Musk's “fully reusable rocket” ambitions welcome verification! SpaceX (SPCX.US) Starship's test flight tomorrow challenges “double recycling”, a key step for large-scale launch data centers

The Zhitong Finance App learned that the 15th SpaceX (SPCX.US) Starship test flight will open the window at 06:45 Beijing time on November 14. The window will be 75 minutes long. Simply put, SpaceX's goal this time is to land the two parts of the starship intact on Earth in its latest mission. This is a test of reusable rocket technology. If successful, it will provide an important basis for Starship (Starship)'s proof of concept, which Elon Musk promised would be the world's first fully reusable rocket.

Level 2 recycling is an economic problem that has been calculated for ten years

The mission had two goals: the super-heavy booster returned to the launch site about 6 and a half minutes after takeoff and was caught by the launch tower's robotic arm; the spacecraft entered orbit, circled the Earth for about 14 hours, and then deorbited and re-entered in the same way, and was caught by the same robotic arm in the same way — the latter was an unprecedented experiment. If both stages are successful, this will be the first time that an orbital stage launch system has been fully recycled.

The reusable competition of the past ten years has only been half done. Falcon 9 recovers a sub-stage — the section that contains the engine and fuel and is responsible for launching the rocket into the thickest part of the atmosphere. After the second sub-stage put the satellite and crew into orbit, they were abandoned as usual.

The reason isn't that technology can't do it, it's that it's not cost-effective. If you want to recycle the Falcon 9's second sub-stage, you have to install heavy insulation, landing legs, grille wings, and landing fuel, all of which directly eat up the capacity — and the cost of the second sub-stage itself is not high compared to the first sub-stage.

Starship turned this account over. Its size and capacity are large enough, and the loss of capacity caused by installing a recycling system can be covered by manufacturing costs saved by “full reuse”.

Boundaries need to be explained: Falcon 9 is partially reused, and the space shuttle's external storage tank is also disposable; what Starship needs to do is return the entire rocket to the launch pad, refuel, and fly again.

The mission used the B22 booster and the S42 spacecraft, both in Block 3 configuration. The booster was the V3 version of the first attempt to capture — the previous three successful captures were all using a V1 booster, two of which have completed the return flight. According to the process, about 6 minutes and 35 seconds after takeoff, the booster was retracted and captured by a robotic arm, then it must be moved immediately to free up space for the spacecraft to work.

The spacecraft's path is much more complicated: it enters an orbit about 275 kilometers high, orbits about 14 times, glides in orbit for about 21 hours waiting for the orbital phase to line up with the star base, then deorbiting, brakes, and then re-enters. This section has to face higher heat flow and structural load due to orbital speed re-entry, thermal protection, reentry attitude, black barrier zone communication, final landing ignition, and approach accuracy. If there is a problem with any ring, the capture must be abandoned.

As a result, SpaceX has set a safety threshold: return capture will only be carried out if it is confirmed that all the systems required for re-entry maneuver, landing ignition, and approach the launch tower are healthy; otherwise, the spacecraft will splash in pre-set waters off the Pacific Ocean near Hawaii or the Chilean coast as planned. This coincided with the outcome of the 14th test flight — the September 28 mission first got into orbit and ended up splashing into the North Pacific about 3 hours after takeoff.

In the latest outlook report, Morgan Stanley gave a success rate of “less than 50%”. The reason was not that it was technically impossible, but rather an execution threshold — the team wouldn't initiate a capture unless all of the hardware was in near perfect condition. However, stopping the capture does not mean that the mission has failed. If the original test session is aborted, Flight 16 will be tried again during the year. Ship 43 and Booster 23 have completed low temperature tests and are awaiting engine installation.

The real bottleneck is not the “catch” that the market is most concerned about

The robotic arm captures the strongest visual impact, but when it comes to “quick reuse,” the card points actually protect against heat. When the starship re-enters, the surface temperature of the arrow body is extremely high. Once the insulation tiles fall off, the steel structure is directly exposed. The insulation cover of the 15th flight followed many of the improvements from the 14th: adding fixing mechanisms in the areas where the tiles are most likely to fall, handling channels where plasma may flow into the tile gaps, and using curved tiles on a large area to reduce heating between the tiles.

What's more ingenious is the verification method: Of the 26 Starlink V3 satellites deployed this time, 3 have been modified and equipped with multi-angle cameras to scan the arrow insulation tiles in orbit and return images for the ground to determine the viability of return flight — the “satellite performs a physical examination of the rocket” has already been tested in previous missions.

