Nvidia and SpaceX Form a "Match Made in Heaven"! Rubin Computing Power Flies into Space Orbit, Musk Anchors SpaceX's Valuation to Kardashev Type II Civilization

According to reports from Zhicheng Finance APP, just as the "AI chip superpower" Nvidia (NVDA.US) released another outstanding earnings report, further raising global AI capital expenditure expectations and sparking a new bull market in the AI compute industry chain, Elon Musk is attempting to push the boundaries of AI infrastructure from the ground into Earth orbit. His company, SpaceX (SPCX.US), plans to launch the first AI data center satellites in the fourth quarter of 2027 using Nvidia's next-generation compute cluster—the Vera Rubin architecture, featuring AI GPU-dominated clusters—with the goal of achieving "significant scale" by 2028. This does not imply that ground-based data centers will be quickly replaced; rather, it bets that orbital compute can bypass major bottlenecks such as power grids, land, and water resources on Earth, becoming an additional layer of AI computing supply.

For Nvidia, this represents an expansion of Vera Rubin's potential market from "ground-based AI factories" to "orbital AI factories." For SpaceX, it means integrating Starship launches, orbital energy, satellite networks, and AI cloud computing into a vertically integrated platform. However, Wall Street financial institutions such as Evercore expect substantial revenue to materialize no sooner than fiscal year 2029, while heat dissipation, radiation, collision, and regulatory issues remain key constraints determining the feasibility of commercialization.

Nvidia and SpaceX Forge "A Match Made in Heaven," Planning to Send Vera Rubin Into Orbit

In a parallel universe, Nvidia CEO Jensen Huang and "Mr. Everything" and world’s richest man Elon Musk might be a pair of cosmic superheroes. But in this universe, on this planet, they are still making plans for an ambitious space AI project that seems almost unimaginable—Nvidia provides the core compute power, while SpaceX offers launch and orbital infrastructure, collectively pushing the physical boundaries of AI computation. Vera Rubin and Starship form a closed-loop space software-hardware system and infrastructure, though current valuations mainly reflect long-term optionality rather than near-term cash flows.

This project is part of the increasingly close collaboration between Nvidia and SpaceX. During Nvidia's August 26 earnings call, CFO Colette Kress stated that the Vera Rubin systems are in full production, with key shipments going to leaders in AI cloud computing and computing rentals such as Oracle (ORCL.US), Amazon (AMZN.US) AWS, and SpaceX (SPCX.US).

"We believe Nvidia's explicit listing of SpaceXAI as a primary partner further demonstrates the depth of the relationship and proves SpaceXAI is among the earliest Vera Rubin customers," the Evercore ISI analyst team wrote in an investor report. "The relationship is also bidirectional: Nvidia has disclosed holding 122.8 million shares of SpaceX stock, making Nvidia both a major supplier and a significant shareholder in SpaceX."

Nvidia’s latest financials showed Q2 FY2027 revenue of $96.22 billion, up 106% year-on-year; data center revenue of $89 billion, up 117% Y/Y; and adjusted EPS of $2.22. The company forecasts Q3 revenue of around $108 billion, plus or minus 2%, and has unusually guided for about 70% revenue growth in FY2028, far surpassing Wall Street's previous expectation of 44%. Nvidia management said at the earnings meeting that Vera Rubin is in full production and being delivered, which largely means Nvidia’s growth constraints are still mainly from memory capacity and TSMC's advanced process manufacturing bottlenecks, rather than a lack of orders.

After Nvidia announced its results, Wall Street raised price targets one after another. As of last Friday, Nvidia shares closed at $217.55. Citi raised its target from $300 to $315 and maintained a "Buy" rating; Goldman Sachs raised its target from $285 to $300; Morgan Stanley from $288 to $300. Their bullish consensus is based on strong, ongoing demand for AI compute, Vera Rubin's scaling, and Nvidia’s strengthening integrated software-hardware platform advantages.

Musk has stated that the data centers designed by SpaceX will be fully powered by Nvidia AI GPU accelerators. In addition, in August, SpaceX completed an all-stock acquisition of Cursor valued at about $60 billion. Jensen Huang recently noted in a PPT presentation that Cursor is now used throughout Nvidia. The latest third-party data show that this AI programming assistant's annualized revenue run rate has surpassed $1 billion.

Regarding commercialization of the space AI data center—a joint focus for Musk and Huang—Evercore expects that the project won't make any substantial contribution to SpaceX until at least fiscal year 2029.

