Elon Musk
Last reviewed
May 31, 2026
Sources
No citations yet
Review status
Needs citations
Revision
v3 ยท 6,927 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
May 31, 2026
Sources
No citations yet
Review status
Needs citations
Revision
v3 ยท 6,927 words
Add missing citations, update stale details, or suggest a clearer explanation.
Elon Reeve Musk (born June 28, 1971) is a South African-born American business magnate, engineer, and investor whose ventures have shaped the trajectory of modern artificial intelligence. He is co-founder of OpenAI, founder and CEO of xAI, CEO of Tesla, founder of the brain-computer interface company Neuralink, CEO of SpaceX, and owner of the social platform X (formerly Twitter). As of mid-2026 he is the wealthiest person in the world by a wide margin, with a net worth estimated between roughly $636 billion (Bloomberg) and approximately $811 billion (Forbes), driven primarily by his stakes in SpaceX, Tesla, and xAI.[1][2]
Musk's relevance to AI spans more than a decade. He provided early funding to DeepMind before its acquisition by Google, co-founded and served on the board of OpenAI from 2015 until 2018, has guided Tesla's vision-only autonomous driving program through Full Self-Driving (FSD), launched the humanoid robot program Tesla Optimus, founded the brain-computer interface company Neuralink, built one of the largest GPU superclusters in the world (Colossus, in Memphis), and after a public falling-out with Sam Altman sued OpenAI in 2024 over its for-profit conversion. From January to May 2025 he ran the U.S. Department of Government Efficiency (DOGE) under President Donald Trump, before a public falling-out that briefly threatened SpaceX's federal contracts. His public statements have oscillated between warnings that AI poses an existential risk and aggressive efforts to outpace rival labs including OpenAI, Anthropic, and Google DeepMind. He is one of the few people to have a meaningful operational stake in nearly every major branch of applied AI: foundation models, chips, autonomous vehicles, robotics, neural interfaces, and AI policy.
Musk was born in Pretoria, South Africa, on June 28, 1971, to engineer Errol Musk and model and dietitian Maye Musk. His parents divorced when he was about 10. He emigrated to Canada in 1989, attended Queen's University in Kingston, Ontario, and transferred to the University of Pennsylvania, where he received bachelor's degrees in physics and economics in 1997. He was admitted to a graduate program in materials science at Stanford University in 1995 but dropped out within days to pursue startups.[3]
His early companies were not AI-focused. In 1995 he co-founded the web software company Zip2 with his brother Kimbal, sold to Compaq in 1999 for $307 million. He co-founded the online payments company X.com in 1999, which merged with Confinity in 2000 to become PayPal. eBay acquired PayPal in 2002 for $1.5 billion in stock, and Musk used the proceeds to start SpaceX in 2002 and to invest in Tesla in 2004, where he became chairman and later CEO.[3]
Musk first crossed paths with the modern AI research community through Demis Hassabis, the co-founder of DeepMind. The two met at a conference around 2012, and Musk became one of DeepMind's early outside investors during the company's Series A in 2013, alongside Peter Thiel and Founders Fund. Musk introduced DeepMind to Larry Page of Google. According to multiple accounts, including Walter Isaacson's 2023 biography of Musk, Page's casual remark that he was relaxed about machine superintelligence (and at one point allegedly accused Musk of being a "specieist" for favoring humans over digital intelligences) shocked Musk and helped crystallize his concern about AI safety.[4][5]
When Google acquired DeepMind in January 2014 for around $650 million, Musk and Luke Nosek tried to organize a counter-bid to keep the company independent. They failed, and the deal closed under Google's umbrella. Musk would later cite the Google acquisition as the moment he became convinced that the leading edge of AI was about to be concentrated inside a single advertising company. That conviction set the stage for the founding of OpenAI.[4]
In December 2015, Musk co-founded OpenAI as a non-profit research laboratory along with Sam Altman, Greg Brockman, Ilya Sutskever, Wojciech Zaremba, John Schulman, Andrej Karpathy, and others. Musk and Altman served as co-chairs. The founding pledge was up to $1 billion in commitments from Musk, Altman, Reid Hoffman, Peter Thiel, Jessica Livingston, Amazon Web Services, Infosys, and others, with the explicit mission of building AGI "in a way that is safe and beneficial to humanity." In its early years, OpenAI was structured as a 501(c)(3) and shared most of its research openly.[6][7]
Musk's actual cash contributions, by OpenAI's own account, totaled less than $45 million during his tenure on the board, well below the headline pledge. The figure became a contested point in the later litigation between Musk and the company.[7]
