In a substantive 10-minute SpaceX company-address excerpt, Elon Musk says AI revenue could exceed SpaceX's other businesses in September, targets 10 GW by the end of 2027, and projects $300 billion to $500 billion in annual revenue. The strategy is real; the timeline and economics remain forecasts that must be tested against reported capacity, customers, spending and revenue.
SpaceX AI Could Hit $500 Billion in Annual Revenue, says Elon Musk · VideoFromSpace
Elon Musk used a recent SpaceX company address to make an extraordinary claim: artificial intelligence could soon produce more revenue than the company's rockets, spacecraft and Starlink businesses combined.
That is a meaningful signal about where SpaceX is putting its attention. It is not proof that the forecast will happen.
*AI Shift News tracks major AI infrastructure claims and separates operating results from executive forecasts. Subscribe for the next evidence check.*
The central video is a 10-minute original-English excerpt published by VideoFromSpace and credited to SpaceX. Unlike the earlier 2½-minute news clip, this segment contains Musk's full SpaceX AI discussion. He talks about internal AI software, terrestrial training clusters, inference in orbit, a 10-gigawatt compute target and a possible $300 billion to $500 billion annual-revenue business.
The video is valuable because it lets viewers hear the strategy in Musk's own words. It also makes clear why every number needs an attribution label.
Musk says SpaceX AI revenue could exceed all other SpaceX revenue in September and become significantly larger in the fourth quarter. He then projects that SpaceX could bring 10 gigawatts of AI capacity online by the end of 2027. At an assumed value of $30 to $50 per watt, he calculates that capacity could support $300 billion to $500 billion in annual revenue. Farther out, he says AI could represent 99% of SpaceX's value within five years.
Those statements are forecasts from SpaceX's chief executive. They are not audited future results, signed customer orders for the full amount, or proof that the capacity will arrive on schedule.
What is real today
SpaceX is no longer only a launch and satellite-connectivity company. Its official updates describe a vertically integrated business spanning rockets, Starlink, direct-to-device communications and AI. Recent public reporting on its quarterly results also shows that AI is already a material operating segment rather than a distant concept.
That distinction matters. There is a real business underneath the forecast.
AP and Axios reported that SpaceX generated $7.81 billion in quarterly revenue while spending heavily on AI infrastructure. The company still recorded a loss, and investors were weighing whether the growth justified the spending. That gives readers a more useful baseline than the biggest number in the video: meaningful current AI revenue, enormous capital requirements and unresolved execution risk.
The video also identifies several parts of the strategy. Musk says SpaceX must succeed in both AI hardware and software. He discusses using the company's engineering information to improve internal AI tools, expanding ground-based compute and eventually moving more inference work into space. He describes solar-powered orbital infrastructure as a path to much larger computing capacity.
These pieces fit together as a strategy. They do not yet prove the economics.
The 10-gigawatt test
Ten gigawatts is a vast amount of computing power. Reaching it requires chips, networking equipment, buildings or orbital hardware, power generation, cooling, permits, supply agreements, skilled operators and customers willing to pay for the capacity.
The revenue projection adds another assumption: that each watt can reliably produce $30 to $50 of annual value. Even if SpaceX reaches the capacity target, pricing, utilization and customer demand will determine whether the revenue follows.
This is where readers should separate engineering ambition from business proof.
SpaceX has a strong record in launch engineering and satellite deployment. That history makes its infrastructure plans worth taking seriously. AI infrastructure is still a different competitive market. SpaceX must compete with hyperscalers, specialized cloud providers and other well-funded AI labs. It must deliver reliable services, win customers and control costs while spending at an exceptional rate.
The company may have advantages. It can combine launch capability, satellite manufacturing, connectivity, energy planning and AI operations inside one organization. If orbital inference becomes practical, SpaceX can build and launch its own hardware instead of depending completely on another launch provider.
But vertical integration is not automatic profitability. Owning more layers can reduce dependencies; it can also concentrate capital risk and operational complexity.
What the video proves—and what it does not
The video proves that Musk is making AI central to SpaceX's internal message. He calls success in AI essential, asks employees to use and improve the technology, and describes hardware and software as the company's most important future challenge.
It also proves that the 10-gigawatt and $300 billion to $500 billion figures came directly from Musk rather than from an anonymous social post or a third-party commentator.
The video does not prove that 10 gigawatts will be operating by the end of 2027. It does not provide a customer list supporting the entire revenue range. It does not show audited projections, utilization assumptions or the capital required to reach the target. It does not establish that orbital inference is already cheaper than terrestrial service.
The “99% of value” statement is even more speculative. Company value depends on revenue, margins, risk, competition and what investors are willing to pay. It cannot be verified today.
Why this matters
The practical story is not that SpaceX has already become a $500 billion AI business. The story is that its chief executive is publicly redefining the company around AI infrastructure and software—and attaching a near-term operating target to that shift.
That changes what customers, competitors and investors should watch.
For customers, the important evidence will be named services, pricing, performance, reliability and contract announcements. For competitors, the key question is whether SpaceX's control of launch and connectivity creates a cost or deployment advantage. For readers, the main job is to track actual capacity and revenue against the timeline in this video.
The September claim creates the first checkpoint. If AI revenue does exceed every other SpaceX business line, the next company results should make that shift visible. The end-of-2027 target creates a second checkpoint: deployed capacity should move materially toward 10 gigawatts, with credible reporting on how much is active and serving customers.
What to watch next
Watch for four pieces of evidence.
First, look for company filings or formal updates that break out AI revenue and spending. Those figures can show whether the September and fourth-quarter claims are materializing.
Second, watch for named customer contracts. Large capacity is only valuable when customers use it at sustainable prices.
Third, track deployed compute rather than announced compute. Power agreements and construction plans matter, but operating capacity is the better measure.
Fourth, look for technical evidence on orbital inference: hardware launched, useful workloads completed, costs disclosed and reliability demonstrated.
There is also a governance question. Using SpaceX's accumulated engineering information in internal AI systems could make those tools more useful to employees, but it raises practical questions about access controls, security boundaries, evaluation and accountability. A capable internal model is not automatically a safe or reliable one. SpaceX will need evidence that the system improves work without exposing sensitive information or turning confident model output into an unchecked engineering decision.
*Subscribe to AI Shift News for the follow-up scorecard when SpaceX reports the results behind these forecasts.*
The honest verdict is straightforward. SpaceX AI is already a serious operating effort, and Musk's full company-address segment is worth watching. His projections describe a possible future, not an achieved one. The next useful story will come from measurable capacity, customers and reported revenue—not another giant number on its own.
Bottom Line
SpaceX AI is a serious operating effort, but Musk's 10 GW and $300 billion to $500 billion revenue figures remain forecasts that require capacity, customer, and reported-revenue proof.