed plans to acquire shares of Series E Preferred St in xAI, effectively tying shareholder capital to a venture that is still in heavy investment mode rather than profit generation, and formalizing Chief Executive Officer Elon Musk’s belief that AI is now core to Tesla’s mission rather than a side project. That commitment sits inside a broader strategy in which Tesla’s traditional model of selling automobiles has to coexist with, and eventually be overshadowed by, software and services built on top of advanced neural networks.
On January 16, 2026, Tesla followed through by entering an agreement to invest approximately $2 billion to acquire those Series E Preferred St shares, a move one analyst bluntly described as funding a “cash furnace” but that Musk frames as necessary to stay at the frontier of AI. The company is also plotting a far larger capital wave, with plans for a $20 billion spending spree on AI infrastructure, robotics and compute that would build what one report called Musk’s AI future, including discussions to potentially build a dedicated data center footprint that goes well beyond today’s in‑house clusters. In that context, the xAI stake looks less like a side bet and more like the software brain for a sprawling ecosystem of vehicles, robots and digital services.
Dojo 3, custom chips and the rise of Cortex 2
Underneath the splashy investment headlines is a quieter but equally radical hardware roadmap. Musk has made clear that Tesla will not rely indefinitely on off‑the‑shelf accelerators, and he has been teasing a pathway of in‑house AI chips in recent public remarks, including a detailed breakdown of future silicon generations in a longform conversation with Elon earlier this month. That vision is now being operationalized: Tesla is aggressively expanding its AI chip design team, with Elon Musk personally advertising roles for designers and stressing that Tesla aims to build a new chip every year, projecting massive production volumes to support autonomous driving and humanoid robots.
The most concrete sign of this shift is the revival of Dojo. After a pause, Musk posted that “Now that the AI5 chip design is in good shape, Tesla will restart work on Dojo 3,” confirming that the internal accelerator program is back on track and slotted into Tesla’s broader AI chip roadmap. A separate analysis of why Musk changed his mind on the AI supercomputer project noted that “Our AI5 chip design is almost done and AI6 is in the works,” suggesting that the company is already planning at least two generations ahead and intends to reuse much of its previous technical work rather than starting from scratch. In parallel, Tesla is building a new Supercomputer cluster called Cortex 2 at Gigafactory Texas near Austin, described as a next‑generation system that will significantly bolster AI training compute capacity and sit at the heart of its newest vehicle and robot programs.
Cybertruck, Cybercab and the pivot away from legacy models
While the chip and data center plans grab the attention of technologists, the most visible manifestation of Musk’s strategy shift is happening in the showroom. Tesla has ended production of the Model S and Model X vehicles, a move that one market watcher argued would free resources so Robots and robo‑taxis can “definitely” drive Tesla stock to $500, even as a Tesla Model X alongside a Model S at an EV dealer becomes a snapshot of a closing chapter. At the same time, Tesla profits slumped 46% last year and the company lost its crown as the top EV seller, a jarring backdrop for such aggressive bets, yet management is signaling that the answer is not more conventional sedans but a new class of autonomous products.
Instead of more traditional vehicles, Tesla is focusing its attention on its Cybercab, a vehicle designed without a steering wheel or pedals that Musk has pitched as part of an “epic future” of robotaxis. The Cybertruck is central to that narrative. Musk has already called it Tesla’s “best ever” product, telling followers that Cybertruck Is An “Incredible” Vehicle, Elon Musk Says, and describing it as “Our best ever from Tesla” compared with traditional ICE‑powered pickup trucks. In a recent update on charging infrastructure, Elon added that Tesla will be transitioning the Cybertruck line to a fully autonomous vehicle line and stated that the Cybert platform will be used for future robotaxi services such as the Robotaxi Network, effectively turning the angular pickup into a rolling testbed for the company’s most ambitious autonomy software.
Supercomputers, Megapacks and space‑based data dreams
All of this autonomy talk only works if Tesla can train and serve enormous AI models, which is why the company is racing to stand up new supercomputers. At Gigafactory Texas near Austin, Tesla said it is building a Supercomputer cluster called Cortex 2 to expand its AI training compute capacity for projects ranging from Optimus robots to its newest vehicle, the Cybercab. To power that system, Tesla Adds Megapacks for Its Next‑Generation Supercomputer, with Megapack batteries installed alongside racks of H200 GPUs at the Cortex 2 site, a pairing that lets the company buffer grid demand and keep training jobs running through power fluctuations.
Musk’s ambitions for compute do not stop at terrestrial factories. In a recent discussion of his most outlandish concepts, one deep dive into his plans noted that his wildest idea might actually work, focusing on space‑based data centers that could, in theory, leverage SpaceX launch capacity and orbital infrastructure to host AI workloads above the atmosphere. On a separate program, commentators who know Musk’s companies well, including Efron, framed these moves as part of a pattern in which he uses SpaceX, Tesla and other ventures in concert, arguing that there is “nothing Elon can’t do” when he aligns rockets, cars and data. For now, the concrete steps are on the ground in Texas, but the rhetoric around space‑linked compute hints at how far he is willing to push the supercomputer story.
X, smarter feeds and the cross‑platform AI play
Musk is also trying to ensure that his AI stack reaches hundreds of millions of people directly, not just through cars and robots. On his social platform X, Musk outlines AI roadmap for X in a post on Thursday, promising that the X AI recommendation system would be “great by mid year” and that ads would become more relevant, making the feed more engaging and personalised. That effort dovetails with the work at xAI, which is expected to supply models that can power both in‑car assistants and social feeds, giving Musk a way to test and refine algorithms at internet scale before pushing them into safety‑critical environments like autonomous driving.
More from Morning Overview