Damo said that what you should pay more attention to than putting the robotic arm together is the V3 engine. Flight 12 and Flight 14 each lost a Raptor Vacuum. Musk attributed Flight 14 to a software failure unique to the engine; yet so far, every V3 booster lost its engine before it was splashed, and Flight 14 lost 2. The B22 happened to be the first V3 booster to be captured by the tower arm, and it also landed on the first capture on launch pad 2. The two pieces were crushed “for the first time” in the same mission. According to the logic of the Damo report, only one engine is dropped after takeoff, and the team is likely to abandon capture and switch to a soft splash; once an error occurs during the capture phase, the cost is more than one arrow — it may damage the ground facilities at the star base, and cause internal and regulatory authorities to question the pace of subsequent launches at the same time. This is much more expensive than “mission failure,” and it is also the most realistic part of “less than 50%.”

Even if the capture is successful, the economy of full reuse will have to pass the hurdle of refurbishment. Musk told investor Ron Barron that when he first proposed this idea, “they looked at me like I was crazy”; his response was that from a physical point of view, nothing stopped this from happening. But physics allows it, doesn't mean that the project is cheap — inspection and repair time between flights is the real threshold for “operating like an airplane.”

Why does it have to be done now?

Falcon 9 won't work because of the launch list at the back.

The 26 Starlink V3 satellites carried this time can add 1 Tbps of capacity to the constellation. The total of 26 satellites is equivalent to the capacity of the V2 Mini launched by Falcon 9 10 times. This is the second batch of upgraded Starlinks, the same 26 from the previous batch, which was put into orbit by the 14th test flight in September.

The bigger bill is behind. SpaceX plans to use Starships to deploy up to 100,000 upgraded Starlink satellites and up to 1 million spacecraft for the Starmind orbital data center. Starmind applied to the FCC in January and named in June. The first AI1 hardware has an average of 120 kilowatts of computing power. It was developed in cooperation with Nvidia based on the Vera Rubin NVL72 architecture, and the first batch is scheduled to be launched in the 3rd quarter of 2027.

These numbers are almost demanding on the frequency of transmission. Musk said that the company's goal is to make 10,000 Starship flights per year by 2030; while the external model assumes 250 flights in 2028, compared to plans of 2 this year and 38 next year, this is an order of magnitude jump. Without full multiplexing, this slope is unlikely to occur.

A big test after listing

This Starship test flight is also SpaceX's first major technical test in the true sense of the word since its launch.

The company was listed on the NASDAQ on June 12, raising basic capital of about $75 billion (including overallotment of about 86 billion dollars), making it the largest IPO in history; the listing valuation is about 1.77 trillion US dollars. Since then, the stock price has fluctuated greatly; it has entered the NASDAQ 100, but not the S&P 500.

The fundamentals are still “walking on one leg”: revenue of $18.674 billion and net loss of $4.937 billion in 2025; revenue of $7.814 billion in the second quarter of 2026 was $7.814 billion, up 92% year over year, and net loss narrowed to $541 million. Starlink is the only profitable sector, with over 12 million active subscriptions at the end of the second quarter. Both launch and AI infrastructure are still being invested, and the growth slope for both depends on Starship reuse.

Restraint in judgment: This success is not equal to commercialization

First, this is still a high-risk test. SpaceX's official website states that “awaiting regulatory approval”, capture failure, mission suspension, and abandonment of return capture are all within the plan; one success only proves that the route is viable; it does not prove that the turnover efficiency is up to standard.

Second, the economy of full reuse depends on the cost of refurbishment, and there is no public data on the cost of refurbishment so far. Direct equating “full recycling” with “a sharp drop in launch costs” is a media narrative rather than a company disclosure.

Third, the dispute over the cost of space computing power is far from being resolved. Google's data center energy director publicly stated in November that a space data center could cost up to 100 times that of the ground. Starmind's schedule is also just Musk's “prediction based on current circumstances.”

For a company with a market capitalization of 1.9 trillion dollars, tonight's real highlight is not the moment the robotic arm is assembled, but the next few months: whether this spacecraft can take off again at an acceptable time and cost. That is the “full reuse” form of redemption.

Damo pointed out that the capture was only an entrance ticket. The report placed the real proof point on reuse: the spacecraft only flew once again after experiencing the thermal and mechanical loads re-entering orbit before it was considered a “complete and rapid reuse”, as early as early as the beginning of next year; the tower arm clamping is a demonstration of ability, and the resumption of flight is the starting point of the cost curve. The numbers on the scale side can be used as a reference: Each time the Starship launches Starlink, it is about equal to the launch volume of Falcon 9 for at least one month; carrying 26 Starlink V3s this time is about the capacity of Falcon 9 10 times.