SpaceX conference call: Elon Musk stated that ARR is expected to reach 1 trillion US dollars next year, and the payback period of computing power may be as short as 6 months.

SpaceX delivered a blowout performance in the third quarter. Elon Musk not only unveiled a violent AI computing power expansion plan that guarantees a minimum of 13GW in 2027 and even targets 20GW, but also boldly stated that the company's annual recurring revenue will sprint to the trillion-dollar mark by the end of next year.

Wall Street CN noted that after U.S. markets closed on Tuesday, October 20, Eastern Time, SpaceX announced that in the third quarter of 2026 ending in September, the company's revenue rose 161% year-over-year to $14.58 billion, more than 10% higher than analysts' consensus estimate of roughly $12.8 billion.

Behind the explosive revenue growth lies the skyrocketing expansion of the super-intelligent cloud service business. The super-intelligent business posted quarterly revenue of $7.6 billion, surging $5 billion quarter-over-quarter and $6.7 billion year-over-year. Meanwhile, net profit turned positive to $4.4 billion from a loss of $1.1 billion in the same period last year, and adjusted EBITDA reached $8.8 billion, up 306% year-over-year.

During the earnings call, company CEO Elon Musk said market demand for super-intelligent computing power is extremely robust, the payback period for supercomputing infrastructure deployment costs has been compressed to less than one year, and short-term contract renewal prices continue to trend upward.

Musk further predicted that the company's 2027 annual recurring revenue (ARR) is expected to land in the range of $500 billion to $1 trillion.

Musk stated that 2.5 to 3GW of computing power will be put into operation by the end of 2026, SpaceX is highly confident of getting 10GW of computing power online in 2027, adding that "I would be surprised if we have less than 13GW of online computing power by the end of next year."

To support the above expansion, SpaceX plans to add an additional 20GW of power supply and cooling infrastructure in 2027. Musk said SpaceX's speed of deploying power, cooling and supporting infrastructure is expected to outpace the sum of all other manufacturers.

Regarding rocket launch and Starlink data, SpaceX COO Gwynne Shotwell said during the call that the company is on track to achieve reusable second-stage rocket launches early next year. Shotwell noted that output of Starlink aviation terminals is expected to double before the end of this year.

Shotwell revealed that monthly active users of Starlink's direct-to-mobile service have exceeded 22 million. The spectrum acquisition deal will make Starlink a major U.S. mobile operator. Both Starlink's enterprise business revenue and government business revenue are expected to exceed consumer business revenue in the future. Starlink is poised to carry a significant share of global internet traffic.

01

Better-than-expected computing power stack and Nvidia orders

Previously, the market widely underestimated the ambition of Musk-affiliated companies in AI computing power. During this meeting, Musk laid out a stunning computing power deployment timeline, sharply raising previous computing power targets.

Musk stated bluntly during the earnings call:

People are starting to realize that the constraints on deploying superintelligence lie more in power, cooling and infrastructure than in chips.

He said SpaceX has transferred its top-tier team that previously captured nearly 90% of the global orbital launch mass share in the aerospace sector to ground computing power deployment. Musk added:

It's a bit like the New York Yankees going to play Little League. That's not fair. But the end result is that we deploy power and cooling infrastructure much faster than anyone else, possibly even faster than everyone else combined. I'll let that sink in for a moment.

On specific Nvidia GPU orders and delivery timelines, the incremental figures are extremely staggering. Musk revealed that the team is currently working around the clock:

Our current progress is that in each remaining month of this year, we will bring roughly 200,000 GB300 units online each month, and maintain this pace next year, with the main deployment focus shifting to Vera Rubin next year.

For total future computing power capacity, Musk offered an extremely aggressive performance guidance:

By the end of this year, we will have roughly 2.5 to 3GW of computing power online. I am very confident we will add 10GW of new computing power in 2027. In other words, I will be surprised if we have less than 13GW of online computing power by the end of next year.

What has drawn even more market attention is that he left huge room for imagination:

To ensure power and cooling do not become bottlenecks, our actual target for next year is to add 20GW of new power and cooling infrastructure... So SpaceX could have more than 20GW of superintelligent computing power online next year.

In addition, he confirmed that by the end of the second quarter of next year, the most powerful training cluster consisting of 440,000 Vera Rubin computers will be launched to support the Grok large language model.

02

Ultra-fast payback in as short as 6 months and computing power shortage

SpaceX expects Q4 capital expenditure to reach $35 billion to $40 billion. Faced with massive capital expenditure, the market is highly focused on the sustainability of its commercial monetization. SpaceX management delivered extremely optimistic feedback.

