Elon Musk's SpaceX has secured an option to acquire the artificial intelligence coding startup cursor for $60 billion, or alternatively pay a $10 billion fee entirely for their joint development work. I mean, the dollar amounts attached to that agreement are just staggering. Yeah, it's a huge number. We are looking at a deal that combines the world's most valuable private aerospace company with a heavily used enterprise software tool. Right. And the specific software arrangement, it operates as the linchpin in a much larger strategy. We're talking about an upcoming initial public offering, the complete absorption of the AI platform, and the deployment of server farms in low Earth orbit. Okay. So how does acquiring a desktop coding application justify a multi-trillion dollar aerospace valuation and lead to launching data centers into space? Well, the $60 billion cursor option exists entirely to bolster the financial narrative of the company's upcoming initial public offering. They are targeting a $1.75 trillion to $2 trillion valuation in this public offering. And they're actively seeking to raise up to $75 billion in new capital. Wow. To put those numbers into perspective for you, raising $75 billion would easily surpass the Saudi Aramco listing. Making it the largest ever. Exactly. The largest public offering in the history of the global capital markets. But the revenue multiple required to reach that specific target is just astonishing. I mean, if you look at the private valuation of the aerospace company just prior to these new financial maneuvers, it sat at roughly $800 billion. Right. Which is already huge. Exactly. But jumping from $800 billion to a public target of $2 trillion means they are asking public markets to price the stock at a 100 to 125 times revenue multiple. That's a huge leap. It is. For an investor to see any mathematical return on a multiple that high, the company would require roughly 40% annualized revenue growth for an entire decade just to break even on the initial pricing. And you just do not typically see those kinds of multiples applied to companies that manufacture heavy industrial hardware. No, definitely not. Because manufacturing rockets requires massive capital expenditure. You have to build factories, procure raw materials like stainless steel and titanium, employ thousands of aerospace engineers, and, you know, deal with the physical reality of building machines that routinely explode during testing. Right. Rockets blow up. Software doesn't. Exactly. Industrial hardware companies traditionally trade at a fraction of that multiple, which is exactly why the financial foundation of the company has fundamentally shifted toward connectivity. The Internet side of things. Right. The underlying financials of the Starlink satellite Internet division generate the bulk of current revenue. The network has expanded to 9.2 million subscribers spread across 150 countries. That's a lot of dishes. It is. Those subscriptions generate roughly $16 billion in annual revenue, and the company reports an $8 billion profit from those specific operations. Okay, but those profit margins require incredibly close scrutiny, you know. Oh, absolutely. The global top line numbers look phenomenal on a slide deck. 50% profit margins sound like a software company, which is the exact narrative they want to push. Of course. But when you examine specific regional filings, the picture changes dramatically. Like there is a specific filing from their European subsidiary operating in the Netherlands that shows a net margin of less than 3%. Less than 3%. Less than 3%. And the gap between a 3% regional margin and a 50% global margin comes down to a very specific accounting mechanism. It's a $4 billion internal launch subsidy. Yeah, the mechanics of that subsidy are fascinating. The company generates 80% of its total launch revenue by paying itself to launch its own Starlink satellites. So they're their own biggest customer. It operates as a completely enclosed economic loop. When a Falcon 9 rocket launches a batch of satellites, the internet division pays the launch division. They're booking revenue for launches that serve to build their own internet infrastructure. Moving money from the left pocket to the right pocket. Exactly. By moving the money around like that, they can present a highly profitable launch cadence while absorbing the actual costs within the broader corporate structure. Which fundamentally changes how public markets must evaluate aerospace economics. No. By bundling the hardware manufacturing business, the launch services, and the high-margin internet subscription business together into one single entity, they successfully shift the focus away from the low-margin realities of building physical rockets. Because nobody wants to invest in low-margin hardware. Right. So public market investors are being asked to evaluate the entire conglomerate using software and connectivity multiples, effectively ignoring the traditional metrics used for industrial hardware valuations. But to justify that astronomical valuation jumped to $2 trillion, the company could not just rely on rockets or even the enclosed economics of satellite internet. They needed more. They had to alter the fundamental identity of the organization by completely absorbing an artificial intelligence firm. Okay, the XAI deal. Yes. They executed an all-stock transaction that merged XAI directly into the aerospace company. That specific merger valued the combined entity at trillion And it specifically valued the artificial intelligence firm XAI at billion And the mechanics of an all transaction are critical to understand here for anyone looking at this deal No actual cash changed hands to facilitate this merger Right. No wire transfers. Instead, they essentially printed new shares of the combined company and used those to absorb the artificial