Fresh off Nvidia's (NVDA.US) blockbuster earnings report that raised the bar for global AI capital expenditure and signaled a renewed bull cycle across the AI computing supply chain, Elon Musk is now pushing the expansion frontier of AI infrastructure from the ground to the space orbit. His brainchild, SpaceX (SPCX.US), plans to launch its first batch of AI data center satellites powered by Nvidia's next-generation computing cluster, the Vera Rubin architecture-based AI GPU cluster, in the fourth quarter of 2027, achieving "significant scale" by 2028. This does not mean ground-based data centers will be rapidly replaced, but rather bets that orbital computing can bypass major bottlenecks such as ground power grids, land, and water usage, thereby becoming an incremental supply layer for AI compute capacity.

For Nvidia, this represents the extension of Vera Rubin's potential market from "ground-based AI factories" to "orbital AI factories"; for SpaceX, it integrates Starship launches, orbital energy, satellite networks, and AI cloud computing into a vertically integrated platform. However, Wall Street financial institutions like Evercore anticipate that material revenue figures won't materialize until fiscal 2029 at the earliest, with heat dissipation, radiation, collision risks, and regulatory hurdles remaining key constraints on its commercialization potential.

In some parallel universe, Nvidia CEO Jensen Huang and "Mr. Everything" and world's richest man Elon Musk might be cosmic superheroes. But in this universe, they are still plotting what seems like the unthinkable — a space-based AI mega-project. Nvidia provides the compute core, SpaceX provides the launch and orbital infrastructure, together expanding the physical boundaries of AI computing.

The Vera Rubin and Starship combination forms a closed-loop of space-based hardware, software, and infrastructure, but current valuations primarily reflect long-dated optionality rather than near-term cash flow. This project is part of the increasingly close partnership between Nvidia and SpaceX. During Nvidia's August 26 earnings call, Chief Financial Officer Colette Kress stated that the Vera Rubin system is now in full production, with key shipments going to AI cloud and compute leasing leaders such as Oracle (ORCL.US), Amazon's (AMZN.US) AWS, and SpaceX (SPCX.US). The Evercore ISI analyst team noted in an investor report that Nvidia's explicit acknowledgment of SpaceX AI as one of its key partners further validates the depth of the relationship and confirms SpaceX AI is among the earliest Vera Rubin customers. The relationship is also mutual, as Nvidia has disclosed holding 122.8 million shares of SpaceX stock, making Nvidia both a key supplier and a significant shareholder of SpaceX.

Nvidia's latest results show fiscal 2027 second-quarter revenue of $96.22 billion, up 106% year-over-year; data center revenue of $89 billion, up 117%; and adjusted earnings per share of $2.22. The company guided third-quarter revenue to approximately $108 billion, plus or minus 2%, and made a rare forecast of about 70% revenue growth for fiscal 2028, far exceeding Wall Street's prior expectation of 44%. During the earnings call, management confirmed that Vera Rubin is in full production and shipping, indicating that Nvidia's growth constraints still primarily stem from memory capacity and TSMC's advanced process manufacturing capacity bottlenecks, rather than insufficient orders. Following the earnings release, Wall Street firms have raised their price targets, with Nvidia shares closing Friday at $217.55. Citi lifted its target from $300 to $315, maintaining a "Buy" rating; Goldman Sachs raised its target from $285 to $300; Morgan Stanley raised its target from $288 to $300. Their consensus bullish view rests on sustained strong AI compute demand, Vera Rubin's ramp, and Nvidia's strengthening integrated software-hardware platform advantage.

Musk stated that the data centers being designed by SpaceX will operate entirely on Nvidia AI GPU accelerators. Additionally, SpaceX completed an all-stock acquisition of Cursor in August, valued at approximately $60 billion. Huang noted in a recent presentation that Cursor is now used company-wide across Nvidia. The latest third-party data shows the AI coding assistant's annualized revenue run rate has surpassed $1 billion.