CFO Bret Johnsen noted that current market conditions are "unprecedented", at current pricing levels, the payback period for new computing power deployment is less than one year. Musk's statement was even more aggressive:

I would say it's not impossible that the payback period for our computing power could shrink to 6 months next year.

Musk firmly believes that the value per watt of computing power will rise as AI models are optimized, so SpaceX prefers to sign short-term lease contracts at spot market rates rather than long-term binding agreements. He used the iconic "Cheese Shop" sketch from the famous British comedy troupe Monty Python to describe the industry-wide hunger for computing power:

We are literally hounded nonstop by every hyperscale cloud provider and dozens of smaller companies to lease more of our computing power. If we said "yes" to everyone, we could lease our computing power out several times over. It's like that guy who goes to a cheese shop to buy cheese... We tried to outsource some of our inference computing power to other cloud service providers, and we ran straight into the "cheese shop scenario". We asked "Do you have any GB300s?" No. "What about GB200s?" No. "TPUs?" No. "H100s?" No, none either. It's absolutely absurd.

Backed by high-margin cloud service agreements, SpaceX's Superintelligence (SI) division generated $7.6 billion in revenue in the third quarter, surging $5 billion quarter-over-quarter. The company is on track to hit its target of $100 billion in annual recurring revenue (ARR) for the business by the end of this year. For the longer term, Musk offered a stunning forecast:

Put simply, that means SpaceX's annual recurring revenue (ARR) will be between $500 billion and $1 trillion by the end of 2027.

03

Fully reusable Starship breakthrough and trillion-dollar ecosystem moat

On the traditional space business front, SpaceX is also nurturing tremendous fundamental qualitative change.

Musk emphasized that Starship's 15th flight will take place next month, marking the first attempt to "simultaneously recover the booster and the spacecraft" during an orbital launch. He said:

This is a far-reaching technological milestone in the history of human civilization... If the 15th flight does not succeed, we will eventually get there.

SpaceX expects Starship to achieve weekly launches in the first half of next year, and two launches per week or even higher frequency in the second half of the year.

The maturity of Starship directly endows the Starlink network with disruptive competitive power. The recently launched V3 satellite deployment in a single mission delivers bandwidth equivalent to 10 Falcon 9 launches.

Musk predicts that Starlink's bandwidth will reach roughly 100 times current levels within three years, and roughly 100,000 new satellites with more than 10 times the bandwidth of current units will be launched to meet the "exponential communication bandwidth demand brought by the explosion of AI agents".

In addition, regarding the widely discussed "in-house developed chips (TerraFab)" initiative, Musk confirmed that the internal goal is to produce the first functional test chip at the Austin lab in May or June next year, and aims to achieve mass production at the main TerraFab plant in the third quarter of 2028, with the ultimate goal of achieving 1 terawatt of annual superintelligent chip production capacity.

04

Musk recuses himself from SpaceX-Tesla merger discussions due to conflict of interest

Regarding a potential merger between his two companies, Musk said SpaceX and Tesla have formed special committees to explore the possibility of a combination. Media outlets note that if the merger goes through, it will become the largest corporate merger in history.

Musk stated:

While there is no guarantee of a deal at this point, I believe merging the two companies creates tremendous value, as we are at the cusp of the superintelligence singularity.

Musk said that due to his own conflict of interest, he is not participating in the committee's discussions; currently, the committees have not reached agreement on merger terms.

Musk cautioned that the merger still faces uncertainties and must win shareholder support. He noted that Tesla needs approval from 50% of all its shareholders, not just a majority of those who cast votes. He said:

Some people think the SpaceX-Tesla merger is a done deal. I think that statement is overstating things.

Full text of SpaceX earnings call (AI-assisted translation):

Head of Investor Relations Andrea Williams:

Hello everyone, welcome to our Starlink factory in Bastrop, Texas, which will also serve as the future production base for Starmind. I am Andrea Williams, Head of Investor Relations. Thank you all for joining SpaceX's 2026 third quarter earnings conference.

Joining us today are Elon, Gwynne and Bret. This conference is being streamed live simultaneously on the X platform. Our third quarter financial results were officially released after 3 p.m. Texas time, and all related materials have been posted to our investor relations website ir.spacex.com.

This earnings event will be split into two parts: first, prepared remarks, followed by a Q&A session. We will first take live questions from sell-side analysts, then answer questions from investors submitted via the dedicated earnings Q&A platform built by the SpaceX team in partnership with Brock.