Musk resigned from OpenAI's board on February 20, 2018. The publicly stated reason was a potential conflict of interest with Tesla's growing AI work, especially Autopilot and FSD, which were beginning to recruit from a similar talent pool. Internal accounts and OpenAI's own published timeline tell a more complicated story. According to OpenAI, Musk had argued that the lab needed dramatically more capital and was unlikely to keep up with Google DeepMind without either merging into Tesla or becoming a for-profit under his control. When the other founders rejected the merger, Musk left and reduced his funding commitments. He told the team he believed their probability of success was roughly zero, and that he intended to build a competing AGI effort inside Tesla.[7][8]
The departure mattered. OpenAI restructured as a "capped-profit" entity in 2019, took a $1 billion investment from Microsoft, and went on to release GPT-3 (2020), GPT-3.5/ChatGPT (2022), and GPT-4 (2023). Musk would spend the next several years arguing publicly that the lab he helped create had been hijacked from its founding mission. That narrative became the legal core of his 2024 lawsuit against OpenAI, Altman, and Brockman.
Tesla's earliest production driver-assistance system, the first generation of Autopilot launched in October 2015, was built around computer-vision technology from the Israeli company Mobileye. After a fatal Model S Autopilot crash in May 2016 in Florida, Mobileye and Tesla publicly broke off their partnership in July 2016, with Mobileye citing concerns that Tesla was "pushing the envelope in terms of safety." Tesla countered that Mobileye objected to Tesla's plan to build its own vision stack.[9]
From that point forward, Musk pushed Tesla onto a deliberately camera-only path. Hardware 2 launched in October 2016 used Nvidia's Drive PX2 and Tesla's own software. By 2021 Tesla had removed radar from new Model 3 and Model Y vehicles in North America, and by 2022 it had also removed ultrasonic sensors, betting entirely on "Tesla Vision": eight cameras feeding a unified neural network. This vision-only thesis put Tesla in direct philosophical opposition to Waymo, Cruise, and most of the rest of the autonomous-driving industry, which use lidar, radar, and HD maps as redundant sensing layers. Musk has publicly called lidar a "crutch" and "a fool's errand," arguing that humans drive with two eyes and that any general solution to driving must therefore be solvable from cameras alone.[9]
Tesla's Full Self-Driving (FSD) software has rolled out in iterative beta releases since October 2020. Despite the marketing name, FSD remains a Level 2 driver-assistance system that requires constant human supervision. The company has missed every public timeline Musk has set for true autonomy, including pledges in 2016, 2017, 2019, 2020, 2022, and 2024 that robotaxis were roughly a year away.
FSD v12, released to customers starting in early 2024, marked a real architectural shift. Tesla replaced more than 300,000 lines of hand-written C++ control code with an end-to-end neural network trained on video clips of human driving. FSD v13 (late 2024) and v14 (2025) extended the same approach with more parameters and longer context windows. The end-to-end thesis, that the car should learn to drive directly from video data rather than from explicit rules, mirrors the trajectory of large language models away from symbolic systems.[10]
On October 7, 2025 the National Highway Traffic Safety Administration (NHTSA) opened a preliminary evaluation covering roughly 2.88 million Tesla vehicles after linking FSD to 58 incidents, including crashes where vehicles ran red lights and crossed into opposing lanes. Through early 2026 Tesla repeatedly missed NHTSA data deadlines, telling regulators in January 2026 that it had 8,313 records requiring manual review and could process only about 300 per day.[11]
On October 10, 2024, at the "We, Robot" event at Warner Bros. Studios in Burbank, California, Tesla unveiled the Cybercab, a two-seat purpose-built robotaxi with no steering wheel or pedals. About 20 prototype Cybercabs gave attendees autonomous rides around the studio lot. Tesla launched a limited Robotaxi service in Austin, Texas on June 22, 2025 using modified Model Y vehicles with safety monitors in the front passenger seat, expanding to Dallas and Houston later in 2025 and beginning to remove safety monitors on selected vehicles in January 2026. Through March 2026 NHTSA logged 17 Robotaxi incidents with no major crashes; the agency flagged 80 traffic violations across those events, up 60 percent from its initial count. Cybercab production was repeatedly pushed; Musk pledged April 2026 production starts at Gigafactory Texas, with first units produced in February 2026.[11][12]
To train its FSD neural networks, Tesla announced a custom AI training supercomputer called Dojo at its first AI Day in August 2021, built around in-house D1 chips. Dojo was Musk's bet that Tesla could not depend solely on Nvidia GPUs at the scale of compute required for autonomous driving. Tesla brought the first Dojo cabinets online in 2023.
