Musk Claims AI Will Drive Global Economic Growth
At the G20 Innovation Ministers' Meeting hosted in the U.S., Musk stated that American companies must develop new energy sources to power data centers; he predicts that AI could drive global economic growth by 20% to 30% and significantly enhance productivity. This growth range is his personal forecast and not an official economic prediction from the G20, the International Monetary Fund, or the U.S. government.
Musk participated in the meeting via video, which took place in Raleigh, North Carolina; other invited tech industry figures included OpenAI CEO Sam Altman, Nvidia CEO Jensen Huang, Meta President Dina Powell McCormick, and former White House AI advisor David Sacks. The U.S. is using this meeting as a precursor to the G20 Leaders' Summit in Miami in December.
He directly linked the power demands of data centers to the expansion of the AI economy: model training and inference clusters require continuous power supply, cooling, and grid access. If the existing power system cannot be expanded, the construction of computing power will be limited. Musk did not specify what "new energy sources" refers to in public reports, whether it means nuclear, natural gas, solar, storage, geothermal, or other technological routes, nor did he disclose the time frame and calculation model for his 20% to 30% global growth estimate.
The policy backdrop of the meeting is the U.S. push for G20 members to adopt a non-binding light regulatory framework for AI. White House officials stated that this framework would encourage members to avoid creating new regulatory bodies for AI, to test emerging technologies in collaboration with the private sector, and to lower barriers to AI deployment; this means the U.S. is integrating AI infrastructure expansion, private sector experimentation, and cross-border regulatory coordination into the same diplomatic agenda.
G20 members account for about 85% of global GDP and two-thirds of the global population. If members reach coordination on data centers, power investments, model deployment, and talent cultivation, the policy impact will cover major economies; however, the proposed documents are not legally binding, and countries can still take different institutional paths based on energy security, labor markets, data sovereignty, and competition policies.
In market mechanisms, Musk's statements reinforce the capital narrative that "power is the bottleneck for AI expansion": model developers and cloud providers will procure GPUs, data center capacity, and long-term power contracts, while suppliers of grids, transformers, gas turbines, nuclear power, storage, and cooling equipment will compete for new investments. Beneficiaries will be infrastructure asset holders with grid-connected power sources, transmission access, land, and approval capabilities; those under pressure will be regions with insufficient power capacity, energy-intensive industries unable to secure stable electricity prices, and end users who must bear the costs of grid expansion.
Source: Public Information
ABAB AI Insight
Musk's reference to "new energy sources" aligns with his capital layout in Tesla, SpaceX, and xAI-related projects: Tesla bets on storage, batteries, and distributed power regulation, SpaceX possesses high-density communication networks and engineering capabilities, while companies like xAI require large-scale continuous computing power. In the past, tech companies focused on algorithms, data, and chips when discussing AI; as data center loads reach hundreds of megawatts or even gigawatts, grid access, transformer lead times, cooling capacity, and power generation assets begin to determine the speed of model deployment.
The capital path is shifting from "buying GPUs" to "locking in power." GPU orders can only determine theoretical computing power; long-term power purchase agreements, independent power generation projects, transmission queue positions, and data center park permits determine whether equipment can actually be powered. Recently, Anthropic signed a long-term computing power agreement worth $45 billion for about 350 megawatts with Nvidia-supported Lambda, with the project developed by Hut 8 in Nueces County, Texas, indicating that large model companies are securing physical capacity through multi-year contracts rather than waiting for spot cloud resources.
This is comparable to the railroad boom of the 19th century, the oil and chemical industry of the 20th century, and fiber optic construction in the internet era: upper-level demand first attracts capital, and then bottlenecks shift to lower-level networks with long construction cycles. The railroad boom was constrained by track rights and steel rails, cloud computing expansion is limited by fiber optics and servers, and now AI is constrained by power and cooling. The difference is that AI load growth may outpace the approval and construction cycles of power grids, nuclear power, and transmission projects, leading to infrastructure mismatches that convert short-term capital expenditures into regional electricity prices, land, and regulatory conflicts.
Essentially, this is about industrial chain restructuring: for the economic benefits of AI to be realized, some tech capital must be transformed into energy and utility capital. In the past, software companies could expand with light assets; the more AI model companies rely on training and inference clusters, the more they need to bear heavy asset constraints such as power generation, transmission, cooling, land, and financing. Those who have accessible power, expandable parks, and long-term customer contracts may shift from being supporting roles in the tech supply chain to becoming core nodes that determine AI capacity and pricing.
ABAB News · Cognitive Laws
The growth rate of AI will ultimately be constrained by the growth rate of power.
GPUs determine the upper limit of computing power, while the grid determines deployment speed.
The smarter the software, the more expensive the infrastructure.