intelligence division. This maneuver allows them to assign these massive private valuations without needing to secure hundreds of billions of dollars in liquid capital from banks. And this execution relies on a financial concept known as the valuation matryoshka. The nesting dolls. Exactly. If you picture those Russian nesting dolls, where one fits perfectly inside the other, that is exactly what they have built with these corporate structures. It operates as a sequence of nesting corporate entities. How so? Well, first, the social media platform X was acquired by XAI at a significantly reduced valuation. In that step, X ceased to be an independent social media business evaluated on its ability to generate advertising revenue. Right. The ads don't matter as much anymore. Right. Instead, it became a dedicated data source used specifically for artificial intelligence training. Then the next all was stacked. The newly combined entity of X and XAI was absorbed by the aerospace company. Yeah. Hold on. Wait. Back up. I want to make sure you catch this because the structure is dizzying. It really is. So an unprofitable social media platform is inside an AI startup, which is now inside a rocket company. That is the exact structural reality of the organization. And the ownership web extends even further than that. Oh. Yeah. Tesla shareholders, following a previous $2 billion investment in XAI, now own preferred stock in an entity controlled entirely by an aerospace company. That is incredibly messy. And it gets better. The artificial intelligence division is carrying high double-digit interest rate debt, which is secured by physical graphics processing units as collateral. The chips themselves. Yes. They pledge their computer chips to secure the loans required to keep the systems running. Wow. Securing high interest debt with depreciating computer hardware is incredibly aggressive. Very. I mean, graphics processing units or GPUs lose value quickly as faster, more efficient chips hit the market. If you default on a loan secured by real estate, the bank takes the building, which likely holds its value. Land is land. But if you default on a loan secured by a two-year-old computer chip, the bank takes a piece of hardware that might be functionally obsolete. Exactly. But, you know, this massive corporate structure limits the ability of outside investors to scrutinize the individual cash burn or profitability of the social media platform or the artificial intelligence division. Because it's all buried. Because they are now shielded completely within the broader aerospace revenue, you cannot easily parse out how much money the large language models are losing. No, you can't. From a corporate finance perspective, I actually view this as a brilliant shielding maneuver. Really? Yeah. They have taken extreme cash burn technologies, social media platforms, and large language model training and hidden them safely inside a highly profitable government subsidized launch business. It protects the incredibly fragile AI valuation from direct public market scrutiny. Okay, I view this entirely differently. Yeah. You might see a brilliant shielding maneuver, but this operates as a severe governance nightmare. How do you mean? Consolidating these massive, highly distinct businesses without any independent fairness opinions creates a complete fiduciary stress test. Right, the boards overlap. Exactly. When you have overlapping boards of directors and preferred shareholders from an automotive company suddenly holding equity in a rocket manufacturer just because the founder decided to merge the entities, it creates a chaotic environment for any institutional investor trying to properly assess risk. That's a fair point. The blending of high interest debt secured by rapidly depreciating computer chips with revenue generated by launching critical federal satellites creates a highly unstable corporate architecture. So with the artificial intelligence division successfully absorbed into that architecture, the newly combined entity immediately utilized its massive computing power to court Cursor, which sits as the fastest growing tool in the entire enterprise software sector. Right. The mechanics of the Cursor deal present a strict binary choice. The agreement grants an option for an outright acquisition of the startup for $50 billion or alternatively, a $10 billion payment specifically allocated for joint development work. And to understand why those dollar amounts are so large for a software application, we have to look closely at the current AI coding tool ecosystem. It's crowded right now. Very crowded. There are three main approaches dominating the market right now. First, you have GitHub Copilot, which is an accessible $10 a month extension that functions essentially as highly advanced autocomplete. Right. You type. It finishes the thought. Exactly. A developer starts typing a line of code and Copilot suggests the rest of the line based on the context. Then you have Quad Code, which takes a completely different approach. It is a terminal native autonomous agent. And it features a one million token context window. And we should probably define what a token context window actually means for you. Yeah, that's a good idea. In artificial intelligence, a token is roughly equivalent to a word or a piece of a word. The context window is the model's short-term memory. It's how much information it can hold in its head at one exact moment. So a million words. Basically. A one million token context window means the AI can read and process an enormous amount of an application's code base simultaneously. It scores 80 on the SWE Bench for complex multi refactoring Which is high Very high The SWE bench is a rigorous testing framework that evaluates an AI ability to solve real world software engineering issues found on platforms like GitHub. Right. Scoring