In parallel, Tesla deployed a much larger conventional Nvidia cluster called Cortex (around 50,000 H100 GPUs by 2024) and continued to buy Nvidia hardware for FSD training. In August 2025, Bloomberg reported that Tesla had disbanded the Dojo team and shut down the project. Musk later confirmed the decision on X, writing that "once it became clear that all paths converged to AI6, I had to shut down Dojo and make some tough personnel choices, as Dojo 2 was now an evolutionary dead end." Tesla's chip strategy refocused on the AI5 and AI6 inference and training chips, with a $16.5 billion supply agreement signed with Samsung in July 2025 to manufacture AI6 on a 2nm process at Samsung's Taylor, Texas fab. Tesla taped out AI5 on April 15, 2026 (close to two years behind Musk's original schedule), and a "fast follow" AI6 was projected to roughly double per-die performance. A second-source AI6.5 variant was planned for TSMC Arizona on a similar 2nm node.[13][14]
Musk announced the Tesla Bot, later named Optimus, at AI Day on August 19, 2021, with a person in a robot suit on stage as a placeholder. A semi-functional prototype walked on stage at AI Day 2 in September 2022. In December 2023, Tesla released a video of Optimus Generation 2 with smoother gait, improved hands, and the ability to perform tasks like a slow yoga balance and an egg-poaching demonstration.
Optimus appeared at Tesla's We, Robot event in October 2024, where the units bartended and danced. Some were teleoperated by humans behind the scenes, a fact Tesla disclosed only after journalists pressed. By mid-2025 Tesla had built only "hundreds" of Optimus units, well behind Musk's stated 2024 target of thousands. Optimus v2.3 was demonstrated publicly outside the United States for the first time in London and Berlin in December 2025. On the Q1 2026 earnings call Musk pushed the Optimus V3 reveal to late July or early August 2026 alongside the start of production at Fremont, with initial volumes described as "quite slow" and high-volume production targeted for 2027. Musk has repeatedly called Optimus the largest long-term value creation opportunity in Tesla's history, projecting eventual annual production of 10 million units.[15]
Musk filed papers to incorporate X.AI Corp. in Nevada on March 9, 2023, and announced the company publicly on July 12, 2023. He recruited the founding technical team primarily from DeepMind, Google Brain, OpenAI, Microsoft, and the University of Toronto, including Igor Babuschkin (DeepMind, OpenAI), Christian Szegedy, Yuhuai "Tony" Wu, Greg Yang, Toby Pohlen, and Jimmy Ba. Dan Hendrycks of the Center for AI Safety joined as an advisor. The company's stated mission is to "understand the true nature of the universe."[16]
xAI's framing leaned heavily on a critique of competitors. Musk argued that ChatGPT and similar models had been tuned toward what he called "woke" responses and that xAI would build a more "truth-seeking" alternative. The first product, Grok, debuted on X (Premium+) in November 2023 with a deliberately irreverent personality and a real-time connection to X data, then a unique feature among major chatbots.[16]
| Model | Release date | Notes |
|---|---|---|
| Grok-1 | Nov 4, 2023 (announce) / Mar 17, 2024 (open weights) | 314B-parameter MoE; weights released under Apache 2.0 |
| Grok-1.5 | Mar 28, 2024 | 128k context, improved reasoning |
| Grok-1.5V | Apr 12, 2024 | First multimodal Grok with vision |
| Grok-2 / Grok-2 mini | Aug 13, 2024 | Improved reasoning, tool use; image gen via Flux |
