Walter Isaacson’s Elon Musk biography is set to be published on Tuesday, and a new preview of the book illustrates details about Tesla’s development of the upcoming Full Self-Driving (FSD) version 12.
In an additional preview of his Musk biography for CNBC, Isaacson discusses the use of AI in the development of Tesla’s FSD v12, in a shift that took place within the last several months. Isaacson talks about Tesla’s recent development of the upcoming FSD v12, which he and Tesla demonstrate has moved away from a “rules-based” approach.
Notably, FSD v12 is expected to use billions of video frames from real-world driving incidents to train its neural network rather than using thousands of lines of code like previous versions. In a conversation with Musk last December, Tesla Autopilot employee Dhaval Shroff had likened the concept to the popular chatbot ChatGPT, instead for use with driving.
“It’s like ChatGPT, but for cars,” Shroff said. “We process an enormous amount of data on how real human drivers acted in a complex driving situation, and then we train a computer’s neural network to mimic that.”
Surprisingly enough, Tesla only shifted toward this “neural network planner” approach recently. By the beginning of this year, however, the neural network had already analyzed 10 million video clips based on the best-case-scenario drivers the system had access to. Musk instructed employees at the company’s Buffalo, New York facility who were in charge of analyzing the footage to train the AI on things “a five-star Uber driver would do.”
Moving from a rules-based to a network-path-based AI approach allowed FSD to use human driving data to avoid obstacles, even if breaking some rules was necessary. Shroff helped demonstrate the idea to Musk with a demo featuring trash bins, debris, and upturned traffic cones, which the car handled surprisingly well.
“Here’s what happens when we move from rules-based to network-path-based,” Shroff explained. “The car will never get into a collision if you turn this thing on, even in unstructured environments.”
Musk quickly took to the idea, as can be seen in a recent livestream of Tesla’s FSD v12 software in Palo Alto with Autopilot software director Ashok Elluswamy. He has repeatedly spoken about the upcoming software version’s impressive driving results, despite one small moment in the drive where the car almost ran a red light.
In any case, Musk could argue that the red-light moment is a good case for the need for self-driving software to continually learn. Given that it will constantly be trained from the video data generated by camera footage from real-world drivers, it should theoretically make it safer over time, according to Musk.
During development, Musk also reportedly latched onto the fact that it took over a million video clips for the neural network to begin performing well, though he looks forward to what significantly more data will do for FSD.
Still, critics and regulators have expressed concerns about the faults of human drivers training AI-based driving systems, and Tesla has repeatedly been questioned by the National Highway Traffic Safety Administration (NHTSA) about its Autopilot and FSD beta systems.
According to Isaacson, Tesla plans to release FSD v12 as soon as regulators approve it. Meanwhile, an ongoing study by the National Highway Safety Board is looking to determine if self-driving cars should be permitted to imitate human driving actions that blur traffic rules, such as creeping up at stop signs.
Musk said in April that he expects Tesla to reach full autonomy within a year, though he has also been known to share ambitious targets for the software in the past.
You can read Walter Isaacson’s full account of the development of Tesla FSD v12 here, in a CNBC preview of the upcoming Elon Musk biography.
What are your thoughts? Let me know at zach@teslarati.com, find me on X at @zacharyvisconti, or send your tips to us at tips@teslarati.com.
Tesla has finally clarified the situation regarding the viral crash in Texas where a Model 3 slammed into a home.
CEO Elon Musk replied to reports on Monday that stated the crash was due to the company’s Full Self-Driving or Autopilot suite, which seemed unlikely to those who are familiar with it. Video showed the car slamming into a house at an excessive rate of speed, making it highly unlikely the crash was due to the suite’s operation, as it does not travel at those speeds in residential areas.
Musk said:
“This makes no sense. FSD drives slowly through neighborhood streets, and this was a high-speed crash!”
Tesla’s Head of AI, Ashok Elluswamy, added context, revealing that the company’s data shows the driver “manually overrode self-driving by pressing the accelerator all the way to 100%.”
He revealed the speed reached by the car was 73 MPH, and the accelerator was still pressed “even after the crash.”
Yup. In this case, the driver manually overrode self-driving by pressing the accelerator all the way to 100% of the accel pedal in this residential area. They reached a speed of 73 mph during the crash, and had the accelerator pressed even after the crash.
— Ashok Elluswamy (@aelluswamy) June 22, 2026
Authorities are reportedly investigating “whether Tesla’s Autopilot system played a role after a Model 3 left the roadway…slammed through a brick house at high speed and fatally struck Matha Avila as she sat inside,” the New York Post reported.
The National Highway Traffic Safety Administration (NHTSA) is now investigating the crash. Tesla will work with the agency to provide them with whatever information they need in order to clarify the cause of the crash.
Similarly, Tesla had claims of a fatal accident in Harris County, Texas, a few years ago. Early reports indicated that Full Self-Driving was the cause of the crash. After the National Transportation Safety Board (NTSB) worked with Tesla, the agency proved there was “no use of the Autopilot system at any time during this ownership period of the vehicle, including the time frame up to the last transmitted timestamp on April 17, 2021.”
Tesla alleged “driverless” crash in Texas: What is known so far
“Application of the accelerator pedal was found to be as high as 98.8 percent,” the NTSB said in their findings. The highest recorded speed in the five seconds leading up to the impact was 67 miles per hour. The area where the crash occurred is residential, and Texas State laws have default speed limits of 30 MPH in residential streets.
This appears to be a similar situation. However, an investigation will prove what happened for sure.
