Why SpaceX acquired xAI - Sync #557
Plus: Claude Opus 4.6; GPT-5.3-Codex; Claude is ad-free; OpenAI Codex app; Gemini has over 650M MAU; Chinese robots vs snow and cold; Waymo raised $16B; and more!
Hello and welcome to Sync #557!
This week, we take a closer look at the merger of SpaceX and xAI, and why it looks more like a bailout than a step towards enabling humanity’s next phase of progress.
Elsewhere in AI, it’s been a big week for Anthropic, which released Claude Opus 4.6, committed to keeping Claude ad-free (while roasting OpenAI along the way), and shared how Claude became the first AI to plan and carry out a real drive on another planet. Oh, and became an F1 team sponsor, because why not at this point? Meanwhile, OpenAI released GPT-5.3-Codex and Codex app, announced OpenAI Frontier, and is reportedly unhappy with Nvidia’s GPUs.
Over in robotics, Waymo raised $16 billion at a $126 billion valuation, but ran into issues rolling out services in Washington, DC. Former Meta employees launched a humanoid robotics start-up, and Chinese robots showed how they fare against snow and extreme cold.
Beyond that, this week’s issue of Sync also includes Asianometry explaining why Silicon Valley views TSMC as a bottleneck, a look into the Chinese AI ecosystem, the first human trial of an epigenetic reprogramming therapy, how full brain emulation became plausible, how microbes can help mining metals, and more!
Enjoy!
Why SpaceX acquired xAI
What do rockets and AI have in common? Until recently, not that much. But on 2 February, Elon Musk merged SpaceX and xAI into a single $1.25 trillion entity—the largest corporate merger by value in American history. Musk describes the result as the “most ambitious, vertically-integrated innovation engine on (and off) Earth,” with the mission to “make a sentient sun to understand the Universe and extend the light of consciousness to the stars.” For now, however, the goal is a bit more down to Earth: deploy solar-powered AI data centres in orbit, launched by SpaceX’s rockets and running xAI’s artificial intelligence.
That is one reading of the deal. Another is that xAI, burning through billions with no clear path to profitability, desperately needed a lifeline—and SpaceX was it.
xAI’s and SpaceX’s diametrically different financial realities
xAI is burning money, and there seems to be no viable path for it to become profitable any time soon. Internal documents reported by Bloomberg show the company posted a net loss of $1.46 billion in Q3 2025 alone, and spent $7.8 billion in cash during the first nine months of the year. Revenue, while growing, was just $107 million in Q3—a rounding error compared to the spending.
The main use of Grok, xAI’s main product, is as an AI assistant on X, where it causes one controversy after another—most recently, regulatory probes after it enabled users to generate sexualised images of children. Beyond X, Grok has struggled to gain traction. It does not have the hundreds of millions of users that ChatGPT has built, nor the enterprise and developer foothold that Anthropic's Claude and Google's Gemini have established. xAI is competing against companies that are both better-funded and further ahead.
Unlike Meta, Amazon, Microsoft and Google, xAI has no cash-generating legacy business to fund its AI ambitions. It relies entirely on fundraising—having raised at least $40 billion to date, including a recent $20 billion round from investors such as Nvidia, the Qatar Investment Authority and Fidelity. Tesla has also committed $2 billion.
SpaceX, meanwhile, is in a completely different situation. It has revolutionised the space industry by driving launch costs down massively thanks to its reusable rockets. And it is a profitable company. Reuters reported that SpaceX generated roughly $8 billion in profit on $15–16 billion of revenue in 2025. Starlink, the satellite internet service with around 9 million customers, has become the largest revenue driver, while launch services continue to generate billions as well.
SpaceX is the world’s most valuable private company—and a private company only for now. On a call with investors following the merger announcement, SpaceX’s CFO confirmed plans to go public this summer, possibly on or around Musk’s birthday. If that happens, SpaceX’s IPO would become one of the largest in history, valuing the company at well over $1 trillion. It would also open new paths for raising capital to fund expensive projects such as space exploration and, conveniently, xAI’s enormous cash burn.
SolarCity déjà vu?
