AI Can Scale Faster Than the Grid Can

Thank you to our Sponsor: VITURE

For most of the AI boom, the race has been framed as a competition over intelligence. Which company has the best model? Which system can reason more effectively, write better code, process more data, or operate more autonomously? Increasingly, however, the limiting factor may not be intelligence at all. It may be electricity.

AI companies can improve models in months, deploy new systems almost instantly, and add millions of users through software updates. Power plants, transmission lines, substations, transformers, and data centers operate on a very different clock. They can take years to permit, finance, construct, and connect to the grid. That mismatch is creating one of the most important tensions in the AI economy.

The next major AI race may therefore be between two very different curves: the speed at which AI can scale and the speed at which the physical infrastructure supporting it can grow. If intelligence keeps accelerating while power infrastructure falls behind, electricity could become one of the most important constraints on AI development.

AI Is Becoming a Physical Infrastructure Business

AI may look like software when someone opens a chatbot, generates an image, or asks an agent to complete a task, but everything happening on the screen depends on enormous physical infrastructure somewhere else. Models run inside data centers filled with specialized chips, networking equipment, cooling systems, backup power, and increasingly massive electricity connections.

As models grow and AI usage spreads into more industries, the physical requirements grow with them. The technology industry is therefore discovering that software can scale almost instantly while electricity cannot.

• AI data centers require enormous amounts of electricity, especially when thousands of high-performance processors operate continuously for training and inference.

• Model usage adds to infrastructure demand, because serving millions of everyday AI requests can consume enormous computing resources even after training is complete.

• Cooling becomes part of the AI infrastructure equation, since powerful computing systems generate large amounts of heat that must be removed continuously.

• Grid connections are becoming strategic assets, because having servers does little good if sufficient electricity cannot be delivered to the facility.

• AI companies are increasingly becoming infrastructure companies, whether they intended to or not, because future growth depends on power, land, construction, networking, and energy supply.

The AI boom may have started as a software revolution, but its next stage increasingly resembles an industrial buildout.

Thank you to our Sponsor: Pokee AI

Software Moves in Months. Electricity Moves in Years.

The central problem is speed. An AI company can launch a new model, attract millions of users, and dramatically increase computing demand in a remarkably short period. Utilities cannot respond that quickly.

Building a large power plant can take years. Transmission projects can take even longer because they may require environmental reviews, rights-of-way, regulatory approval, financing, equipment, construction, and coordination across multiple jurisdictions.

That produces a fundamental mismatch between the digital and physical economies.

• AI demand can appear suddenly, while utilities traditionally plan electricity infrastructure years in advance.

• New transmission lines can face long approval processes, particularly when they cross multiple communities or regulatory jurisdictions.

• Large electrical equipment can have long manufacturing lead times, making transformers, turbines, switchgear, and related infrastructure potential bottlenecks.

• Data centers can sometimes be constructed faster than the grid can support them, leaving companies waiting for power connections.

• Traditional utility planning was not designed for enormous computing facilities appearing rapidly, sometimes requesting power comparable to major industrial operations.

The faster AI advances, the more visible this mismatch becomes. The industry can manufacture intelligence faster than society can manufacture the infrastructure that keeps that intelligence running.

Thank you to our Sponsor: Omane Media

Data Centers Are Becoming Power-System Players

Data centers were once viewed primarily as technology facilities. AI is turning some of them into energy projects.

A large AI campus may require hundreds of megawatts of power, and future projects could require much more. At that scale, the arrival of a data center can affect utility planning, generation investments, transmission capacity, and regional electricity demand.

That gives technology companies a new kind of influence over energy systems.

• Data-center developers are increasingly negotiating directly with utilities over how quickly new facilities can be connected.

• Technology companies may help finance new generation, particularly when existing regional capacity cannot meet future demand.

• Long-term electricity contracts could become strategically important, allowing AI companies to secure reliable power before competitors do.

• Location decisions may increasingly depend on energy availability, rather than simply proximity to technology talent or major cities.

• Regions with abundant electricity could become important AI hubs, giving energy-rich areas a larger role in the technology economy.

The geography of AI could therefore start following the geography of electricity. The next important computing cluster may be built where power is plentiful rather than where software companies have historically been concentrated.

Thank you to our Sponsor: Partnerly

The Fight Over Who Pays Is Just Beginning

The electricity required by AI data centers does not magically appear. Utilities may need to build new substations, transmission lines, generation capacity, and grid infrastructure to support that demand, and someone ultimately has to pay for those investments.

That is where the AI infrastructure boom becomes political. Communities may welcome jobs, investment, and tax revenue from data centers while questioning whether households and small businesses should absorb any of the cost of supporting them.

• Utilities may need billions of dollars in new infrastructure to connect rapidly expanding data-center demand.

• Ratepayers may resist paying higher electricity bills for projects primarily serving some of the world's largest technology companies.

• Technology companies may face pressure to fund more grid expansion themselves, particularly when their facilities drive the need for new investment.