SpaceX announced today that it commenced its first-ever public bond offering, marking a significant step in the newly public company’s capital markets strategy.
The company announced an offering of senior unsecured notes expected to raise at least $20 billion.
The move comes just a short time after SpaceX completed one of the largest initial public offerings in history. In mid-June, the company priced shares at $135 and raised more than $85 billion, propelling founder Elon Musk’s net worth past the trillion-dollar mark and giving the firm substantial liquidity.
🚨 SpaceX has announced its inaugural offering of senior unsecured notes.
The net proceeds will be used to repay outstanding loans under its bridge loan facility in full.
This inaugural debt offering represents a financing milestone for SpaceX, which previously depended… pic.twitter.com/pcOZuVbTRv
— TESLARATI (@Teslarati) June 22, 2026
According to the company’s SEC filing, the net proceeds from the notes will be used primarily to repay in full the outstanding borrowings under its existing bridge loan facility, cover related fees and expenses, and fund general corporate purposes. The offering is being conducted under Rule 144A, as well as Regulation S, targeting qualified institutional buyers and non-U.S. investors. Notes will be unsecured obligations ranking equally with other unsubordinated debt.
The $20 billion bridge loan was used to refinance approximately $17.5 billion in higher-cost “junk” debt tied to X and xAI. SpaceX had merged with xAI in February 2026 in an all-stock deal. The bridge facility, which matures in September 2027, had represented the bulk of SpaceX’s long-term debt.
SpaceX officially acquires xAI, merging rockets with AI expertise
In connection with the bond launch, SpaceX disclosed it held approximately $100.8 billion in cash and cash equivalents as of June 19. Investor calls began on the announcement date, with pricing and launch expected shortly thereafter. Rating agencies have assigned investment-grade ratings to the proposed bonds, reflecting confidence in SpaceX’s dominant position in commercial launches and the growth trajectory of its Starlink internet offering.
The debt raise also allows SpaceX to optimize its balance sheet by replacing short-term, higher-cost bridge financing with longer-date, lower-cost fixed-income securities. This provides greater financial flexibility to support capital-intensive initiatives, including the development of Starship, the expansion of the Starlink constellation, and the integration of AI capabilities following the xAI combination.
SpaceX shares (NASDAQ: SPCX) fell sharply on the news, dropping over 16 percent overall on the market on Monday. The stock had surged initially after debuting but pulled back amid profit-taking and broader market dynamics.
Overall, the bond offering underscores SpaceX’s transition to a mature public company with access to diverse funding sources. It positions the firm to pursue its long-term vision of multiplanetary expansion and AI infrastructure, while maintaining a disciplined approach to its capital structure in a high-growth but capital-heavy industry.
SpaceX confirmed today that it has officially signed its third massive compute deal, providing compute at its Colossus data center in Southaven, Tennessee.
Reflection AI will gain immediate access to NVIDIA GB300 chips at SpaceX’s Colossus 2 data center. In return, Reflection will pay SpaceX $150 million per month starting on July 1, with total payments reaching approximately $6.3 billion if the contract runs through its duration, which is until 2029. Either party can terminate the agreement with 90 days’ notice after the initial three-month period.
CNBC first reported the deal.
🚨 SpaceXAI has agreed to a new compute deal with Reflection AI.
Reflection gets access to NIVIDIA GB300s, and will pay $150M per month to SpaceXAI for the compute. pic.twitter.com/bNPare8U5u
— TESLARATI (@Teslarati) June 22, 2026
This latest partnership highlights SpaceX’s strategy of commercializing its massive Colossus supercomputing infrastructure, originally developed to power Elon Musk’s Grok AI models. The company has rapidly expanded its customer base in the AI sector following its February 2026 merger with xAI, a transaction that valued the combined entity at $1.25 trillion.
SpaceX has previously signed significant compute deals with other major players.
It granted Anthropic exclusive access to the full capacity of its Colossus 1 data center, which exceeds 300 megawatts and includes over 220,000 NVIDIA GPUs. Details from SpaceX’s IPO filings indicate Anthropic will pay $1.25 billion per month through May 2029, potentially generating around $45 billion over the term of the deal.
Additionally, Google agreed to pay SpaceX $920 million per month for compute capacity from October 2026 through June 2029. This 32-month period will provide Google access to roughly 110,000 NVIDIA GPUs, along with supporting processors and memory. Capacity ramps up through September at a reduced fee, with termination options after the first year.
SpaceXA also established arrangements for computing power with Cursor, an AI coding startup. SpaceX acquired them in a $60 billion all-stock deal.
These arrangements position SpaceX’s collective position as an AI infrastructure powerhouse with high-margin revenue potential. The Google deal alone could generate nearly $29.5 billion over its term, while the Reflection contract adds another $6.3 billion.
Combined with the Anthropic arrangement, SpaceX stands to realize tens of billions in revenue from compute leasing in the coming years, which diversifies beyond SpaceX’s traditional rocket launches and Starlink operation.
The deals underscore growing demand for advanced AI training and inference capacity amid chip shortages and surging model development needs. Reflection, valued at $25 billion and focused on “American open intelligence” with government and national security ties, cited recent restrictions on closed models as validation for open-source approaches.
For SpaceX, the partnerships transform capital-intensive data centers into flexible revenue sources while supporting its broader AI ambitions after the company has gone public.
Tesla finally clarifies fatal Texas crash, confirms driver manually overrode acceleration
SpaceX makes $20 billion move to optimize its balance sheet
