America’s electricity challenge is often framed as a shortage of power, especially as AI increases demand from data centers. But the source article points to another pressure point: a large amount of new generating capacity is already waiting to connect to the grid.
Google and PJM now want to see whether AI can help reduce that bottleneck. Their partnership, announced Thursday with Alphabet moonshot Tapestry, aims to build AI models and planning tools for a grid approval process that has become a major constraint on new power projects.
Why grid connection has become a technology problem
Before new power capacity can serve customers, it has to be approved for connection to the grid. That interconnection process is technical, administrative, and slow. Developers submit requests, grid operators evaluate whether the system can handle the new supply, and both sides work through data, engineering, and planning requirements.
The issue is not limited to one region. The source states that all grid operators in the U.S. face similar backlogs. PJM’s backlog, however, stands out because of its size and importance.
PJM manages the flow of electricity in the mid-Atlantic states, Ohio, and eastern Kentucky. Its queue includes over 3,000 active requests to connect 286.7 gigawatts of capacity in the region, according to the Berkeley Lab. The organization became so overwhelmed that it stopped accepting applications for new connections in 2022 and will not review new requests until mid-2026.
That delay matters because electricity demand has become a strategic concern for large technology companies. AI computing requires power, and companies building or operating major data center infrastructure have been trying to secure more generation.
What Google, PJM, and Tapestry plan to build
The partnership between Google, PJM, and Tapestry is focused on speeding up key parts of the application process. The organizations plan to develop AI models that can help on both sides of a grid connection request.
According to the source, the work will include assistance with data verification and new centralized planning tools for project submission. Those tools are also intended to help PJM analyze how best to integrate variable power sources like renewables.
In plain terms, the goal is not to create more electricity directly. The goal is to make it easier to process projects that are already trying to enter the system. If the process becomes faster and more organized, more capacity could move from queue to grid.
That distinction is important. A power shortage can be caused by a lack of generation, but it can also be worsened by an approval system that prevents available projects from connecting. The source article argues that untangling the bureaucracy could go a long way toward addressing the problem.
The backlog is bigger than today’s power plant fleet
The scale of the queue explains why this administrative problem has drawn attention. Nationwide, 2.6 terawatts of generating capacity are waiting for approval, according to the Lawrence Berkeley Lab. The source says that is double what every U.S. power plant combined is capable of generating today.
Those numbers show why interconnection has become a core energy issue rather than a narrow procedural concern. If a large amount of capacity cannot move through review, the grid may remain constrained even while developers are proposing new supply.
The source highlights several major figures:
- 2.6 terawatts of generating capacity are waiting for approval nationwide.
- PJM has over 3,000 active requests in its queue.
- Those PJM requests represent 286.7 gigawatts of capacity.
- PJM stopped accepting new connection applications in 2022.
- PJM will not review new requests until mid-2026.
For tech companies, the backlog intersects directly with AI infrastructure. The source notes that Amazon, Google, Meta, and Microsoft have either invested in or pledged to buy significant amounts of nuclear power. It also says they have been steadily acquiring large amounts of solar power.
That mix reflects a broader race for dependable electricity. Data centers need power, and the companies behind AI services are trying to secure it through multiple sources.
Renewables are central to the queue
The source article says renewables have been penalized the most by the slow process. Nationwide, over 1 terawatt each of solar and storage are waiting for permission to send electrons to the grid.
Even in the PJM region, which the source says is not typically considered a hotbed of renewable development, solar and storage dominate the queue. Natural gas power plants account for just 2.4% of applicants.
This creates a tension for the grid operator. PJM’s system has historically been dominated by fossil fuels. Over the last decade or so, natural gas-fired power plants have displaced coal as fracking drove gas costs down.
The source also notes that PJM recently developed a new approval process that critics argue allows fossil fuel plants to unfairly skip the line ahead of renewable projects. In unveiling the partnership with Google, PJM Executive Vice President Aftab Khan said the organization’s grid will remain “fuel agnostic,” according to E&E News.
Google spokesperson Amanda Peterson Corio said the company was “committed to our goals to decarbonize our electricity footprint.” Those two statements frame the balance the partnership will have to navigate: improving grid throughput while addressing the role of renewables, fossil fuels, and growing AI power demand.
What success would mean for AI and the grid
If the AI tools work as intended, they could make the interconnection process more efficient without changing the basic need for careful grid review. Data still has to be verified. Projects still have to be analyzed. Grid operators still have to decide how new sources of electricity can be integrated.
But a better process could reduce delays that keep proposed capacity from becoming usable capacity. For data centers, that could ease concerns about underpowered infrastructure. For energy developers, it could make the path from application to connection less uncertain.
The larger implication is that AI may become part of the solution to a problem partly intensified by AI demand. Google’s interest is clear: the company needs power for computing growth, while also maintaining its stated electricity decarbonization goals. PJM’s interest is also clear: its queue is long, and its review process has already been paused for new requests.
The partnership does not erase the hard engineering and policy questions around the grid. It does, however, target a practical chokepoint. With terawatts of capacity waiting for approval, faster and better organized grid connection reviews could become one of the most important energy tools available.