AI’s massive power consumption is making energy infrastructure a hot topic. In a brand new report, Anthropic says the US is seriously lagging China on recent energy development and lays out what’s needed to keep up the country’s AI lead.
Training today’s largest AI models requires data centers that draw tens if not a whole lot of megawatts of power at peak load. Anthropic predicts that by 2028, leading developers would require training clusters with as much as five gigawatts of capability.
With several corporations competing to coach the most important models, that would add as much as around 25 gigawatts of recent power requirements for training alone. Anthropic predicts that not less than as much power will likely be needed to run finished models for purchasers, suggesting the US must deploy one other 50 gigawatts of capability in the following three years. And that’s on top of what is required to fulfill already rising energy demands.
But getting recent energy projects up and running within the US might be cumbersome, Anthropic says, which is putting the country at a serious drawback in comparison with China, which deployed an eye-watering 400 gigawatts of recent capability last 12 months. In a white paper titled, “Construct AI in America,” the corporate outlines regulatory and policy changes it thinks are needed to support the domestic AI industry.
“For america to guide the world in AI, we must make substantial investments in computing power and electricity that make it possible to construct AI in America,” the corporate wrote in a blog post.
The report outlines three key areas where the US is moving too slowly—constructing data centers themselves, constructing generation facilities, and constructing the transmission systems required to get electricity from one to the opposite. It also identifies the three biggest barriers holding these efforts back.
The primary is the array of permits that developers must secure before starting construction on any of those projects, specifically those pertaining to the environment. The second is transmission approvals that have to be sought from the state public utility commissions before constructing recent power lines, which might take years. And the third is the interconnection approvals from utilities that allow facilities to connect with the grid and may also take years for sign-off.
Anthropic proposes a two-stream solution. To hurry the event of recent AI infrastructure, the report suggests allowing data centers to be built on federal lands to avoid local zoning processes and streamlining environmental review of those projects.
It also suggests the Department of Energy should partner with private firms to speed up the event of recent power lines and demanding transmission upgrades. And the federal government should encourage utilities to hurry up the interconnection of power sources and data centers, even using national-security powers to further speed up the method.
The second pillar of their proposal focuses more on broader improvements to the country’s energy infrastructure. This includes streamlining permitting for brand spanking new geothermal, natural gas, and nuclear power plants and developing special high-capacity transmission corridors to serve areas with high AI datacenter growth.
Additionally they suggest using loans and guarantee programs to encourage greater domestic production of critical grid components like transformers and turbines and even making a national reserve for this stuff. Finally, they suggest creating training and entrepreneurship programs to assist boost the energy-industry workforce.
Certainly one of the corporate’s wishes already seems to have come true. President Trump announced plans to streamline datacenter and energy project permitting in his recent AI Motion Plan.
Whether the remainder of the proposals come to fruition stays to be seen. But there appears to be a growing consensus that winning the AI race would require some pretty hands-on industrial policy.