This situation looks similar to when SolarCity, Musk’s solar energy company, was folded into Tesla. SolarCity’s financials before the merger looked quite troubled. The company was experiencing a liquidity crisis stemming from its business model, which prioritised product development, acquiring market share, and achieving scale, all funded largely by debt. In 2016, Tesla acquired SolarCity for $2.6 billion in an all-stock transaction.
Many shareholders viewed the deal sceptically. Some saw the purchase as more of a bailout, particularly given that Musk owned over 20% of SolarCity, sat on both boards, and the company’s CEO and CTO were his relatives. This led to lawsuits, and eventually Musk had to defend the acquisition in court, where a Delaware judge ultimately ruled the price paid was fair.
The parallels are hard to ignore. Musk is once again attaching a struggling company to a more successful business, asking one set of shareholders to absorb the risks of another. And once again, he controlled both sides of the transaction—holding roughly 42% of SpaceX and a controlling stake in xAI. Like the SolarCity deal, the merger is structured as an all-stock transaction, with xAI shares converting into 0.1433 shares of SpaceX, according to an employee email viewed by the Wall Street Journal. Morgan Stanley, Musk's go-to investment bank, advised both companies—an unusual arrangement since each side in a merger typically hires its own advisers to ensure a fair deal. According to the Journal, the bank did not provide an independent fairness opinion on the valuations. But since Musk controls both companies, there was effectively no one on the other side of the table to demand one. Both boards settled on clean, round numbers—$1 trillion for SpaceX and $250 billion for xAI—despite Morgan Stanley's own estimates valuing them higher, at up to $1.26 trillion and $294 billion, respectively.
SpaceX investors signed up for rockets and satellite internet, a business generating $8 billion in annual profit. They are now also invested in an AI startup that burns through $1 billion a month, owns a controversial social media platform under regulatory scrutiny, and is competing against far better-resourced rivals. As investor Ross Gerber said, quoted by the Guardian: "If I was a SpaceX shareholder I would be pissed." Michael Sobel, president of Scenic Management, which trades in secondary stakes of private companies, put it more diplomatically in an interview with the same outlet: by folding in xAI, "you change the financial profile of the company overnight."
The $1.25 trillion bet
Whether this merger proves to be visionary or a repeat of the SolarCity playbook remains to be seen. But the SolarCity comparison understates the scale of what is happening here. That was a $2.6 billion deal. This one is worth $1.25 trillion, and the technological bet is far more radical. Musk is, after all, talking about a path toward a Kardashev II civilisation—one capable of harnessing the full power of its star. But the gap between vision and execution is vast.
The entire plan hinges on Starship, SpaceX’s next-generation rocket, which has been in flight testing since 2023 and has not yet deployed an operational payload. Without a fully operational Starship launching frequently and cheaply, orbital data centres remain a PowerPoint slide, not a product.
The merger also sets the stage for what is shaping up to be a mega year for tech IPOs. SpaceX, OpenAI and Anthropic are all expected to go public in 2026—meaning the three companies could become publicly traded within months of each other. For SpaceX, going public would unlock vast new pools of capital, but it would also subject the company—and xAI’s cash burn—to the scrutiny of public market investors for the first time. Public market investors will be less patient than private ones.
There are regulatory uncertainties, too. It is unclear whether the merger requires review by the Committee on Foreign Investment in the United States (CFIUS). SpaceX holds tens of billions of dollars in federal defence contracts and plays a critical role in US national security infrastructure. Meanwhile, xAI's most recent funding round included sovereign wealth funds from Qatar and Abu Dhabi. The combination of sensitive defence technology and foreign investors is precisely the kind of arrangement CFIUS exists to scrutinise—yet neither company has confirmed whether a review is underway.
Musk has proved doubters wrong before. He was mocked for proposing reusable rockets, and SpaceX now dominates the launch industry. But he has also oversold and underdelivered—full self-driving, the Tesla Semi timeline, and the original Starship schedule all come to mind. The question for investors, regulators, and the rest of us is which pattern this merger will follow.
If you enjoy this post, please click the ❤️ button and share it.