over 80 percent means the AI can look at multiple interconnected files, understand how a change in one affects the others, and successfully rewrite the code without breaking the application. Finally, you have Cursor, which is a $20 a month AI native integrated development environment. An IDE. An IDE, yeah. Unlike Copilot, which just sits inside an existing program, Cursor is an entirely self-contained workspace. Cursor excels at visual multi-file editing and utilizing background agents. Meaning the AI does work while you're busy doing something else. Exactly. A developer can highlight a section of code, ask the background agent to rewrite it to be more efficient, and the AI will complete that task while the developer continues working on a completely different part of the application. And Cursor has experienced hyper growth within that specific ecosystem. They have reached $2 billion in annual recurring revenue. That's wild. For an enterprise software company, hitting $2 billion in recurring revenue demonstrates extreme stickiness with their client base. Their software is currently used by 67% of the Fortune 500. Wow. Yeah, they have successfully captured the high-end enterprise developer market. But the underlying economics of Cursor's business model show exactly why they would agree to this arrangement with the aerospace company. The API cost. The API cost. Cursor currently pays retail prices to companies like Anthropic and OpenAI for direct access to their underlying language models. Right. Every single time a Cursor user prompts the system to generate a block of code, Cursor pays an API fee to the very companies they are competing against in the broader AI market. This is a huge disadvantage. They are effectively funding their own direct competitors just to keep their software functioning. Which is why the $10 billion alternative payment opens up a Compute for Equity handshake. handshake. Cursor gets direct access to the Colossus supercomputer, which possesses computing power equivalent to 1 million NVIDIA H100 chips. That is an unbelievable amount of power. Is. This access allows Cursor to train their own proprietary coding models without relying on their competitor's infrastructure. And what does SpaceX get? In exchange, the aerospace company gains immediate access to a massive stream of enterprise software revenue, which creates a highly attractive software narrative to bolster the financials of their upcoming public offering. The massive volume of compute required to train proprietary models for tools like Cursor, combined with the energy demands of the Colossus supercomputer, is rapidly exhausting the physical limits of Earth's power grid. We're just running out of power. We are simply running out of places to plug these machines in. This severe resource exhaustion has led to the most extreme proposal the aerospace company has ever produced. They submitted a formal filing to the Federal Communications Commission requesting authorization for an entirely new non-geostationary orbit system. A new satellite constellation. Yes. The application details a network of up to 1 million individual satellites designed specifically to function as fully operational orbital data centers. The physics and orbital parameters of this plan are fascinating to examine. The satellites would operate at altitudes between 500 and 2,000 kilometers above the Earth, specifically utilizing sun-synchronous orbits. Which is a very deliberate choice. Right. This exact orbit matters immensely because the satellite rides the Terminator line. The Terminator line is the dividing boundary between day and night on Earth. Right. By matching the satellite's orbital speed with the Earth's rotation around the sun, the satellite never falls into the planet's shadow. This specific placement allows for near constant solar exposure. And because the solar panels are operating in the vacuum of space, there is absolutely no atmospheric filtering of the sunlight. So it's perfectly clear. Yes. On Earth, clouds, dust, and the atmosphere itself scatter the light before it ever hits a solar panel. In space, the light hits the panels with complete intensity, and because they are riding the Terminator line, there are no nighttime cycles to interrupt power generation. Constant power. A solar panel in this specific orbit can generate up to 40 times more usable energy than an identical panel placed on the surface of the Earth. But the technical infrastructure required to link one million server nodes in space is highly complex. The system relies entirely on high-bandwidth optical inter-satellite links. Oh, these servers. Yeah, they are essentially creating a vast laser mesh network in a vacuum to rapidly route data traffic between the servers and down to the ground stations. Right. Furthermore, placing these incredibly heavy compute nodes into space relies entirely on the Starship vehicle's 200-ton payload capacity. A standard communications satellite weighs a fraction of what a server rack weighs. No other launch vehicle currently in existence possesses the raw lift capability required to physically transport this much heavy server hardware into orbit. Wait, let me just pull back here for a second because we are talking about launching massive server farms into the vacuum of space. Yeah. Simply because we literally do not have enough electricity available on the ground to power the artificial intelligence industry. Exactly. We are out of juice down here, so we have to go up. And the financial equation provided by the company in their filings attempts to rationalize the physics of doing this. How do they justify the cost? They claim that with the launch costs of the Starship vehicle dropping below per kilogram orbital computing could actually become 25 cheaper than running ground operations really cheaper yes think about what goes into building a terrestrial data center you have the rising cost of commercial real estate you have massive terrestrial grid interconnection delays where power companies tell developers they will have to wait years just to hook up a new facility because the grid can't handle it right you also have the immense cost of cooling infrastructure. Keeping a million running computer chips from melting down requires millions of gallons of water and massive industrial air conditioning systems. By moving the servers to a vacuum that sits near absolute zero, the cooling problem is functionally solved by the environment itself. That's fascinating. And they aim to eventually generate 100 gigawatts of compute capacity through this orbital network. 