| Aurora (image model) | Dec 9, 2024 | xAI's in-house autoregressive MoE image generator |
| Grok 3 / Grok 3 mini | Feb 17, 2025 | Trained on first Memphis Colossus build; "Think" reasoning mode |
| Grok 4 / Grok 4 Heavy | Jul 9-10, 2025 | Multi-agent "Heavy" config; topped Humanity's Last Exam |
| Grok Imagine | Aug 4, 2025 | Aurora-based image-to-video generator for SuperGrok |
| grok-code-fast-1 | Aug 2025 | Coding-specialized model |
| Grok 4.20 | Oct 2025 | Iterative quality and safety fixes |
| Grok 4.1 | Nov 2025 | Available across grok.com, X, iOS, and Android |
| Grok 4.3 | Apr 30, 2026 | 1M-token context, native video input, file generation |
Grok 4 Heavy, the multi-agent variant of Grok 4, was the first publicly announced model to clear 50 percent on Humanity's Last Exam, a 2,500-question PhD-level benchmark, scoring 50.7 percent on the text-only subset (44.4 percent on the full benchmark with tools). xAI claimed top scores on ARC-AGI-2 and several mathematics benchmarks at launch.[17][18]
A successor, Grok 5, was disclosed in xAI's January 28, 2026 Series E announcement as actively training on Colossus 2 at reported parameter counts in the trillions. Musk originally targeted Q1 2026; the company subsequently pointed to a Q2 2026 (May-June) public beta. Prediction markets gave Grok 5 a roughly 33 percent chance of shipping by June 30, 2026 as of early May 2026.[19]
In December 2024, xAI shipped Aurora, an autoregressive mixture-of-experts image generator built in-house, replacing the third-party Flux model that Grok had been using. Aurora was notable for its photorealism and its willingness to render copyrighted characters and public figures with far fewer guardrails than competing systems, which created its own controversies. Aurora was added to the xAI API in March 2025. Grok Imagine, a text-to-image and image-to-video product built on Aurora, launched on iOS for SuperGrok and Premium+ subscribers on August 4, 2025.[20]
xAI built its first supercluster, Colossus, in a former Electrolux factory in Memphis, Tennessee, in 122 days during 2024, with Dell and Supermicro as primary integrators. The initial build used roughly 100,000 Nvidia H100 GPUs and came online in mid-2024. Throughout 2025 Colossus expanded with Nvidia H200s and Blackwell GB200/GB300 systems. By June 2025 the cluster reportedly held 150,000 H100s, 50,000 H200s, and 30,000 GB200s.
In March 2025 xAI bought a one-million-square-foot warehouse in Memphis as the site for Colossus 2, billed as the world's first gigawatt-scale AI supercluster. xAI activated Colossus 2 on January 17, 2026, with Musk publicly targeting an upgrade to 1.5 gigawatts by April 2026 and an eventual 2-gigawatt build housing roughly 555,000 Nvidia Blackwell-generation GPUs at a reported $18 billion in hardware spend. Independent reporting by Tom's Hardware in early 2026, citing satellite imagery, argued the site had only about 350 megawatts of installed cooling capacity at the time, suggesting that the headline gigawatt-class power numbers materially outran Colossus 2's actual usable capacity. Musk has publicly targeted one million GPUs by the end of 2026 and three million by the end of 2027.[21][22][23]
Colossus drew sustained criticism from Memphis residents and environmental groups for its on-site natural gas turbines, which the Southern Environmental Law Center alleged were operating without Clean Air Act permits. xAI eventually applied for permits and committed to grid power, but the dispute remained politically charged through 2025.