🦾 More than a human
FDA clears first human trial of epigenetic reprogramming therapy
The FDA has approved Life Biosciences to start the first human trial of a partial epigenetic reprogramming therapy, marking an important moment for longevity research. The Phase 1 trial will focus on testing the safety of ER-100, a gene therapy designed to rejuvenate damaged retinal cells in people with serious age-related eye diseases, using a modified set of the Yamanaka factors to lower cancer risk. Although the study is mainly about safety rather than proving benefit, it moves cellular rejuvenation from theory into real clinical testing and could pave the way for similar treatments for other age-related diseases.
🧠 Artificial Intelligence
Introducing Claude Opus 4.6
Anthropic released Claude Opus 4.6, an updated version of its flagship model, with reported improvements in coding and agentic performance, including better planning, longer autonomous task-running, more reliable work in large codebases, and stronger code review/debugging. It also introduces a 1M token context window (beta), adds agent teams in Claude Code (research preview), and is positioned as more capable for everyday work tasks like research, financial analysis, and creating documents, spreadsheets, and presentations. Opus 4.6 is available now on claude.ai, via the API, and on major cloud platforms, with pricing unchanged. To celebrate the Claude Opus 4.6 launch, Anthropic is offering a limited-time $50 in extra usage to Pro and Max users (details on how to claim the offer are here).
Anthropic has announced that Claude will remain ad-free. In what reads as a direct shot at OpenAI, Anthropic laid out the reasons why ads are incompatible with what the company wants from Claude and would undermine trust in their AI assistant, which Anthropic aims to be genuinely helpful for work and for deep thinking. Anthropic, however, does not rule out agentic commerce, where Claude helps users search and purchase products or services, but it stresses that this should happen only when the user chooses to do so. Moreover, Anthropic prepared a series of ads for the Super Bowl that are directly attacking and mocking OpenAI’s choice to introduce ads in ChatGPT. Sam Altman responded by calling Anthropic’s post “dishonest”, accusing it of “doublespeak” and claiming Anthropic shows “authoritarian” tendencies.
Introducing GPT-5.3-Codex
OpenAI launched GPT-5.3-Codex, a new version of its model focused on coding, which it describes as its most capable agentic coding model so far, with reported state-of-the-art benchmark performance, a 25% speed boost, and stronger long-running abilities. Beyond coding, OpenAI says it’s designed to support the full software lifecycle and broader professional computer work like debugging, deployment, writing product requirements, research, and producing slides or spreadsheets. GPT-5.3-Codex is available on paid ChatGPT plans everywhere Codex runs (app, CLI, IDE extension, and web), with API access planned soon.
OpenAI: Introducing the Codex app
OpenAI launches Codex app, a new desktop app that makes it easier to work with multiple AI coding agents at the same time. It lets developers run tasks in parallel, manage changes safely, use skills and automations to handle more than just code, and move smoothly between the app, CLI, IDE and cloud. The Codex app is available on macOS for ChatGPT Plus, Pro, Business, Enterprise and Edu users, with limited-time access for Free and Go users and temporarily higher rate limits on paid plans.
Introducing OpenAI Frontier
OpenAI Frontier is a new platform that helps large organisations build, deploy, and manage AI agents that can do real work across the business. It aims to close the gap between what AI models are capable of and what companies can actually roll out safely and reliably. Frontier gives agents shared business context, access to tools and systems, ways to measure and improve quality over time, and clear identities, permissions, and security controls. Early adopters include HP, Intuit, Oracle, State Farm, Thermo Fisher, and Uber, with broader availability expected in the next few months.
Indonesia ‘conditionally’ lifts ban on Grok
Indonesia has lifted its ban on xAI’s chatbot Grok, joining Malaysia and the Philippines, after X promised to take steps to prevent misuse. The countries had banned Grok because it was used to create millions of non-consensual, sexualised images, including images of real women and minors. Indonesia said the ban is only conditional and could be brought back if new violations are found. That is not going well, as Grok reportedly still generates sexualised images. Additionally, French prosecutors raided X's offices in Paris, and the UK privacy watchdog opened an inquiry into X over Grok.