100 gigawatts. In the regulatory filing, They specifically describe this energy capacity as a step toward achieving a Kardashev-Takka-level civilization. Oh, the Kardashev scale. Yeah, the Kardashev scale is a method of measuring a civilization's level of technological advancement based entirely on the amount of energy it is able to harness. Okay, remind me how that works. A Type I civilization can harness all the energy of its home planet. A Type II civilization can harness all the energy of its home star. By moving power generation and computation directly into space to capture unfiltered solar energy, they are framing this infrastructure project as an evolutionary leap in human technology. This physically relocates the primary bottleneck of artificial intelligence development. The constraint-limiting AI growth is no longer the physical power grid infrastructure or the availability of real estate on Earth. The new bottleneck is strictly launch cadence and payload capacity in space. If a company wants more computing power, they physically have to launch more rockets to put more servers into orbit. But relocating critical computing infrastructure into orbit using a complex mix of international software and hardware components immediately triggers intense government scrutiny. Oh, I'm sure regulators love this. Not at all. The regulatory backlash against this plan is already servicing rapidly. U.S. senators have formally demanded probes into the company regarding strict foreign ownership, control, or influence rules. Oh, because of the funding. Exactly. There are specific, detailed reports outlining undisclosed funds being routed through various entities in the Caribbean and the British Virgin Islands, specifically to purchase private shares in the aerospace firm. And this intense scrutiny connects directly back to the foundational code of the artificial intelligence models themselves. There is intense controversy currently surrounding Cursor's Composer 2 model. Right, the code origin. It was discovered that the system was secretly built upon the Chinese artificial intelligence model known as Kimi-K 2.5. And the broader legislative actions responding to these discoveries are severe. Very severe. There is the No Deep Seek on Government Devices Act moving through the legislature, along with bipartisan bills, specifically barring federal contractors from utilizing any artificial intelligence models affiliated in any way with foreign adversaries. But, you know, utilizing open-weight international models like Kimi or DeepSeek is entirely standard practice for rapid iteration in the technology sector. Sure, in standard tech. Right. Developers take the absolute best available open source tools, refine them locally, and build specialized application layers directly on top of them. That is exactly how the software industry continuously achieves hyper growth. Yep. Punishing a company for using the most efficient, effective code available stifles global innovation and completely misunderstands how global software development actually functions in reality. I get that. But a defense contractor cannot have undocumented code originating from foreign adversaries embedded within its core infrastructure. You think it's too risky. It's not simply a matter of software efficiency or skeet of development. It is strictly about data security and the severe potential for espionage. Okay, fair. When an aerospace company is responsible for managing highly classified national security payloads, and they decide to integrate an AI tool built on a foreign adversary's architecture into their operational network, the risk of hidden backdoors or data exfiltration becomes a profound vulnerability. The geographic origin of the code strictly dictates the operational security of the entire system. And this intense regulatory scrutiny severely limits their operational freedom and directly threatens a $24.4 billion backlog in Department of Defense and National Reconnaissance Office contracts. $24 billion on the line. Yeah. If the federal government strictly enforces these foreign influence rules, it could potentially force the company to completely rip out the foundation of its newly acquired artificial intelligence tools, crippling the very software narrative they are utilizing to sell the public markets on a $2 trillion valuation. Which is wild. By wrapping rockets, satellite internet, and advanced coding software together into a single, massive financial package, the company is attempting to construct an entirely independent, off-planet computing infrastructure while simultaneously commanding the largest public valuation ever seen. So will public market investors agree to price this sprawling conglomerate based strictly on its off-planet computing dreams? Or will the intense regulatory risks and the staggering cost of launching the hardware drag the entire valuation right back to Earth? If you're not subscribed yet, take a second and hit follow on whatever app you're using. It helps us keep making this. We appreciate you being here. Also, check out our YouTube channel for more business and tech updates. There's a link in the description.
