Morgan Stanley has given SpaceX (SPCX.US) a new label: a "safety cushion" amid AI turbulence, with a target price of $300.

Since last Friday, a debate over whether the development of frontier AI models should "actively slow down" has rapidly spread from the tech community to the capital markets. Anthropic CEO Dario Amodei released a three-step plan on September 12, calling to "set the pace for frontier AI development." OpenAI's Sam Altman subsequently stated he would adopt the "independent evaluators with employee-level access" proposal put forward by Amodei, and even Elon Musk, long known for his aggressive stance, rarely joined this camp. On the first trading day after the news gained traction, global tech stocks briefly sold off sharply. Yet amid the panic selling in AI concept stocks, Morgan Stanley analyst Adam Jonas delivered a rather unconventional judgment to clients: SpaceX (SPCX.US) may in fact be the AI name with relatively stronger defensive qualities in this round of turbulence. In a research note published on September 15, Jonas reiterated his "Overweight" rating and $300 price target on SpaceX, offering the following reasons. A diversified foundation: not just an "AI concept stock" Jonas opened the report with a rather candid statement. He wrote: "You can't have your cake and eat it too." The implication is that the vast majority of SpaceX's capital expenditure and long-term earnings growth is indeed driven by the enterprise AI business, and any material slowdown in the AI trade would be a headwind for the stock. But the other side of the issue is that SpaceX's revenue structure is not a pure bet on the AI compute narrative. According to the company's Q2 2026 earnings report released on August 4, Starlink's global subscriber base has reached 12 million, double the 6 million in the same period last year, with a net increase of 1.7 million in the quarter. Continued expansion among enterprise and government customers has made Starlink SpaceX's largest source of revenue. Jonas believes the growth DRIVE of this cash cow business comes from market penetration of satellite communications itself, rather than the industry-wide AI capex cycle. Meanwhile, SpaceX's Starship program is also advancing at its own pace. According to the latest notice from the Federal Aviation Administration, Starship's 14th test flight has been adjusted to as early as September 22, launching from Starbase in Texas. This mission will attempt orbital insertion and deploy real Starlink satellites for the first time. More noteworthy is the 15th flightif the 14th mission goes smoothly, SpaceX plans to conduct the next launch in less than 30 days and attempt to catch both the Super Heavy booster and the Starship vehicle simultaneously using the launch tower's "chopstick" robotic arms, a key validation step for a fully reusable system. Jonas argues that the Starship launch cadence is a classic engineering-driven milestone, almost unrelated to AI model training progress or regulatory wrangling. Long-dated option: orbital AI and Siasun Robot&Automation synergies If Starlink and Starship constitute SpaceX's "non-AI fundamentals," then the space compute project is what Jonas sees as a long-term option. In early 2026, SpaceX applied to the FCC for a license to launch as many as 1 million data center satellites. In July, the company formally disclosed an orbital data center project called "Starmind." The first AI satellite, AI1, has peak compute of about 150 kilowatts, equipped with a 70-meter CECEP Solar Energy array wing span and a liquid radiator cooling system. SpaceX has also partnered with NVIDIA Corporation (NVDA.US) to customize a space-optimized AI computing system based on the Vera Rubin NVL72 architecture for the orbital environment, with the first launch targeted for Q4 2027. Jonas acknowledges that scaled deployment of this vision will most likely occur after 2030, essentially "a fairly long-dated option." But he also notes that the market currently assigns almost no valuation to SpaceX's collaboration with Tesla, Inc. (TSLA.US) in Siasun Robot&Automation and domestic chip manufacturing. According to public information, the two companies are jointly advancing the "Terafab" semiconductor manufacturing project, aimed at integrating logic chips, memory, packaging, and testing within a single facility to supply chips for Tesla, Inc.'s Optimus humanoid Siasun Robot&Automation and Cybercab, while also supplying components for the space data centers SpaceX is developing. Tesla, Inc. has taken a stake in SpaceX, and the two sides have also signed agreements covering joint projects including Terafab and "digital Optimus." Musk himself has described the relationship as "mutual dependence," warning that "without Terafab, Tesla, Inc.'s ability to scale Optimus production will be constrained." Financial flexibility: cash-rich, spending only what it can borrow On the balance sheet front, Jonas specifically highlighted the capital allocation discretion of SpaceX management. As of the end of Q2, the company held about $100 billion in total cash and about $6 billion in net cash. Jonas expects the company to raise substantial financing through debt markets in the future, pushing the debt/EBITDA ratio to 1.5x to 2x, but the key is that both the pace and scale can be flexibly controlled by management"it only spends what it can borrow, and only spends it if the business plan works." After its IPO, SpaceX has obtained investment-grade credit ratings from all three major international rating agencies and has issued $20 billion in short-term unsecured notes to repay bridge loans maturing next year. This financing capacity provides an additional layer of buffer in an AI slowdown scenario. Valuation: the market has barely priced in the AI business From a valuation perspective, Jonas calculates that of SpaceX's roughly $2 trillion market capitalization, about three-quarters is contributed by the space and communications businesses. Excluding the AI business, the implied share price is about $125that is, the space business corresponds to about $8 per share and the satellite communications business $118 per sharewhile at the current share price, the AI business (enterprise AI plus X and Grok) receives only about 10% of implied valuation, or about $17 per share (based on its $183 AI business valuation). Back in July, Jonas already put forward the thesis that "SpaceX at $100 means AI value goes to zero." He further ran a sensitivity analysis: each additional 1GW of nameplate compute (at $50 per watt, 70% incremental margin, 10x EBITDA capitalization) could add about $27 per share of value, equivalent to 15% to 20% of the current share price. Morgan Stanley's AI compute forecast for fiscal 2027 is 4.9GW, while the company's own target is "close to 10GW." The "Musk factor" Jonas's fifth reason returns to Musk himself. He wrote in the report that over two decades of observing this CEO's management style, "he is not the kind of person who would waste a good crisis." Tesla, Inc.'s performance during the 2018-2019 "production hell" and the 2020-2021 COVID-19 pandemic is seen by Jonas as historical evidence of the company achieving dual operational and share price breakthroughs under high-pressure conditions.