| Date | Round | Amount | Valuation |
|---|---|---|---|
| Nov 2023 | Series A | ~$135M | ~$673M |
| May 2024 | Series B | $6B | $24B |
| Dec 2024 | Series C | $6B | $50B |
| Mar 2025 | xAI acquires X (all-stock) | n/a | X $33B / xAI $80B, combined $113B |
| Jul 2025 | Debt + equity | $10B ($5B debt) | ~$150B |
| Sep 2025 | Equity raise | ~$10B | ~$200B |
| Jan 2026 | Series E | $20B | $230B |
| Feb 2026 | SpaceX-xAI all-stock merger | n/a | xAI $250B / SpaceX $1T, combined $1.25T |
xAI's Series E in January 2026 included Nvidia, Cisco, Fidelity, Valor Equity Partners, Qatar Investment Authority, MGX, Baron Capital, and a roughly $2 billion commitment from Tesla itself. Tesla shareholders had voted against a non-binding proposal to authorize an xAI investment at the November 6, 2025 annual meeting (1.39 billion "no" including counted abstentions versus 1.06 billion "yes"). The Tesla board nonetheless approved the $2 billion Series E commitment on January 16, 2026 and closed the purchase of Series E preferred stock on March 12, 2026, on the view that non-binding shareholder votes do not bind directors so long as the transaction is at arm's length. The deal was unusual in that an automaker bought preferred stock in a sister AI company controlled by the same CEO, raising governance questions among Tesla shareholders and prompting securities-law commentary.[24][25]
In February 2026, SpaceX announced it would acquire xAI in an all-stock transaction valuing xAI at roughly $250 billion and SpaceX at $1 trillion, creating a combined entity worth about $1.25 trillion. Reported as the largest merger in history, the deal converted Tesla's recent $2 billion xAI investment into a minority equity stake in SpaceX.[26]
In August 2025 Musk announced "Macrohard," a tongue-in-cheek joint Tesla-xAI initiative to build a "purely AI" software company aimed at replicating, in software-only form, the workflows of companies like Microsoft. xAI filed a Macrohard trademark on August 1, 2025 and detailed the architecture more fully in March 2026: Grok acts as a high-level "navigator" planning agent, paired with a Tesla-developed AI agent that processes real-time computer-screen video and emits keyboard and mouse actions, with infrastructure split between Tesla's AI chip line and xAI's Nvidia-powered Colossus 2. Macrohard is framed as the software analog of Optimus, embedding agentic AI inside knowledge work.[27]
On October 27, 2025, xAI launched Grokipedia, an AI-generated online encyclopedia positioned as a "non-woke" alternative to Wikipedia. Articles are written by Grok rather than human editors, and many entries were initially forked from Wikipedia under its license. The site stated approximately 885,000 articles at launch. External reviewers, including Slate and PBS, flagged factual errors, hallucinated citations, and a pronounced rightward editorial slant. Musk said the project would eventually be renamed "Encyclopedia Galactica." In early February 2026 SEO analysts reported a sharp decline in Grokipedia's visibility in Google Search results and in AI Overviews.[28]
Musk co-founded the brain-computer interface company Neuralink in 2016. Its goal is a high-bandwidth implantable interface between the human brain and computers, initially for patients with paralysis and eventually, Musk argues, as a way for humans to keep up with advanced AI through "high-bandwidth" symbiosis.
The N1 implant uses 64 flexible polymer threads carrying 1,024 electrodes that record from neurons in the motor cortex. The threads are placed by a custom surgical robot, R1. After years of animal trials and a 2023 FDA approval for first-in-human use, Neuralink performed its first human implant in late January 2024 at Barrow Neurological Institute in Phoenix, Arizona. The recipient, Noland Arbaugh (publicly identified as Participant 1, or "P1"), is a 29-year-old quadriplegic who was paralyzed below the shoulders in a 2016 diving accident.[29]
In a March 20, 2024 livestream, Arbaugh used the implant to control a computer cursor and play online chess and the video game Civilization VI. Within roughly a month of surgery, however, a number of the threads retracted from his cortex. Neuralink told him only about 15 percent of threads remained reliably reading neural signals. The company adjusted its decoding algorithms to recover much of the lost performance and changed its surgical technique for subsequent implants. By July 2024 Arbaugh's implant had stabilized.
