Musk says AI agents could drive a 1,000x jump in bandwidth demand as SpaceX recasts Starlink’s future

Musk says AI agents could drive a 1,000x jump in bandwidth demand as SpaceX recasts Starlink’s future

N
News Editor
2026-10-25 19:03:38
SpaceX used its Oct. 20 third-quarter earnings call to frame Starlink as more than a satellite internet service for rural users, aircraft, ships, or military communications. CEO Elon Musk said global bandwidth demand could rise 1,000x over the next three to five years, driven by AI agents that use far more bandwidth than humans and could eventually outnumber them. He also said Starlink’s capacity could expand 100x within about three years, helped by a larger satellite constellation and higher-capacity satellites launched by Starship. The company’s latest numbers show why SpaceX is leaning into that narrative. As of Sept. 30, Starlink had 13.6 million consumer subscribers, up from 7.2 million a year earlier, while its connectivity segment posted $5.248 billion in quarterly revenue and $2.367 billion in operating profit. Enterprise and government revenue rose 91% year over year, and direct-to-cell monthly users topped 22 million. Still, the case is far from settled. Musk’s supply and demand figures do not line up neatly, pricing assumptions remain unclear, consumer ARPU is below last year’s level, and scaling depends heavily on Starship launches. Technical questions also remain, including whether low-Earth-orbit networks can handle the broader protocol and congestion-control challenges tied to machine-driven traffic.

SpaceX CEO Elon Musk used the company’s Oct. 20 third-quarter earnings call to argue that Starlink’s future market is much larger than rural broadband, in-flight connectivity, maritime internet, or battlefield communications. His case was blunt: global bandwidth demand could rise 1,000x over the next three to five years, and AI agents are the reason.

Musk said on the call that, on a per-agent basis, AI agents would use far more bandwidth than a person, and that the number of agents would eventually far exceed the number of people. In that framing, Starlink is not just a network for human users. It becomes infrastructure for AI agents, connected vehicles, and robots.

What Musk said on the earnings call

Musk started with supply. He said Starship’s 14th test flight deployed 26 Tbps of Starlink bandwidth in a single mission, equivalent to 10 Falcon 9 launches, and added that the vehicle was only about half full.

He said Starlink bandwidth could reach 100x the capacity of the current constellation in roughly three years. The path, as he described it, has two parts: the number of satellites rises by an order of magnitude to around 100,000, and bandwidth per satellite climbs by more than 10x versus the current system. He also said Starlink revenue would be at least roughly proportional to the total bandwidth delivered.

On the demand side, Musk said bandwidth demand could be 1,000x current levels within three to five years, adding that communications bandwidth would need to increase by several orders of magnitude to meet that demand.

SpaceX President and COO Gwynne Shotwell made a similar point on the call. She said communications growth would move far beyond traditional person-to-person traffic, and that connected cars, robots, and the spread of superintelligence would push demand higher by orders of magnitude rather than by percentage points. She also said Starlink could eventually carry a significant share of global internet traffic.

During the Q&A, Musk said SpaceX sees a clear path for its connectivity business to reach hundreds of billions of dollars in annual revenue, though he said the details would be discussed in a separate presentation.

Cloudflare had already floated a similar 1,000x figure

The idea did not appear out of nowhere. In August, Cloudflare had already raised a similar 1,000x scenario.

According to The Register, Cloudflare CFO Thomas Seifert said on the company’s second-quarter earnings call in August that if current trends continue, non-human traffic could be 1,000x human traffic within five years. He said humans could become a rounding error on the internet, not because human traffic is shrinking, but because non-human traffic is growing so quickly. He also added a caveat: 「I get every forecast wrong.」

Musk says AI agents could drive a 1,000x jump in bandwidth demand as SpaceX recasts Starlink’s future 3

Cloudflare had previously expected machine traffic to overtake human traffic in 2027. Instead, that crossover happened in May 2026.

Musk responded on X on Aug. 9, writing: 「AI agentic Internet traffic will obviously VASTLY exceed human usage. Not a close call at all. Cloudflare’s forecast is accurate.」 Minutes later, he added: 「The only system that can support the insane bandwidth growth needed for AI is Starlink.」

That said, the two 1,000x figures are not measuring the same thing. Cloudflare was talking about requests and traffic flowing through its network. Musk was talking about global bandwidth demand. Cloudflare’s estimate also came with explicit caution.

Musk’s revenue math points to a much bigger communications business

Musk had already laid out a rough revenue model on X on Aug. 8. He prefaced it by saying that some of what he says is speculation or aspiration, and some is not.

The non-speculative part, in his telling, is that V3 satellites launched by Starship have an order-of-magnitude higher capability than V2 satellites launched by Falcon 9, and that SpaceX will also launch an order of magnitude more satellites. That, he said, would mean more than 100x the bandwidth of the Starlink V2 system.

