According to reports, on September 28, SpaceX (SPCX.US) launched its 14th integrated Starship test flight from the Starbase facility in Texas, reaching Earth orbit for the first time and deploying 26 third-generation Starlink (V3) satellites through a deployment mechanism on the spacecraft nicknamed the "candy dispenser." The market clearly cares about more than just this single launch, as Wall Street investment banks have broadly turned bullish.
TD Cowen initiated coverage on Monday with a "Buy" rating and a $200 price target; alongside Royal Bank of Canada (RBC), Clear Street, and CLSA, the four institutions set target prices ranging from $200 to $250, implying roughly 37% to 72% upside based on Monday's closing price. In this wave of bullishness, what keeps coming up is not rockets, but compute leasing.
Starship's First Orbital Flight: All Satellites Deployed, Two Orbits Before Return
According to reports, this mission used the B21 booster and S41 spacecraft, both part of the third-generation Starship system; one of the spacecraft's six engines malfunctioned, and live commentary initially judged that orbital insertion would be impossible, but the engineering team reassessed and confirmed that an orbital burn could still be attempted. About 25 minutes after launch, Starship entered orbit, then spent roughly half an hour releasing satellites one by one, establishing contact with all 26 satellites, with Musk confirming all were operating normally.
The flight was not flawless. One engine shut down early during ascent, and a mission originally planned to orbit Earth about 6 times at roughly 275 kilometers altitude over nearly 10 hours was ultimately shortened to about 3 hours and just 2 orbits, with the upper stage's remaining engines extending their burn as compensation. The Super Heavy booster splashed down in the Gulf of Mexico, and no attempt was made to catch it with the launch tower's mechanical arms this time; the spacecraft completed its first deorbit burn and splashed down in a designated area of the Pacific Ocean west of Chile.
The significance of this flight lies in this: since the first integrated test flight in 2023, all 13 previous test flights had remained suborbital, and reaching orbit is the basic threshold for a launch vehicle. Aerospace engineer Dean Sladen called delivering 26 V3 satellites into their target orbit a "huge victory," also marking Starship's first revenue-generating commercial flight—these satellites are worth millions of dollars. Georgetown University Center for Security and Emerging Technology senior analyst Catherine Curley took a more measured stance: the mission made significant progress but was not a "super success," noting that "by SpaceX's own metrics and goals, they did not complete the 10-hour flight, and the engine did have problems."
1 Tbps Per Satellite: The Capacity Leap Brought by V3
According to documents SpaceX submitted to regulators, V3 satellites are designed with downlink capacity of about 1 Tbps, equivalent to each satellite supporting roughly 10,000 households simultaneously running full 100 Mbps broadband, with uplink capacity of 160 Gbps; compared with V2, downlink is about 10 times higher and uplink about 22 times higher, with antennas supporting 2,048 uplink and downlink beams (V2's phased array had 192 downlink and 144 uplink). According to SpaceX's technical notes, a single Starship launch can deliver into the constellation roughly 20 times the capacity of a Falcon 9 launching V2 satellites. Weight is another threshold: a single V3 weighs about 2,000 kilograms, roughly double the previous generation, and Falcon 9 cannot lift it.
Third-party commentary is more optimistic, though these are personal statements: First Principles Group managing partner Rob Maurer said on X that "with just today's single Starship launch, the Starlink capacity SpaceX added in the past hour exceeded the total from Starlink's entire first year and a half of launches"; former Apple executive and current Rivian senior director Phil Beisel said a single V3 is designed for 1.16 Tbps of user capacity and 2.4 Tbps of inter-satellite optical links, which could ultimately form a network backbone capable of routing more traffic and even connecting computing resources to space. Musk wants Starship to launch 60 V3 satellites per flight once it enters regular operations; by estimates, for Starlink service overall to improve noticeably, about 1,000 V3 satellites need to be in orbit, requiring at least 17 more launches at full capacity. SpaceX has applied to the U.S. Federal Communications Commission (FCC) to operate up to 100,000 V3 satellites; as of September 2026, Starlink had launched about 13,000 satellites cumulatively, with about 11,000 in orbit, serving tens of millions of users across more than 160 countries and regions.
Wall Street Raises Prices: Compute Is the Near-Term Engine
TD Cowen, led by analyst John Blackledge, initiated coverage with a "Buy" rating and a $200 price target, calling SpaceX's broader opportunities in AI and space "enormous." The firm estimates AI compute leasing will contribute about 60% of revenue by 2027, with revenue growing at a 62% compound annual rate from 2026 to 2031, and low-Earth-orbit launches potentially approaching 1,000 by 2031. TD Cowen also expects about 35% of 2026 sales to come from compute-related business, which could surpass Starlink—the "crown jewel"—as early as the first quarter of 2027, account for 65% of overall revenue by 2028, and see nearly half of planned compute capacity over the coming years leased to outside companies. Deutsche Bank analyst Edison Yu gave a "Buy" rating and a $235 price target, estimating that five signed customers correspond to about $54.5 billion in revenue run-rate, saying that as capacity comes online "several more large orders are expected to land." Other institutional moves the same day: RBC reiterated "Outperform" and a $225 price target, calling the 14th flight an important milestone for Starship's launch capability and V3 deployment; Clear Street maintained "Buy" and a $217 price target, believing the mission advanced the company's ability to expand its high-capacity constellation, though engine problems were a reminder that execution risk remains; CLSA initiated coverage with "Accumulate" and a $250 price target; Bernstein SocGen maintained "Buy" with a $248 price target. According to Tipranks data, the average analyst price target is $233.
