SpaceX's 15th Starship test flight will open its window at 06:45 Beijing time on November 14, with the window lasting 75 minutes.
The goal of this mission is to bring both parts of Starship back to Earth intact, a test of reusable rocket technology. If successful, it would provide important evidence for the Starship concept, which Elon Musk has promised would be the world's first fully reusable rocket.
Second-Stage Recovery: An Economic Problem a Decade in the Making
This mission has two objectives: the Super Heavy booster returns to the launch site about six and a half minutes after liftoff and is caught by the launch tower's mechanical arms; the ship then enters orbit, circles the Earth about 14 times, flies for more than 22 hours, deorbits and reenters, and is caught in the same way by the same pair of arms — the latter being an unprecedented attempt.
If both stages succeed, it would be the first complete recovery of an orbital-class launch system.
The reusable race over the past decade only did half the job. Falcon 9 recovers the first stage — the section carrying the engines and fuel that pushes the rocket through the thickest part of the atmosphere. The second stage, after delivering satellites and crew to orbit, is routinely discarded.
The reason is not that it is technically impossible, but that it is not cost-effective. Recovering Falcon 9's second stage would require adding a heavy heat shield, landing legs, grid fins, and landing fuel, all of which directly eat into payload capacity — and the second stage itself is not expensive relative to the first stage.
Starship flips that calculation. Its size and capacity are large enough that the payload penalty from adding a recovery system can be offset by the manufacturing cost savings from full reuse.
The boundary must be stated: Falcon 9 is partially reusable, and the space shuttle's external tank was also single-use; what Starship aims to do is return the entire rocket to the launch pad, refuel, and fly again.
This mission uses Booster B22 and Ship S42, both of the Block 3 configuration. The booster is the V3 version making its first capture attempt — the previous three successful captures used V1 boosters, two of which have already completed reflights.
In terms of sequence, about 6 minutes and 35 seconds after liftoff the booster turns back and is captured by the mechanical arms, after which it must be moved immediately to free up workspace for the ship.
The ship's path is far more complex: entering an orbit about 275 kilometers high, circling the Earth about 14 times, coasting in orbit for about 21 hours while waiting for the orbital phase to align with Starbase, then performing a deorbit burn and reentry.
This phase must withstand higher heat flux and structural loads from reentry at orbital velocity, and if any link fails — thermal protection, reentry attitude, communications through the blackout zone, the final landing burn, or approach precision — the catch must be abandoned.
SpaceX has therefore set a safety threshold: only if all systems required for the reentry maneuver, landing burn, and approach to the launch tower are confirmed healthy will it proceed with the return catch; otherwise the ship will splash down as planned in a preset area of the Pacific near Hawaii or off the coast of Chile.
This is consistent with the outcome of the 14th test flight — that mission on September 28 reached orbit for the first time and ultimately splashed down in the North Pacific about three hours after liftoff.
In its latest outlook report, Morgan Stanley put the success rate at "less than 50%," not because it is technically impossible but because of the execution threshold — unless all hardware conditions are near-perfect, the team will not initiate the catch.
However, a catch abort does not equal mission failure; aborts are a normal part of testing, Flight 16 will retry within the year, and Ship 43 and Booster 23 have completed cryogenic testing and are awaiting engine installation.
The Real Bottleneck Is Not the Market's Favorite "Clamp"
The mechanical arm catch has the strongest visual impact, but for "rapid reuse," the actual sticking point is thermal protection. During Starship's reentry, the vehicle surface reaches extremely high temperatures, and once heat shield tiles fall off, the steel structure is directly exposed.
Flight 15's heat shield follows several improvements from Flight 14: adding fastening mechanisms in the areas most prone to tile loss, addressing channels through which plasma could penetrate tile gaps, and widely adopting curved tiles that reduce heating between tile gaps.
Even more clever is the verification method: among the 26 Starlink V3 satellites deployed on this flight, 3 have been modified to carry multi-angle cameras that scan the vehicle's heat shield tiles in orbit and transmit images back to the ground for use in judging return feasibility — "satellites giving the rocket a health check" has already been tested on previous missions.
Morgan Stanley says that what deserves more attention than the arms closing is the V3 engine. Flight 12 and Flight 14 each lost one Raptor Vacuum, and Musk attributed the Flight 14 loss to a software fault specific to that engine; and so far every V3 booster has lost an engine before splashdown, with Flight 14 losing 2.