As for the commercialization of the space-based AI data centers on which Musk and Huang are jointly focused, Evercore projects that the project will not deliver any meaningful contribution to SpaceX before fiscal 2029. "Notably, we have not included any orbital computing revenue in 2026 or 2027, and we expect the first gigawatt of space-based AI orbital compute capacity to come online in fiscal 2029; by the end of that fiscal year, orbital compute capacity will reach 8 gigawatts, compared to 10 gigawatts of ground-based capacity," Evercore stated. "If orbital construction scheduled for fiscal 2029 is accelerated to 2028, it could significantly unlock inference-side AI compute demand; and based on the $30 to $50 per watt economics Musk has discussed, our forecasts could see meaningful upside." Evercore has assigned SpaceX an "Outperform" rating with a price target of $230. As of Friday's market close, SpaceX shares were trading at $141.50.

"The key question is what Musk means by 'significant scale' in 2028 — whether it represents compute capacity capable of generating revenue, or 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 significant technical hurdles. "Ground-based data centers use both air cooling and server liquid cooling systems, dissipating heat through thermal conduction and convection," the Washington, D.C.-based think tank stated. "They first transfer heat from chips to liquid coolant, then from the coolant to the surrounding atmosphere of the facility. For data centers floating in the vacuum of space, this cooling strategy is nearly impossible to implement because there is no atmosphere to absorb heat. In fact, thermal radiation may be the only way to dissipate heat from orbital data centers, but some scientists estimate that using this method to properly exhaust heat from a single orbital data center would require an astonishing 2.15 million square feet of radiators." Other issues include equipment damage from continuous exposure to intense solar ultraviolet bombardment, and an increasing probability of orbital collisions as the number of satellites grows. Regulatory approvals could also become a concern.

SpaceX's prospectus explicitly defines "building a continuously expanding space civilization, ultimately advancing toward a Type II civilization capable of harnessing the Sun's full energy" as a long-term paradigm shift. Musk himself has stated that lunar satellite factories, mass drivers, and the deployment of over 100 terawatts of AI hardware annually would drive "non-trivial progress" toward a Type II civilization. Therefore, orbital AI data centers are not isolated projects but rather the first layer of infrastructure connecting "Earth-constrained compute power," "solar-system-scale energy," and "multi-planetary civilization." The so-called Kardashev Type II civilization refers to the ability to utilize all of a star's energy, while Type III would command the entire galaxy's energy — far beyond SpaceX's current narrative.

SpaceX's claimed $28.5 trillion total addressable market (TAM) further capitalizes on this civilizational vision: traditional space business accounts for only $370 billion, Starlink-led high-speed connectivity business $1.6 trillion, while the AI-integrated business market size reaches a staggering $26.5 trillion — approximately 93% of the total — including $2.4 trillion in AI-integrated compute infrastructure, $760 billion in consumer subscriptions, $600 billion in digital advertising, and $22.7 trillion in enterprise applications. The underlying judgment of Musk and other top SpaceX executives is that global data center compute demand may reach 235 gigawatts by 2030, with 70% used for AI, while Earth's power grids, land, approvals, and environmental capacity are insufficient to support terawatt-scale expansion. The Sun, however, accounts for approximately 99.8% of the solar system's energy. SpaceX therefore plans to begin commercializing modular orbital compute by the end of this decade, with a long-term goal of deploying 100 gigawatts of AI compute capacity to orbit annually — if running continuously year-round, its energy consumption would be roughly equivalent to one-fifth of the entire U.S. annual electricity generation in 2025.

But from an investment perspective, $28.5 trillion represents a theoretical market boundary, by no means a revenue or profit forecast. Orbital AI is essentially a very long-dated, high-convexity technology option. Achieving 100 gigawatts of annual deployment would require Starship to conduct thousands of launches per year, sending approximately one million tons of equipment into orbit, while solving problems related to radiation dissipation, irreparable chip failures, satellite mass manufacturing, and regulatory issues. Meanwhile, SpaceX's AI business has already spent $12.727 billion in capital expenditures in 2025 with an adjusted EBITDA loss of $1.237 billion, and in Q1 2026 further spent $7.723 billion in capex with an adjusted EBITDA loss of $609 million. The company also acknowledges that sustained profitability will require a multi-year investment cycle. Therefore, near-term valuation should be anchored by launch and Starlink cash flows, with orbital AI contributing long-term upside optionality; what investors really need to track are Starship's per-unit launch cost, launch frequency, compute per ton, thermal dissipation reliability, and compute utilization — rather than directly underwriting current valuations with the $28.5 trillion figure.