The details of the compute contracts are the key to understanding this round of expectations. Anthropic pays $1.25 billion per month to lease compute at SpaceX's Colossus data center in Memphis; Google Cloud signed a compute agreement worth $920 million per month, effective next month; Reflection AI pays $150 million per month; and another unnamed customer will pay $1.11 billion per month starting in December—totaling about $3.4 billion per month, or roughly $40.8 billion annualized, if fully performed. However, only three of the five disclosed customers have had their identities publicly confirmed, and the fourth, disclosed in July, has been speculated by some analysts to be the U.S. Department of Defense; these leases also include "unusually short" 90-day termination clauses, making it relatively easy for customers to exit.
The parent company's results provide a floor. According to an earnings summary, SpaceX's second-quarter revenue (released August 4) was $7.8 billion, up 92% year over year and above the market's $6.81 billion estimate, with adjusted EBITDA of $3.5 billion, up 191%, and a net loss of $541 million, including AI segment revenue of $2.6 billion, up 247%. Chief Financial Officer Bret Johnsen said the company is on track to reach a $100 billion annual recurring revenue (ARR) run-rate by the end of 2026; management plans compute infrastructure scale of 15 to 20 gigawatts by the end of 2027, while Musk moved the $1 trillion annual revenue target forward from 2031 to 2030.
The source of this compute narrative is the February merger this year: SpaceX acquired xAI in an all-stock deal (consideration of about $250 billion, with a combined valuation of about $1.25 trillion), the AI business was officially renamed SpaceXAI in July, and Colossus thereby became an asset of the public company. Musk explained the merger logic very bluntly—orbital data centers: the company has applied to the FCC to operate up to 1 million satellites for AI compute in orbit and plans to begin launching them with Starship as early as the end of next year.
The Other Side: A Roughly $1.9 Trillion Market Cap, 20 Times Price-to-Sales, and a Wave of Lockup Expirations
The other side of the stock price is valuation. Based on the September 28 closing price, SpaceX's total market capitalization is about $1.92 trillion; by the original account, the stock has fallen 10% year to date but is still up 1% for the month. By calculations, 2025 revenue was about $18.7 billion and the company is still losing money, while Wall Street expects that to jump to $44.8 billion and $108.3 billion in the next two years respectively—even if all of that is achieved, the price-to-sales ratio based on 2027 expected revenue would still be close to 20 times, which is not cheap for a capital-intensive company, with a forward price-to-earnings ratio of about 204 times. "Big Short" investor Michael Burry said at the time of the company's listing that there was nothing in the prospectus to support even a $1 trillion valuation, and later described it as "a small space business, a niche telecom business, a struggling social media platform, plus a 'lite version of CoreWeave.'"
Share supply is a nearer-term problem: the company issued only about 4% of shares in its IPO, and lockup periods have been expiring progressively—one batch unlocked on September 24, up to 328.4 million shares each could become tradable on October 9 and October 24, and as many as 1.3 billion shares could flood the market after the third-quarter report; Musk himself is locked up until June next year, but employees and institutional investors are not. Financing costs are also rising: according to the same report, the company issued $25 billion in bonds shortly after listing, with coupons of 5.35% to 6.65%; against the backdrop of the Federal Reserve having just raised rates and hinting at possible further hikes, if more debt is issued to finance the next round of compute and launch site construction, the path to profitability will be tighter.
Retail sentiment, meanwhile, has recovered somewhat. On the Stocktwits platform, SPCX's retail sentiment index rebounded from "bearish" the previous day to "neutral," with 24-hour message volume up 74%; one long-term holder commented that "the money will come, all the major bank analysts have confirmed it," while another user emphasized that the flames and water fireball during reentry were "completely normal for this stage of testing" and called the 14th flight a "historic success."
Falling-Behind Rivals, and NASA's Bet on Starship for the Moon Landing
The situation of competitors has instead highlighted SpaceX's position. Blue Origin's New Glenn made its first flight in January this year, but a mission in May failed during the powered landing phase, destroying the rocket and its only launch pad, with the return-to-flight target set for before year-end; United Launch Alliance's Vulcan has been grounded since February due to a booster anomaly. Amazon's Leo constellation's most recent successful deployment was still on July 2, with 396 satellites launched cumulatively, a clear gap versus Starlink's more than 11,000 in orbit, and its target of starting commercial service in mid-2026 has already passed. More critically, there is NASA's dependence. According to industry media estimates, Starship is the only crewed landing system for the Artemis lunar program, and that mission requires 10 to 20 "tanker Starship" flights to transfer cryogenic methane and liquid oxygen in orbit, a process with no precedent to date; SpaceX has twice postponed the in-orbit propellant transfer demonstration, with the latest target being "no earlier than the end of 2026," while Artemis III plans an uncrewed lander Earth-orbit demonstration by the end of 2027 and Artemis IV plans the first crewed lunar landing in early 2028, both depending on the success of that demonstration.
Back to Musk's own timeline: he wrote on Sunday that "Starship is 2 to 3 years away from one flight per hour"; the longer-term goal is 10,000 launches per year by 2030—for reference, Falcon 9 flew 165 times in 2025, already a company record and roughly half of that year's global orbital launches. TD Cowen's forecast of nearly 1,000 low-Earth-orbit launches by 2031 falls roughly in the range of "less than 3 per day." The results of the engine anomaly investigation will determine whether the 15th flight can attempt to catch the spacecraft with the launch tower's mechanical arms as originally planned.
Comments