He added that the V2 system’s annualized revenue this year would reach $20 billion. Even if revenue per Gbit falls by 10x, SpaceX communications revenue would still exceed $200 billion a year.

Earlier the same day, replying to a user on X, Musk said that even if the communications market only doubles, he expects Starlink to capture at least 25% market share outside China, producing more than $500 billion in annual revenue. Over the long run, he said it is not impossible for Starlink to carry more than half of global internet traffic, which could imply more than $1 trillion a year.

What SpaceX reported for the third quarter

SpaceX’s third-quarter results show why the company is leaning into this story. As of Sept. 30, Starlink had 13.6 million consumer subscribers, up 1.6 million from the prior quarter and up from 7.2 million a year earlier. The company also said it has hundreds of thousands of enterprise users that are not included in that consumer figure.

Musk says AI agents could drive a 1,000x jump in bandwidth demand as SpaceX recasts Starlink’s future 4

The connectivity segment generated $5.248 billion in quarterly revenue, up 69% year over year, and $2.367 billion in operating profit, up 99%. Consumer revenue was $3.121 billion. Enterprise and government revenue was $2.127 billion, up 91% year over year. Shotwell said on the call that enterprise and government revenue will eventually surpass consumer revenue.

Over a longer stretch, Starlink consumer subscribers rose from 5 million at the end of the first quarter of 2025 to 13.6 million at the end of the third quarter of 2026.

Aircraft connectivity and direct-to-cell are the next growth areas

On the enterprise side, in-flight connectivity stands out. SpaceX said more than 3,800 aircraft had Starlink installed as of the end of September. Shotwell said roughly 12,000 aircraft are under contract, close to one-quarter of the addressable fleet, and three of the four largest U.S. airlines have signed on.

Direct-to-cell is the next step. Shotwell said monthly users of Starlink’s direct-to-cell service have passed 22 million, up 37% from the previous quarter. She also said the Federal Communications Commission approved SpaceX a few weeks ago to launch 15,000 second-generation direct-to-cell satellites.

In October, SpaceX announced an acquisition of 800 MHz low-band spectrum. The company said in its filing that the transaction is expected to close in the first half of 2027. Bloomberg also reported that SpaceX is in talks for at least 80,000 cell site locations.

Shotwell described the division of labor this way: S-band spectrum would carry most traffic, 800 MHz would handle building penetration, and SpaceX would also place small cell equipment at homes, businesses, and cell sites where Starlink is already installed. The target is to launch an integrated satellite-and-terrestrial service in the U.S. by early 2028.

Five questions the thesis still does not answer

1. Supply and demand do not line up cleanly

Musk said bandwidth supply could rise 100x in three years, while demand could rise 1,000x in three to five years. He did not say how much of that traffic Starlink expects to carry, or what customers would pay.

Telecom industry outlet RCR Wireless made the same point: a 100x increase in bandwidth and a 1,000x increase in demand are not the same thing, and the earnings call did not explain what share Starlink expects to win.

Musk says AI agents could drive a 1,000x jump in bandwidth demand as SpaceX recasts Starlink’s future 5

2. Subscriber growth is strong, but ARPU is below last year

Starlink consumer ARPU was $69 in the third quarter, up from $66 in the prior quarter but down from $83 a year earlier. Musk’s own model also assumes revenue per Gbit could fall by 10x, with total bandwidth growth making up the difference.

For reference, SpaceX did not separately disclose consumer ARPU for the fourth quarter of 2025. It gave a full-year average of $81.

3. The technical challenge is not only bandwidth

Cloudflare CTO Dane Knecht replied under Musk’s post on X: 「Bandwidth is half of it. Most congestion control still assumes a stable path and a stable RTT. LEO is neither.」

He said the entire network stack needs innovation and added that Cloudflare and Starlink should be building that together. No partnership between the two companies has been announced.

4. Scaling depends on Starship

The 100x capacity case depends on Starship launching V3 satellites at scale. Musk said on the earnings call that the 15th test flight in November would attempt to recover both the booster and Starship, though he did not guarantee success. He said he hopes to reach a weekly launch cadence in the first half of 2027.

5. Control over the network remains a live issue

This month, Ukraine’s military version of Starlink, Starshield, was throttled. The Wall Street Journal reported that connectivity improved after Donald Trump called Musk. Senator Richard Blumenthal has opened an inquiry into the throttling decision, questioning who controls a critical capability that the U.S. government has bought with billions of dollars.

What SpaceX has put forward is a much broader argument for Starlink: not just internet access for people, but communications infrastructure for AI agents and machine systems. Whether that argument holds will depend on launch execution, network engineering, pricing, market share, and control over the service itself.

This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan.
9600

Disclaimer:

The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.

Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.