▶️ Silicon Valley Thinks TSMC is Braking the AI Boom (20:32)
In this video, Asianometry explains why Silicon Valley views TSMC as a bottleneck in AI development and criticises its conservative business approach, rather than going all in on AI as US big tech companies have done. It also explains how the semiconductor supply chain works and why the blame cannot be placed solely on TSMC, which is just one part of a much larger system that, due to long lead times and other constraints, cannot be sped up or wound down on demand.
OpenAI is unsatisfied with some Nvidia chips and looking for alternatives
Reuters reports that OpenAI is looking for alternatives to some of Nvidia’s AI chips, particularly for tasks that need faster responses, because it is unhappy with their performance in areas such as coding tools. While Nvidia remains OpenAI’s main supplier, the company has explored options from AMD, Cerebras and Groq, hoping they could provide faster, inference-focused chips as AI development increasingly prioritises speed and efficiency.
The first AI-planned drive on another planet
Anthropic’s Claude became the first AI to plan and carry out a real drive on another planet by helping engineers at NASA’s Jet Propulsion Laboratory work out a 400-metre route for the Perseverance Rover on Mars. The test cut planning time by about half and gives us an early look at how AI like Claude could make future missions to the Moon, Mars, and beyond more independent and efficient.
Mozilla is building an AI ‘rebel alliance’ to take on industry heavyweights OpenAI, Anthropic
Mozilla wants to challenge the growing power of big AI companies by pushing for AI that is open, transparent and trustworthy. Mozilla is promising to use $1.4 billion from its cash reserves and investment arm to support startups and nonprofits working on mission-driven AI, and do for AI what it once did for the open web. Mark Surman, Mozilla’s president, believes that a broad, open-source AI ecosystem can still become mainstream and make money.
Google’s Gemini app has surpassed 750M monthly active users
According to Alphabet’s fourth-quarter 2025 earnings, Gemini has surpassed 750 million monthly active users (MAU). That’s 100 million more users than in the previous quarter, putting Gemini ahead of Meta AI, which reportedly has 500 million MAU. OpenAI’s ChatGPT is still on top with an estimated 810 million MAU. Additionally, Alphabet revealed that its first-party models, like Gemini, now process over 10 billion tokens per minute via direct API.
Snowflake and OpenAI partner to bring frontier intelligence to enterprise data
OpenAI and Snowflake have signed a multi-year, $200 million deal to bring OpenAI’s models directly into Snowflake’s platform. This will let Snowflake’s customers build AI agents, ask questions in plain language, and get insights from their own data without moving it outside Snowflake’s secure system.
Nvidia’s Campaign to Sell AI Chips to China Finally Pays Off
Shift in US policy has allowed Chinese tech companies to buy large numbers of Nvidia’s advanced H200 AI chips, reversing earlier Biden-era rules that blocked such sales over national security concerns. Under the Trump administration, officials now argue that controlled, legal sales are better than losing the Chinese market entirely and allowing chip smuggling to continue unchecked, a view supported by Nvidia. China gains access to much-needed computing power for AI development, while still tightly controlling who can buy the chips to protect its own chipmakers. Some experts warn that mixed US policy signals may end up speeding up China’s drive to become more self-reliant in semiconductors.
The All-Star Chinese AI Conversation of 2026
ChinaTalk shares a conversation between Zhipu, Moonshot, Qwen, and Tencent founders and researchers from the AGI-Next summit in Beijing, which covers where China stands in the global AI race and the challenges it still faces. The discussion looks at limits in compute, chips, and enterprise markets, questions whether open source really closes the gap with the US, and explores future directions such as agents, multimodal models, and long-term memory. The speakers argue that China is strong at moving fast once ideas are proven, but that leading globally will require more risk-taking, deeper innovation, and stronger business-focused AI adoption.
Chinese AI Developers Say They Can’t Beat America Without Better Chips
This article says that some leading Chinese AI researchers now believe China is falling further behind the US in AI, mainly because it lacks access to the most advanced chips due to US export controls. These limits force Chinese firms into indirect and less efficient arrangements and reduce their ability to invest in cutting-edge research compared with American rivals. Although companies such as DeepSeek and Alibaba have narrowed the gap through efficiency gains and open-source models, weaker domestic chips and lower overall investment mean chip shortages remain the main obstacle.