B22 happens to be the first V3 booster caught by the tower arms, and it also falls on Pad 2's first catch, stacking two "firsts" into the same mission.
According to the logic of Morgan Stanley's report, if even one engine is lost after liftoff, the team will most likely abandon the catch and switch to a soft splashdown; and once an anomaly occurs during the catch phase, the cost is more than one rocket — it could damage Starbase ground facilities and cause both internal and regulatory doubts about the pace of subsequent launches.
That is far more expensive than a "mission failure," and it is the most concrete part of the "less than 50%" estimate.
Even if the catch succeeds, the economics of full reuse still have to pass the refurbishment hurdle. Musk told investor Ron Baron that when he first proposed the idea, "they looked at me like I was crazy"; his response was that from a physics standpoint, nothing prevents it from happening.
But physics allowing it does not mean engineering makes it cheap — the inspection and refurbishment time between two flights is the real threshold for "operating like an airplane."
Why It Must Be Done Now
Because the launch manifest ahead cannot be carried by Falcon 9.
The 26 Starlink V3 satellites carried on this flight can each add 1 Tbps of capacity to the constellation, and 26 together roughly equal the capacity of 10 Falcon 9 launches of V2 Mini. This is the second batch of upgraded Starlink satellites; the previous batch, also 26, was delivered to orbit by the 14th test flight in September.
A much larger bill comes later. SpaceX plans to use Starship to deploy up to about 100,000 upgraded Starlink satellites, as well as up to 1 million spacecraft for the Starmind orbital data center.
Starmind applied to the FCC in January and was named in June; its first-generation AI1 hardware averages 120 kilowatts of computing power per unit, was developed in cooperation with Nvidia based on the Vera Rubin NVL72 architecture, and the first batch is planned for launch starting in the third quarter of 2027.
These numbers impose almost brutal requirements on launch frequency. Musk has said the company's goal is 10,000 Starship flights per year by 2030; external models assume 250 flights will be needed in 2028, compared with this year's plan of 2 and next year's 38, this is an order-of-magnitude leap.
Without full reuse, that slope is impossible.
A Major Test After Going Public
This Starship test flight is also SpaceX's first real technical test since going public.
The company listed on Nasdaq on June 12, raising about $75 billion in base proceeds (about $86 billion including the over-allotment), the largest IPO in history; its listing valuation was about $1.77 trillion.
The stock has fluctuated sharply since then; it has entered the Nasdaq 100 but not the S&P 500.
Fundamentals remain "walking on one leg": 2025 revenue was $18.674 billion and net loss was $4.937 billion; second-quarter 2026 revenue was $7.814 billion, up 92% year over year, with the net loss narrowing to $541 million.
Starlink is the only profitable segment, with more than 12 million active subscriptions at the end of the second quarter.
Launch and AI infrastructure are still in the investment phase, and the growth slope of both depends on Starship reuse.
A Restrained Judgment: Success This Time Does Not Equal Commercialization
First, this is still a high-risk test. SpaceX's official website states "pending regulatory approval," and a failed catch, mission abort, or abandoning the return catch are all within the plan; one success only proves the route is feasible, not that turnaround efficiency meets the standard.
Second, the economics of full reuse depend on refurbishment cost, and there is still no public data on refurbishment cost. Equating "whole-rocket recovery" directly with "a sharp drop in launch cost" is media narrative, not company disclosure.
Third, the cost debate over space computing is far from settled. Google's data center energy head said publicly in November that space data centers could cost up to 100 times as much as ground-based ones. Starmind's timeline is also only Musk's "prediction based on current circumstances."
For a company valued at $1.9 trillion, the real focus tonight is not the second the arms close, but the next few months: whether this ship can fly again at an acceptable time and cost. That is how "full reuse" will be redeemed.
Morgan Stanley notes that the catch is only the ticket for entry, and the report places the real proof point on reuse: only after the ship endures the thermal and mechanical loads of orbital reentry and flies again will "complete and rapid reuse" be considered proven, at the earliest early next year; the tower arms catching it is a capability demonstration, while reflight is the starting point of the cost curve.
The scale figures can serve as a reference: each Starship launch of Starlink is roughly equal to at least a month of Falcon 9 launches; this flight carries 26 Starlink V3 satellites, roughly equivalent to the capacity of 10 Falcon 9 launches.
Comments