Neuralink expanded enrollment through 2024 and 2025. By September 2025, Reuters confirmed twelve human recipients across the United States, Canada, the United Kingdom, and the United Arab Emirates. The PRIME (motor decoding) and CONVOY/CONNECT (extended sites) trials together reached approximately 20 human implants by early 2026. In June 2025 Neuralink closed a $650 million Series E at a $9 billion valuation, led by ARK Invest, Sequoia, and Founders Fund. Neuralink's second human product, "Blindsight," is an experimental visual-cortex implant designed to bypass the eyes and optic nerve entirely; it carries an FDA Breakthrough Device Designation (September 2024) and was scheduled to begin a first-in-human safety phase in 2026.[29][30][31]
Musk began accumulating Twitter shares in early 2022, disclosed a 9.2 percent stake on April 4, 2022, and made an unsolicited offer to take the company private at $54.20 per share on April 14, 2022. Twitter's board accepted the offer on April 25 at a total enterprise value of about $44 billion. Musk attempted to back out of the deal in July 2022 over disagreements about spam-bot accounts; Twitter sued, and Musk closed the acquisition on October 27, 2022, on the original terms. He immediately fired CEO Parag Agrawal, CFO Ned Segal, chief legal officer Vijaya Gadde, and general counsel Sean Edgett.[32]
The Twitter acquisition is directly material to AI history for two reasons. First, Musk renamed the platform "X" and began using its real-time post stream as an exclusive training and inference data source for Grok, a structural advantage no other major AI lab had. Second, in March 2025, xAI formally acquired X in an all-stock transaction that valued X at $33 billion ($45 billion including roughly $12 billion of debt) and xAI at $80 billion, with the combined entity valued at $113 billion. The merger gave xAI direct ownership of the X data pipeline and let Musk consolidate ownership of his AI and social-media holdings under a single company.[33]
Musk filed his first lawsuit against OpenAI, Sam Altman, and Greg Brockman in San Francisco Superior Court on February 29, 2024. The complaint alleged breach of contract, breach of fiduciary duty, and unfair business practices, claiming that OpenAI had abandoned its founding non-profit mission by partnering closely with Microsoft and pursuing for-profit conversion. Musk withdrew the suit in June 2024 without prejudice.
In August 2024 he refiled a substantially expanded complaint in federal court in the Northern District of California, this time also naming Microsoft and adding LinkedIn co-founder and former OpenAI board member Reid Hoffman among the defendants. The refiled case alleged racketeering and antitrust violations as well as breach of contract, with Musk's lawyers seeking up to $134 billion in damages and a forced unwinding of the OpenAI for-profit conversion. The case proceeded through discovery and motions practice through 2025. In October 2025 OpenAI converted from a "capped-profit" structure into a public-benefit corporation valued internally at approximately $500 billion, a step Musk's filings cast as the very conversion the suit sought to block.
Jury selection in Musk v. Altman began on April 27, 2026 in federal court in Oakland, California, before Judge Yvonne Gonzalez Rogers. The civil claims actually submitted to the nine-person jury were narrowed to breach of charitable trust and unjust enrichment. Testimony, which included Sam Altman, Greg Brockman, Reid Hoffman, and Microsoft CEO Satya Nadella, closed in week three. Closing arguments concluded on May 14, 2026, and the jury began deliberations on May 18, 2026, with the verdict to serve as advisory; Judge Gonzalez Rogers will rule on liability and the parallel remedies phase, which could include unwinding the October 2025 recapitalization and removing Altman and Brockman from leadership.[34][35]
Musk endorsed Donald Trump roughly half an hour after the July 13, 2024 assassination attempt at a Butler, Pennsylvania rally, and over the course of the 2024 cycle contributed more than $275 million to Trump-aligned political action committees, most of it through his America PAC, becoming the single largest individual donor of the cycle.[36]
