Starship Soars as SpaceX Looks to AI
SpaceX’s colossal Starship launch vehicle lifted off July 24th from the company’s launch facility in South Texas on its thirteenth integrated flight test. The test was, broadly speaking, a repeat of the vehicle’s twelfth all-up test in May, which was itself a repeat of previous flight tests. As with earlier tests, the second V3 Starship took off from Starbase, Texas at 5:51PM Central Daylight Time on a suborbital lap around the planet. After engine cutoff, the Starship vehicle successfully deployed twenty next-generation Starlink satellites, marking the first time that the upgraded internet-beaming satellites reached space. While functional, the satellites’ stay in space was brief, as they reentered the atmosphere alongside the main Starship vehicle minutes later. As the satellites were incinerated, the Starship itself, protected by its heat shield, made its way through the atmosphere to a pinpoint soft landing in the Indian Ocean, remaining intact and afloat after its splashdown. The Super Heavy booster, however, stumbled on its return. After an early end to its boostback burn, the massive booster failed to relight enough engines to safely slow down for its planned splashdown in the Gulf of Mexico.
The broadly successful test flight, however, did not produce enthusiasm on Wall Street. Shares in the newly-public company sank in the week that followed the flight, ending the week down 3.5%. As revealed in SpaceX’s S-1 filing, the Starship system is central to multiple aspects of the business. In the S-1, SpaceX notes that they expect future launches of Starship to carry up to 60 V3 Starlink satellites in a single launch. SpaceX says that they expect a single Starlink-laden Starship launch to deliver around twenty times the capacity to the Starlink network that a single Falcon 9 launch of current-generation satellites delivers. The S-1 further indicates how essential Starlink is to SpaceX’s business: for Q1 of 2026, SpaceX’s connectivity business generated approximately five times the revenue of their space business. With SpaceX promising to begin deliveries of Starlink satellites to orbit aboard Starship by the end of the year, Thursday’s test flight represents a hope that SpaceX can conclude its multiyear suborbital test campaign of Starship and push towards orbital flight and the full reusability of Starship their plans demand.
In addition to Starlink, Starship is also essential to SpaceX’s recently-absorbed AI business. SpaceX merged with the Musk-owned xAI in February, granting it control of xAI’s mammoth data centers, as well as ownership over the company’s Grok chatbot and popular social media platform X (formerly Twitter). The newly combined entity inked deals with AI industry titans Anthropic and Google in May and June respectively, with each company agreeing to pay SpaceX on the order of a billion dollars monthly for use of SpaceX’s data centers. Revenues for the company’s AI segment already outstrip the revenues for the Space segment for Q1 2026, and SpaceX expects AI to be by far the largest segment of their business in the future, with enterprise AI solutions representing the lion’s share of the eye-watering 28.5-trillion-dollar total addressable market SpaceX claims in their S-1 filing (the gross domestic product of the United States is approximately $32 trillion).
SpaceX aims to use Starship as the launch vehicle for its future constellation of AI compute satellites. The company disclosed in a January filing plans to deploy a constellation of up to one million AI satellites in low earth orbits to power both the company’s own models, as well as provide capacity for external users. Part of SpaceX’s deal with Anthropic, for instance, includes Anthropic studying the possibility of using SpaceX’s space-based compute capacity in the future. SpaceX points to readily available solar power in sun-synchronous orbits and public backlash to terrestrial data centers as reasons to relocate its compute to space. However, the plan to launch such a staggering volume of satellites has drawn significant scrutiny. Internet giant Amazon filed a public comment opposing the plan in March, one of well over a thousand public comments filed with the FCC on the subject. Concerns expressed by the public over the plan ranged from light pollution degrading astronomy to overcrowding valuable orbital bands. The basic premise of performing AI compute in space is also not without detractors – critics argue that high launch costs combined with thermal and radiation challenges will combine to make space-based compute uneconomical.
Some of those headwinds are the stiff competition SpaceX faces in the AI space. In addition to the test flight of Starship, recent weeks have been marked by a flurry of major product launches across the AI space. SpaceX led the charge of new model releases with Grok 4.5 on July 8th with bold promises of frontier-level performance. Intense reorganization within the AI division post-merger, including an option to acquire AI programming company Cursor, has produced a Grok model that far outstrips its predecessors in coding ability. Independent analysis group Artificial Analysis ranked the model on release in fourth place on their Intelligence Index, behind rival Anthropic’s Fable 5 and Opus 4.8 models, as well as OpenAI’s ChatGPT-5.5. The Intelligence Index measures model performance across a series of benchmarks covering general knowledge as well as programming-specific tasks.
Despite the capability jump for Grok, the rest of the industry moved rapidly to erase SpaceX’s newfound lead. Mere days after its launch, Grok 4.5 had fallen from fourth to seventh place on the Intelligence Index. Two of chatbot juggernaut OpenAI’s new ChatGPT-5.6-family models eclipsed Grok’s performance on the index after releasing on July 9, as did open-weights model Kimi K3 from the Beijing-based Moonshot AI just under a week later. At the time of writing, Grok 4.5 remains in seventh place, behind the leading models from OpenAI, Anthropic, and Kimi.
The brevity of SpaceX’s lead perhaps contributed to the low confidence investors have thus far shown in SpaceX’s AI business. Analysts at Morgan Stanley said that many investors expect the share price to fall even further, towards or below $100/share, as lock-ups end, allowing some insiders to sell. Morgan Stanley said that at $100/share, investors would be assigning essentially zero value to the company’s AI business. While the analyst at Morgan Stanley who produced that report expressed considerably more bullish sentiment on SpaceX’s future, with a $300/share price target, there is far from a broad consensus on that point. Pre-IPO, Morningstar valued SpaceX at a mere $63/share.
What is clear, however, is that Starship’s success is essential to SpaceX’s future. It is the cornerstone to expanding their connectivity business, as well as the foundation of their planned AI expansion that justifies the bullish cases. Last week’s test flight, while short of demonstrating a mature system, was the strongest indication yet that Starship may finally be ready for prime-time.