Advancing AI benchmarking with Game Arena
Kaggle is updating its Game Arena with two new games—Werewolf and poker—to benchmark how models navigate social dynamics and calculated risk. OpenAI’s GPT-5.2 tops the leaderboard in poker, while Gemini 3 Pro leads in Werewolf, as well as in both chess categories.
Alexa+, Amazon’s AI assistant, is now available to everyone in the U.S.
Amazon has rolled out Alexa+, a new AI-powered version of Alexa, to all US customers. Prime members get full access for free, while everyone else can use a limited free version on the app or website, or pay for a subscription. Amazon says most people in testing liked it, and usage increased, even though some users complained it was too talkative at first.
Xcode 26.3 unlocks the power of agentic coding
Agentic coding comes to Xcode with built-in support for coding agents such as Anthropic’s Claude and OpenAI’s Codex, which can work more independently to help developers build apps faster. The update builds on the coding assistant introduced in Xcode 26, lets agents assist across the development workflow, and supports the open Model Context Protocol, giving developers the option to connect other compatible agents and tools.
ElevenLabs raises $500M from Sequoia at an $11 billion valuation
ElevenLabs has raised $500 million in a new funding round led by Sequoia Capital, valuing the voice AI startup at $11 billion—more than three times its value in January 2025. The company has now raised over $781 million in total and said it will use the money for research, building new products, and expanding into new markets. ElevenLabs also plans to go beyond voice by developing AI agents that can talk, type, and take action, and may add video features.
Qwen3-Coder-Next
Qwen introduces Qwen3-Coder-Next, an open-weight AI model designed for coding agents and local development. According to Qwen, the new model delivers strong performance while keeping costs low. It uses a smart training approach focused on real coding tasks and tool use, allowing it to reason over long tasks, fix its own mistakes, and perform well on benchmarks despite having far fewer active parameters than larger models.
Voxtral Transcribe 2
Mistral has released Voxtral Transcribe 2, a new speech-to-text family featuring Voxtral Mini Transcribe V2 for high-accuracy batch transcription (with diarisation, context biasing, and word-level timestamps) and Voxtral Realtime for live transcription with latency configurable down to sub-200ms. Mistral claims state-of-the-art accuracy, best-in-class efficiency, and leading price–performance, with strong multilingual support across 13 languages. Mini Transcribe V2 is available via API and in Mistral Studio’s new audio playground, while Voxtral Realtime is available via API and as open weights on Hugging Face.
▶️ Can humans make AI any better? (23:38)
Here is another excellent video from Stephen at Welch Labs that explores whether LLMs are hitting a barrier because they rely on human knowledge rather than learning independently. He highlights how DeepMind trained AI to achieve superhuman performance in Go by playing against itself, and argues that the next major breakthrough in AI may come from models that can discover knowledge on their own—much like AlphaGo and AlphaZero uncovered new, almost “alien” ways of playing games.
▶️ Inside a Chinese AI Lab: How MiniMax Builds Open Models (31:38)
Olive Song, a senior researcher at MiniMax, discusses her work at one of China’s leading AI labs in this video. The conversation covers reinforcement learning, alignment, and building open AI models, as well as how AI is transforming everyday life and productivity, what working at MiniMax is like, and what the future of AI may look like.
🤖 Robotics
Waymo raises $16 billion investment round
Waymo has raised $16 billion in new funding, valuing the company at $126 billion, with support from Alphabet and several major investors. It claims its self-driving technology is now safer than human driving, reporting a 90% drop in serious injury crashes across 127 million autonomous miles. Waymo also says it has grown quickly, reaching 15 million rides in 2025 and over 20 million total, and plans to expand to more than 20 new cities in 2026, including Tokyo and London.
Waymo Hits a Rough Patch in Washington, DC
Waymo’s plan to expand its robotaxi service has stalled in Washington, DC, where unclear local rules and a lack of political support have left it unable to launch, even though it is already testing there. With no national law for self-driving cars, Waymo has to deal with different rules in each state and city, and DC officials are waiting for a delayed safety report while worrying about safety, oversight, and effects on ride-hail jobs. Meanwhile, Waymo is still trying to enter new places like Boston, where it is asking state lawmakers to allow fully driverless vehicles.