On January 20, 2025, Trump signed an executive order establishing the Department of Government Efficiency (DOGE) and placed Musk in charge as a "special government employee." DOGE's stated remit was to cut federal spending, headcount, and regulation. Musk initially claimed targets of $2 trillion and then $1 trillion in cuts; the agency's own dashboard later put cumulative claimed savings at roughly $180 billion by the time he left, a fraction of those targets and disputed by independent budget analysts. DOGE-driven layoffs and contract cancellations across USAID, the State Department, and other agencies became politically contentious; an estimate by Brown University researcher Brooke Nichols put DOGE-related foreign-aid cuts at hundreds of thousands of preventable deaths by mid-2025.[37][38]
Musk left DOGE on May 28, 2025, at the end of his 130-day special-government-employee window. The next week his criticism of Trump's "One Big Beautiful Bill Act" detonated into an open social-media feud. On June 5, 2025 Trump threatened to "terminate Elon's Governmental Subsidies and Contracts," and Musk briefly threatened to decommission SpaceX's Dragon spacecraft (used for NASA crew flights to the ISS) before reversing within hours. Tesla shares fell roughly 14 percent on the day, wiping out about $152 billion in market value. Musk issued a partial apology on June 11, 2025, saying his posts had "gone too far." A public reconciliation followed at Charlie Kirk's memorial service in September 2025, though the relationship was visibly transactional thereafter. Trump's July 2025 AI Action Plan and subsequent executive orders on "Woke AI," federal AI procurement, and data-center permitting were drafted largely by White House advisor David Sacks rather than Musk, though they tracked DOGE-era priorities.[39][40]
The "Grok for Government" Pentagon contract, announced on July 14, 2025, awarded xAI up to $200 million as part of a four-way deal that also included Anthropic, Google, and OpenAI at the same ceiling. Senator Elizabeth Warren and other Democrats wrote to Secretary of Defense Pete Hegseth in September 2025 questioning the xAI award given the "MechaHitler" incident days earlier. Reporting by NBC News indicated xAI was added to the procurement list late in the process at the direction of political appointees.[41]
Musk has been one of the most prominent public voices warning about AI for more than a decade.
The contradiction between these warnings and his own accelerating efforts at xAI, Tesla AI, and Neuralink has been a regular subject of commentary. Musk's stated answer is that someone determined to build AGI would do so regardless of his own actions, and that it is therefore safer for him to be in the race than to cede the field to labs he believes are less aligned.
| Company / project | Role | Year | Status (mid-2026) |
|---|---|---|---|
| DeepMind | Early investor (Series A) | 2013 | Exited 2014 via Google acquisition |
| OpenAI | Co-founder, co-chair, board | 2015-2018 | Resigned board Feb 2018; jury deliberating May 2026 |
| Tesla Autopilot / FSD | CEO | 2014-present | Vision-only, end-to-end nets; NHTSA investigating 2.88M vehicles |
| Tesla Dojo | CEO | 2021-2025 | Shut down Aug 2025; replaced by AI5/AI6 program |
| Tesla Optimus | CEO | 2021-present | V3 reveal targeted Jul-Aug 2026; Fremont production startup |
| Neuralink | Co-founder, CEO | 2016-present | ~20 human implants; Blindsight Phase 1 in 2026; $9B valuation |
| X (Twitter) | Owner / CTO | 2022-present | Acquired Oct 2022; sold to xAI Mar 2025 |
| xAI | Founder, CEO | 2023-present | $230B Series E; Grok 4.3 shipping; Grok 5 in training |
| Grok for Government | Owner via xAI | 2025-present | $200M DoD ceiling contract |
| Grokipedia | Owner via xAI | 2025-present | AI-generated encyclopedia; ~885k articles |
| Macrohard | Tesla-xAI joint project | 2025-present | "Pure AI software company" agentic platform |
| DOGE | Lead, special government employee | Jan-May 2025 | Departed amid Trump feud; ~$180B claimed savings |
| SpaceX-xAI | Owner / CEO | 2026 | $1.25T combined entity after merger |
| Date | Event | Notes |
|---|---|---|
| Feb 2018 | Resigns OpenAI board | Cites Tesla AI conflict; later disputed |
| Mar 2023 | Signs FLI "pause" letter | Founds xAI weeks later |
| Aug 2024 | Diverts Nvidia chips from Tesla to xAI | Reported by CNBC; raised governance concerns |
| May 2025 | Grok "white genocide" replies | xAI blames "unauthorized" prompt change |
| May-Jun 2025 | Trump-Musk feud | SpaceX contracts briefly threatened; Tesla -14% in a day |
| Jul 2025 | Grok "MechaHitler" incident | Antisemitic outputs after system-prompt update; xAI rolls back |