"Notably, we haven't included any orbital compute revenue for 2026 or 2027, and expect the first 1 GW orbital AI compute capacity to go online in FY2029; by that fiscal year-end, orbital compute capacity will reach 8 GW, while ground compute will be 10 GW," said Evercore. "If the originally planned orbital build-out for FY2029 moves forward to 2028, it will significantly unleash inference-side AI compute resource demand. Assuming Musk's economics of $30 to $50 per watt, our forecasts could see significant upside."

Evercore has given SpaceX an "Outperform" rating with a price target of $230. As of last Friday's US market close, SpaceX stock was at $141.50.

"The key question is, what exactly does Musk’s so-called ‘significant scale’ in 2028 really mean—does it refer to the amount of compute capacity generating revenue or is it a broader technical validation phase before commercialization?" Evercore added.

However, according to a recent report from the Brookings Institution, deploying data centers in space faces major technical hurdles.

"Ground-based data centers use both air cooling and server liquid cooling systems, conducting and convecting heat away," said the Washington-DC-based think tank. "They transfer heat from chips to the liquid coolant and then from the coolant to the surrounding atmosphere. For data centers floating in the vacuum of space, this cooling method is nearly impossible because there’s no atmospheric medium to absorb the heat. In fact, radiation may be the only way to disperse heat from orbital data centers, but some scientists calculate that to effectively dissipate the heat generated by a single orbital data center, you’d need a staggering 2.15 million square feet of radiators," Brookings reported.

Other issues include damage to equipment from constant and powerful solar ultraviolet bombardment, as well as an increasing probability of orbital collisions as the number of satellites grows. Regulatory approval could also present challenges.

AI Data Centers Set to Scale Up? AI Super Factories Preparing to Shift From the Ground to Orbit

SpaceX’s offering prospectus explicitly defines "building an ever-expanding space civilization, ultimately moving towards a Type II civilization capable of harnessing all the Sun’s energy," as the company’s long-term paradigm shift. Musk himself has said that lunar satellite factories, mass drivers, and annual AI hardware deployment exceeding 100 terawatts would help humanity make "nontrivial progress" toward Type II civilization. Therefore, orbital AI data centers are not standalone projects, but rather the first layer of infrastructure linking "Earth-bound compute—solar system energy—multi-planetary civilization."

The so-called Kardashev Type II Civilization is one that can utilize all the energy of its host star, while Type III would control the energy of an entire galaxy—far beyond SpaceX's current narrative.

SpaceX's stated $28.5 trillion total addressable market (TAM) further capitalizes on this civilization vision: traditional space business is only $370 billion, Starlink-led high-speed connectivity business is $1.6 trillion, while the AI comprehensive business market is as large as $26.5 trillion, accounting for about 93% of the total—including $2.4 trillion in comprehensive AI compute infrastructure, $760 billion in consumer subscriptions, $600 billion in digital advertising, and $22.7 trillion in enterprise applications.

The underlying assessment from Musk and other SpaceX top executives is that by 2030 global data center compute demand could reach 235 gigawatts, 70% of which would be for AI, but the Earth’s grid, land, permitting, and environmental capacity cannot support terawatt-level expansion; the sun, meanwhile, accounts for about 99.8% of the solar system’s energy. Therefore, SpaceX plans to begin commercial modular orbital compute later this decade, with the long-term aim of deploying 100 gigawatts of AI compute in orbit per year—if run continuously, the energy consumed would be about one-fifth of projected US total annual electricity output in 2025.

However, from an investment perspective, $28.5 trillion is a theoretical market boundary—not a forecast for revenue or profit; orbital AI is essentially a high-duration, high-convexity technological option. To achieve 100 GW annual deployment would require Starship to complete thousands of launches annually, lifting around one million tons of equipment into orbit and solving issues like radiation heat dissipation, unrepairable chip failures, mass satellite manufacturing, and regulatory challenges.

At the same time, SpaceX’s AI business capital expenditure already reached $12.727 billion in 2025 with adjusted EBITDA losses of $1.237 billion, and Q1 2026 capex further increased to $7.723 billion with adjusted EBITDA losses of $609 million. The company admits that ongoing profitability will require years of investment. Therefore, recent valuations should be underpinned by cash flow from launches and Starlink, with orbital AI providing long-term upside for the future; what investors should truly watch are Starship’s per-ton to-orbit costs, launch frequency, compute per ton, thermal reliability, and utilization rates—not the $28.5 trillion as a direct endorsement of current valuations.