TL;DR
Thorsten Meyer AI has framed the AI infrastructure race around power access, arguing that the grid queue is now a binding constraint for growth. Public data from U.S. energy agencies and the IEA supports the broader pressure point: data centers are raising electricity demand faster than many grids can add firm supply and connections.
Thorsten Meyer AI has published a headline-only analysis framing the next AI infrastructure bottleneck as the electricity grid queue rather than the chip supply chain, a claim that matters because public energy data shows data centers are now a major driver of U.S. power demand growth.
The supplied source material contains only the headline, “The queue. Why the grid, not the chip, is the binding constraint on AI.” No article body, dataset or author explanation was available, so the core claim must be treated as analysis rather than a confirmed finding from that source.
The confirmed backdrop is clear: U.S. electricity demand has been rising since 2020 after more than a decade of slow growth, and the U.S. Energy Information Administration says data centers are helping drive that increase. In its 2026 outlook work, EIA forecast U.S. electricity load growth of 1.9% in 2026 and 2.5% in 2027, with the fastest growth expected in ERCOT and PJM, two regions with heavy data center activity.
Other public data points in the same direction. The Department of Energy has cited an Electric Power Research Institute estimate that data centers could consume up to 9% of U.S. electricity generation by 2030, compared with about 4% of load in 2023. FERC staff reported that data centers used 4.4% of U.S. electricity in 2023 and estimated that more than 50 gigawatts of in-service data center capacity existed by the end of 2025.
Why It Matters
The analysis matters because AI capacity is often discussed through access to advanced chips, but chips do not run without reliable, large-scale electricity. If power delivery becomes the tighter limit, AI developers face constraints that cannot be solved only through semiconductor procurement.
Grid limits can affect where AI campuses are built, how fast new models can be trained, what cloud capacity costs, and whether local customers face higher power prices. They can also shift the energy mix. The International Energy Agency has said that in a high-growth data center case, long grid connection queues would mean much of the extra demand beyond added renewables is met by fossil-fuel generation.
The issue is also local. Data centers can require power levels comparable to large industrial sites, and FERC staff said the average size of data centers entering service rose from about 25 megawatts in 2020 to almost 80 megawatts in 2025. Future sites may be larger, which can require new generation, transmission upgrades or new tariff structures before they connect.
APC UPS Battery Backup for Power Outages, 600VA/330W Surge Protector, 7 Outlets, USB Charging, BE600M1 Uninterruptible Power Supply for Computers, Wi-Fi Routers, and Home Office Electronics
KEEP YOUR COMPUTER, WI-FI AND ROUTER RUNNING THROUGH POWER OUTAGES: Supplies short‑term battery power during outages to maintain…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background
The chip shortage story dominated the early AI buildout because advanced GPUs were scarce and expensive. That constraint has not disappeared, but the grid issue has moved higher because many proposed AI campuses require firm power at a specific location, not only broad national electricity supply.
Interconnection queues are a long-running power-sector problem. Lawrence Berkeley National Laboratory reported that active U.S. generation and storage capacity in queues fell 12% in 2024 but still stood at about 2,290 gigawatts at the end of that year. Those queues refer mainly to new supply resources, yet they show the same system friction now facing large new loads: studies, upgrades, equipment, siting and cost allocation can take years.
For AI companies, this changes the competitive map. Locations with available substations, transmission headroom, firm generation contracts or behind-the-meter power may become more valuable than sites chosen only for tax incentives, fiber access or land.
“The queue. Why the grid, not the chip, is the binding constraint on AI.”
— Thorsten Meyer AI headline
“Electricity use by data centers is driving the electricity demand growth”
— U.S. Energy Information Administration
“up to 9% of U.S. electricity generation annually by 2030”
— U.S. Department of Energy, citing EPRI
“long grid connection queues”
— International Energy Agency
CyberPower CP1500PFCLCD PFC Sinewave UPS Battery Backup and Surge Protector, 1500VA/1000W, 12 Outlets, AVR, Mini Tower, UL Certified
1500VA/1000W PFC Sine Wave Battery Backup Uninterruptible Power Supply (UPS) System designed to support active PFC and conventional…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Remains Unclear
It is not yet clear how much of future AI deployment will be delayed specifically by grid queues rather than by chips, capital, permitting, cooling, water, local opposition or model demand. The supplied Thorsten Meyer AI source is headline-only, so its evidence, definitions and time frame cannot be checked from the provided material.
It is also unclear which regions will feel the sharpest constraints. Public forecasts point to pressure in ERCOT, PJM and other data center markets, but actual outcomes will depend on utility planning, project withdrawals, new tariffs, generation additions and whether AI companies can shift workloads across regions.
AEDIKO 2pcs High Voltage Generator DC 6-12V to 1000kV Boost Step-Up Inverter Arc Pulse Generator Power Module High Voltage Transformer
High Voltage Generator:Input Voltage: DC 6V to 12 V;Output Voltage: About 500KV~1000KV;Input Current: 2A – 5A
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What’s Next
The next milestones are grid-operator load forecasts, utility tariff cases for large data center customers, interconnection reform efforts and company disclosures about power-backed AI capacity. Investors and customers will be watching whether AI firms can secure megawatts as readily as they secure chips.
for Sunon PF40561BX-Q310-S99 40mm Server Fan | 24W Power Cooling for Inverters/Industrial Equipment/Data Centers
High-quality flat cable: selected durable materials, good toughness and anti-aging, stable transmission.
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What does “the queue” mean in this story?
It refers to the waiting process for connecting new power supply or large new electricity users to the grid. For AI data centers, the issue is whether enough power can be delivered at the right location and time.
Is the chip shortage over?
No. Advanced AI chips remain a key constraint for many companies. The analysis argues that power access may now be an equal or greater limit for new AI infrastructure in some markets.
Why can’t data centers just build their own power?
Some are trying. But private generation still faces fuel supply, permitting, land, transmission, backup, emissions and reliability rules. Connecting private power to a large computing campus can still involve grid studies and utility approval.
Could this raise electricity prices?
Yes, in some regions. EIA has warned that if demand grows faster than supply, stress can appear through higher wholesale prices or reliability risks. The scale depends on local grid conditions and cost-sharing rules.
What remains unknown?
The exact share of delayed AI projects caused by grid queues is not public. The source provided for this article was headline-only, so the underlying Thorsten Meyer AI evidence could not be reviewed.
Source: Thorsten Meyer AI
Source: Thorsten Meyer AI