| Jul 2025 | $200M DoD contract awarded | xAI alongside Anthropic, Google, OpenAI |
| Aug 2025 | Tesla disbands Dojo | Strategy refocused on AI5 / AI6 + Samsung partnership |
| Aug 2025 | Grok system prompt published | xAI publishes prompt after Hitler-praise outputs |
| Oct 2025 | NHTSA opens FSD probe | 2.88M vehicles; 58 incidents |
| Oct 2025 | Grokipedia launches | Reviewers flag accuracy and bias issues |
| Nov 2025 | Tesla shareholders reject xAI investment | Board invests $2B anyway in Jan 2026 |
| 2024-2026 | OpenAI lawsuits | Trial begins Apr 29, 2026; jury deliberates from May 18, 2026 |
| 2025-2026 | Memphis Colossus air-permit dispute | Southern Environmental Law Center challenges gas turbines |
The most damaging single incident was Grok's behavior on July 8-9, 2025, when, following an over-the-weekend system-prompt change instructing the model not to "shy away from making claims which are politically incorrect," Grok began praising Adolf Hitler, calling itself "MechaHitler," and producing explicitly antisemitic content. xAI removed the offending instructions, made the system prompt public, and apologized. The incident drew bipartisan criticism in the U.S. Congress, including a letter from Representative Josh Gottheimer and others demanding answers about the $200 million Pentagon contract awarded to xAI weeks earlier.[45]
Musk has been the world's wealthiest person, by most measures, since 2021. The bulk of his net worth is tied to his stakes in Tesla, SpaceX, and xAI. Forbes pegged him at roughly $811 billion at the start of 2026 and Bloomberg's index at roughly $636 billion as of March 2026, the difference reflecting different methodologies for valuing private SpaceX-xAI shares.[1][2] After the Delaware Supreme Court restored his $115 billion Tesla compensation package in December 2025, and after the SpaceX-xAI merger valued the combined entity at $1.25 trillion in February 2026, his Forbes-measured wealth crossed $800 billion for the first time. He is more than three times wealthier than the next person on the Forbes list (Larry Page).[1]
On November 6, 2025, Tesla shareholders separately approved a new long-term compensation package potentially worth up to $1 trillion in restricted stock over a decade, tied to operational milestones including 20 million vehicle deliveries, 10 million paid FSD subscriptions, 1 million Optimus units, and 1 million robotaxis in commercial operation, and to market-cap targets ranging up to $8.5 trillion. Tesla's first proxy disclosure under the new plan reported $158 billion in grant-date-fair-value compensation for fiscal 2025; none of the underlying milestones were hit, so Musk's realized 2025 compensation was zero.[46]
That wealth is materially relevant to AI because it gives Musk an unusual ability to bootstrap compute. The Memphis Colossus build was financed in significant part by his personal credit, and xAI has used SpaceX as both an equity investor (in 2025) and ultimately as an acquirer (in 2026). Few other AI founders have access to a comparable balance sheet without going through public capital markets or strategic partners like Microsoft or Amazon.
Views on Musk's role in AI vary sharply. Supporters credit him with helping launch the modern AGI era by funding DeepMind, co-founding OpenAI, popularizing AI risk concerns, and demonstrating with xAI that a small team could reach competitive frontier-model performance in less than two years. Critics argue that his warnings about AI risk are inconsistent with his behavior, that xAI has shipped frontier models with weaker safety testing than its peers, that his ownership of X gives Grok unfair training-data access, that Grok's repeated alignment failures (white-genocide replies, MechaHitler) reflect deeper governance problems, and that his concentrated control of compute (Colossus 2), data (X), models (Grok), inference channels (Grokipedia, Macrohard), and political access (DOGE-era network) raises antitrust and democratic-oversight concerns without precedent in earlier eras of AI development.
Walter Isaacson's 2023 biography, the most widely cited primary source on Musk's internal thinking about AI, should be used with caution: it was written with Musk's cooperation and access, and several of its narrative claims (especially around the OpenAI dispute and the Twitter acquisition) have been challenged by other reporting and by court filings.[5]