Former Meta Employees Launch Sprout, an Adorable Humanoid Robot for Developers
Fauna, a startup founded by former Meta employees, has emerged from stealth to present its new humanoid robot platform, Sprout, designed specifically for developers, researchers, and institutions rather than industrial deployment. Sprout aims to fill a long-standing gap in robotics by offering a more accessible, flexible, and relatively affordable humanoid system that supports experimentation with AI and real-world robotics.
Waabi raises $1B to advance autonomous trucks and robotaxis
Waabi has raised $750 million in a Series C funding round—the largest ever in Canada—to speed up the launch of its autonomous trucks and move into robotaxis. Waabi says its AI platform can power both trucks and passenger vehicles using the same core technology. Additionally, the company secured further investment from Uber, which will provide milestone-based funding, and both companies plan to deploy at least 25,000 autonomous vehicles.
In this video, Boston Dynamics shows off how natural and human-like Atlas’s movements are. The results are impressive, but they didn’t come easily, and there were a few accidents along the way.
▶️ World’s First: Unitree Humanoid Robot Autonomous Walking Challenge in −47.4°C Extreme Cold (0:44)
Unitree’s G1 humanoid robot has proven that it can not only operate in extremely cold snowfields, but also take over 130,000 steps to create a giant logo of 2026 Winter Olympics in snow.
▶️ TRON 1 Goes Cold | The First Bipedal Robot Skiing in -20°C Extreme! (0:48)
Thanks to LimX and their TRON 1 bipedal robot, we can now add skiing to the list of things robots can do.
▶️ What the iRobot Saga Teaches Us About Building Robots That Last (55:33)
Colin Angle, co-founder and former CEO of iRobot, talks about the rise of consumer robots, the failed Amazon acquisition, and what might come next for physical AI. He reflects on how his thinking has changed after 30 years in the industry, and why trust and creating real value are still the toughest challenges. He also shares what excites him about the future, from emotionally aware machines to smarter, more human-centred homes. It’s an honest, wide-ranging conversation about entrepreneurship, regulation, and what it takes to build robots that last.
🧬 Biotechnology
Automating Biology
In this article, Alex Washburne describes life as a set of layered, self-replicating biological machines and argues that biology is ready for large-scale automation. Drawing on his experience across biology, AI, and engineering, he reframes “AI” as digital automation and explains how, together with physical automation like cloud labs, it could make biological research faster, cheaper, and more scalable. He also notes challenges around trust, flexibility, and rapid change, but ultimately presents automated biology as a moonshot that could unlock major scientific discoveries and advances in medicine and biomanufacturing.
Microbes could extract the metal needed for cleantech
The article explains how new biotechnology methods could help mining companies get more metals from lower-quality ore and old mines as demand for materials like nickel and copper grows. Startups are using microbes, genetic tools, and microbial by-products to improve metal extraction, building on techniques the industry has used for years. Although these approaches show promise, they face challenges in scaling up, proving reliability, and meeting the mining industry’s slow and cautious adoption process.
💡Tangents
Building Brains on a Computer
This article argues that full brain emulation—once seen as absurd due to the extreme cost and difficulty of mapping neural wiring—has become plausibly achievable within the next few decades thanks to breakthroughs like expansion microscopy, neuron “barcoding,” and AI-based tracing tools such as Google’s PATHFINDER. It explains that the main bottleneck is not computing power but acquiring enough high-quality neural wiring and activity data, and it outlines a roadmap from small-organism emulations to mice and eventually humans, with enormous but increasingly realistic budgets and timelines.
Thanks for reading. If you enjoyed this post, please click the ❤️ button and share it!
Humanity Redefined sheds light on the bleeding edge of technology and how advancements in AI, robotics, and biotech can usher in abundance, expand humanity's horizons, and redefine what it means to be human.
A big thank you to my paid subscribers, to my Patrons: whmr, Florian, dux, Eric, Preppikoma and Andrew, and to everyone who supports my work on Ko-Fi. Thank you for the support!
My DMs are open to all subscribers. Feel free to drop me a message, share feedback, or just say "hi!"