During this webcast, we will make forward-looking statements about the company's business prospects, all based on our current forecasts and expectations. Actual events or performance results may differ materially from expectations due to various risks and uncertainties, including those risk factors referenced in our latest earnings releases and filings submitted to the SEC (including the final prospectus). During the presentation, we may discuss several non-GAAP financial measures. For related disclosures and reconciliations to comparable GAAP measures, please refer to today's earnings release and SEC filings.

Below are the 2026 third quarter financial results (compared to the same period last year):

Operating revenue: $14.6 billion, up 161% year-over-year (prior year period: $5.6 billion); Net profit: $4.4 billion, an improvement of $5.5 billion year-over-year (prior year period: net loss of $1.1 billion); Adjusted EBITDA: $8.8 billion, up 306% year-over-year (prior year period: $2.2 billion).

Now I will pass it over to Elon for remarks.

Co-founder and CEO Elon Musk: Thank you.

Right now, SpaceX is hitting important inflection points across three domains: launch, connectivity, and superintelligence.

On the Starship front, the 14th flight was a critical milestone — Sasha reached orbit for the first time and successfully deployed operational Starlink V3 satellites. We will play a recap video later. The 15th flight is scheduled for next month, where we will attempt to recover both the booster and the ship simultaneously. If successful, this will mark the first full recovery in the history of orbital launch vehicles, a major technological milestone.

Fully reusable orbital launch vehicles are the only path to enabling sustained extension of human consciousness beyond Earth. That's why I consider this a far-reaching technological milestone in the history of human civilization. We will now play an animation simulating the 15th flight to demonstrate how the mission would proceed in a successful scenario.

Even if the 15th flight is not successful, we will keep pressing on. There is no 100% guarantee of success on the first few flights, but I think the odds of success for the next flight are better than 50%. If it does not succeed, we will continue scheduling Starship flight attempts before the end of this year, and further increase the flight frequency next year.

We also announced development plans for our Louisiana space launch site. This will be an extremely valuable asset for the United States and the whole world, expected to eventually host roughly 20 or more launch towers, capable of transporting millions of tons of payload to orbit and beyond each year, which is of extraordinary significance.

We expect Starship to achieve weekly flights as early as the first half of next year, two flights per week in the second half of next year, and even higher frequencies by the end of next year. With the breakthrough of rapid full reusability technology, Starship launch frequency will rise dramatically.

Once full reusability is achieved, there will initially be a small amount of maintenance work required between each flight, but we will gradually eliminate the need for maintenance, and eventually enable the rocket to launch again immediately after landing, just like an airplane. As I said before, this is a far-reaching technological milestone, even though most people have not realized its importance yet, in hindsight it will be obvious.

On the Starlink front, we are also just getting started. The 14th flight deployed 26 Tbps of Starlink bandwidth, equivalent to the total bandwidth of 10 Falcon 9 launches. That means a single Starship flight — even at partial payload capacity — delivered the equivalent of 10 Falcon 9 launches.

I expect that before too long, within roughly three years, the bandwidth of our Starlink constellation will be roughly 100 times current levels — because we will launch satellites with an order of magnitude higher bandwidth, while also increasing the total number of Starlink satellites by an order of magnitude to roughly 100,000, each with more than 10 times the bandwidth of existing Starlink systems. Starlink's revenue will grow roughly proportionally to total bandwidth delivered.

We will also have high-bandwidth direct-to-cell capability, and will build a ground network to provide ubiquitous connectivity across all regions where SpaceX and Starlink are authorized to operate. I believe that in the next 3 to 5 years, bandwidth demand will far exceed today's levels — possibly 1000 times current demand. The reason is that AI agents will consume far more bandwidth than individual human users, and the number of agents will also far outnumber the human population, requiring several orders of magnitude improvement in communication bandwidth to meet massive connectivity needs.

On the superintelligence front, I call this SpaceX SI (SpaceX Superintelligence). People are gradually realizing that the bottleneck restricting superintelligence deployment lies more in power, cooling and infrastructure, not in the chips themselves.

SpaceX has an exceptional team that already captures nearly 90% of Earth's orbital launch mass share, with China accounting for roughly 6%, and the rest of the world (including other U.S. players) combining for roughly 4%. The achievements of this team outperform all our space rivals combined by an order of magnitude. Now we are applying that same team to ground computing power deployment. It's like the New York Yankees going to play minor league baseball, the outcome is barely in doubt. And the end result is that we can deploy power and cooling infrastructure far faster than any other company, and in some cases possibly even faster than all other companies combined. Let that sink in for a moment.

I personally have spent a lot of time in Memphis recently, often four days a week, working around the clock to bring the Memphis supercomputing cluster online. Our plan is to steadily bring roughly 200,000 GV300 units online each month