In 30 seconds
Elon Musk argues that electricity could constrain AI deployment and predicts solar power and reusable rockets will enable computing in space. In his Davos conversation with Larry Fink, he also forecasts robot sales and AI progress. Musk says China deploys over 1,000 GW of solar annually; subsequently published data put 2025 additions at about 317 GW and cumulative capacity at about 1,200 GW. His aging and product timelines are projections.
- 1Electricity is Musk’s proposed AI bottleneckHe says AI costs may fall while the pace of adding electrical power limits how many chips can be operated.
- 2His China solar figure differs from later installation statisticsMusk cites over 1,000 GW annually; later data put 2025 additions around 317 GW and total installed capacity around 1,200 GW.
- 3Optimus public sales were a conditional forecastMusk projected sales by the end of 2027 if the robot achieved sufficient safety, reliability and functionality.
- 4Full Starship reuse is central to Musk’s cost argumentHe contrasts Falcon 9’s reusable booster with its discarded upper stage and predicts much lower costs if the whole rocket can be reused.
Takeaways
- Musk identifies electricity supply as a potential bottleneck for deploying AI chips.
- Later data put China’s 2025 new solar capacity at about 317 GW, compared with roughly 1,200 GW cumulative capacity.
- Musk’s projected Optimus sales and space-computing cost reductions depend on future technical and commercial progress.
- Tesla’s supervised FSD requires an attentive driver ready to take over.
Read the edited text by chapter
A multiplanetary future and robot abundance
Musk describes his companies’ goals, then Fink asks about the distribution and meaning of robot-driven growth.
After joking about the audience’s applause, Larry Fink welcomes Elon Musk to the World Economic Forum in Davos and cites Tesla’s historical shareholder returns before asking about engineering across AI, robotics, energy and space.
Musk says his companies address difficult technical challenges under a broad goal of improving civilization’s prospects and extending life beyond Earth. He describes SpaceX’s rocket development as a means of eventually reaching the Moon, Mars and other star systems.
He jokes about being an alien, then says he knows of no alien spacecraft encounters by SpaceX satellites. He compares consciousness to a small candle in darkness and argues that establishing life on more than one planet could protect it from an Earth disaster.
Musk connects Tesla’s work in sustainable technology with AI and robotics, which he predicts could greatly expand economic output. He also says the technologies need careful handling and invokes a fictional dystopia as an outcome to avoid.
Fink asks whether the gains can be broadly shared and how people would find purpose if machines performed necessary work. Musk imagines more robots than people and enough goods and services to satisfy human needs, including possible help with family care.
Aging, AI access and the power constraint
Musk offers speculative ideas on aging and identifies energy as a possible limit to AI deployment.
Asked whether aging could be reversed, Musk predicts that it may be possible and speculates about a synchronizing biological clock. Aging is associated with multiple biological processes; a single clock and an effective way to reverse human aging are not established by this exchange.
Fink asks about the cost of chips, manufacturing and energy and whether AI access will be broad. Musk expects falling costs and competition for customers to widen access, and identifies electricity supply as a likely constraint on deploying more chips.
Solar power and planned robots
The conversation covers China’s solar scale, US manufacturing plans and the proposed Optimus rollout.
Musk says China is deploying more than 1,000 gigawatts of solar a year. Later published figures put China’s 2025 new solar installations at about 317 gigawatts and cumulative capacity at about 1,200 gigawatts; installed capacity and annual additions are different measures.
He estimates solar and batteries could provide substantial power for AI, then argues that the Sun’s enormous output motivates space-based solar and computing. He predicts SpaceX will launch solar-powered AI satellites in coming years.
Fink asks how much land the United States would need to produce its electricity with solar panels. Musk estimates an area of about 100 by 100 miles, while saying installations should be distributed; he argues solar tariffs increase deployment costs.
Musk says SpaceX and Tesla are separately working toward large US solar manufacturing capacity, giving an aspirational figure of 100 gigawatts a year and an estimated three-year timeframe. He encourages other companies to build capacity as well.
Asked about robot deployment, Musk says Tesla’s Optimus prototypes are performing simple factory tasks. He predicts more complex tasks later in 2026 and public sales by the end of 2027 if safety, reliability and functionality are sufficiently high.
Musk discusses updates to Tesla’s driving software and predicts wider robotaxi service and European approvals. He calls self-driving essentially solved, but Tesla’s supervised FSD is a driver-assistance feature requiring a fully attentive driver ready to take over.
Reusable rockets and space computing
Musk predicts lower launch costs and describes solar-powered computing in orbit.
Fink asks how automation changes the economics of space. Musk says SpaceX hopes to demonstrate full Starship reusability in 2026, contrasting it with Falcon 9, whose booster can return but whose upper stage is discarded.
Musk predicts full reusability could dramatically reduce launch costs. He compares throwing away a rocket after each flight with discarding an aircraft after a trip; the cost reduction is his projection rather than a demonstrated operating price.
He argues that solar panels in space could produce more energy and envisions solar-powered AI data centers with radiators. He predicts that space-based computing could become the lowest-cost option in two or three years.
AI predictions and curiosity about Mars
He gives AI timelines, recalls science-fiction influences and discusses a possible Mars trip.
Asked about a ten-year horizon, Musk says forecasting is difficult but predicts AI could surpass any individual human by the end of 2026 or in 2027, and human intelligence collectively around 2030 or 2031.
Fink asks what inspired Musk’s curiosity. Musk recalls science fiction, fantasy and comics, and says he wants to understand the universe and turn aspects of science fiction into reality.
Asked whether he would visit Mars, Musk says yes and notes the long intervals between launch opportunities. Fink closes the conversation, and Musk encourages the audience to remain optimistic about the future.
More from this channel
30:20EN
26:00EN
13:30EN
32:15EN
59:55EN
38:45EN