• State and local governments may compete aggressively for data centers, creating tension between economic-development incentives and infrastructure costs.

• Public opposition could become a meaningful constraint on AI growth, especially when communities associate new data centers with higher prices, land use, water consumption, or grid strain.

The political argument could become remarkably simple: everyone may want AI investment, but fewer people may want to pay for the infrastructure required to support it.

Thank you to our Sponsor: EezyCollab

AI Could Change the Energy Mix

The enormous electricity requirements of AI are also forcing technology companies to think differently about where power comes from. Purchasing renewable-energy credits is one thing; ensuring that a massive AI facility receives dependable electricity every hour of every day is something else.

That could make AI companies important customers for nuclear power, natural gas, renewables, batteries, geothermal energy, and other generation technologies. Instead of merely purchasing electricity, technology firms may increasingly help shape which types of electricity get built.

• Nuclear energy could become increasingly attractive because it can provide large amounts of continuous low-carbon electricity.

• Natural gas may remain important for reliability, particularly in regions where demand is growing faster than renewable generation and storage.

• Solar and wind will continue expanding, but their variability means storage, transmission, or complementary generation may also be required.

• Advanced geothermal and other emerging technologies could receive more investment, especially if they can provide dependable around-the-clock power.

• Technology companies may become major energy investors, signing long-term agreements or directly supporting generation projects to guarantee future supply.

AI could therefore accelerate energy innovation while simultaneously increasing pressure on existing electricity systems. The same technology creating the demand could help finance the next generation of power infrastructure.

Efficiency May Not Solve the Problem

There is an appealing assumption that more efficient AI models and chips will eventually solve the electricity problem. Better processors can perform more calculations with less energy, and smaller models can reduce the computing required for many tasks.

The problem is that cheaper AI may encourage dramatically more AI use. When the cost of something falls, people often consume more of it, and AI could follow the same pattern.

• More efficient chips reduce the cost of individual AI operations, making AI economical for more applications.

• Cheaper inference encourages companies to deploy AI more widely, potentially increasing total demand even as each request becomes more efficient.

• AI agents could operate continuously, generating computing demand without waiting for a human to initiate every interaction.

• Physical AI could create entirely new workloads, as robots, vehicles, factories, and machines continuously process sensor information.

• Efficiency gains could therefore increase total electricity consumption, because the number of AI-powered activities grows faster than energy use per activity falls.

AI could become dramatically more efficient while still consuming more electricity overall. That means efficiency matters, but it may not eliminate the infrastructure challenge.

The Grid Could Become the New AI Bottleneck

For years, the major AI bottlenecks were chips, training data, engineering talent, and computing capacity. Electricity increasingly belongs on that list.

A company may have access to advanced GPUs, enormous amounts of capital, and world-class researchers, yet still be unable to deploy computing capacity if it cannot secure enough reliable electricity.

That changes competitive advantage.

• Access to power could become as strategically important as access to chips, particularly for companies operating enormous computing clusters.

• Grid connection queues could influence where AI infrastructure gets built, shifting investment toward regions where power can be delivered faster.

• Energy contracts could become competitive weapons, allowing companies to lock in scarce generating capacity years in advance.

• AI infrastructure strategy may increasingly require energy expertise, bringing utility planning directly into technology-company decision-making.

• The winners of the AI race may include companies that solve physical infrastructure problems, not simply those producing the smartest models.

The AI industry may discover that intelligence itself is no longer the hardest resource to scale. The harder resource may be everything required to keep that intelligence running.

That creates an unusual race. On one side are AI companies improving models at extraordinary speed. On the other are utilities, regulators, construction companies, equipment manufacturers, and energy developers trying to build the physical infrastructure required to keep up.

For years, the AI industry has been obsessed with how quickly machines can become more intelligent. The next question may be much more practical: Can we build enough electricity infrastructure fast enough to let them?

New AI news site from Unaligned: 

We built an AI agent that reads 40,000 posts on X every day from the AI community and it finds the best info from all of that: https://alignednews.com/ai

We call it “Aligned News” and the same agent will start sending out its own newsletter every day. Please support it, it’s from our team. It will show you all the best discussion of news, AI papers, models, events, and much more.

Looking to sponsor our Newsletter and Scoble’s X audience?

By sponsoring our newsletter, your company gains exposure to a curated group of AI-focused subscribers which is an audience already engaged in the latest developments and opportunities within the industry. This creates a cost-effective and impactful way to grow awareness, build trust, and position your brand as a leader in AI.

Sponsorship packages include:

• Dedicated ad placements in the Unaligned newsletter

• Product highlights shared with Scoble’s 600,000+ X followers

• Curated video features and exclusive content opportunities

• Flexible formats for creative brand storytelling

📩 Interested? Contact [email protected], @samlevin on X, +1-415-827-3870

Just Three Things

According to Scoble and Cronin, the top three relevant and recent happenings

​​AI Companies Prepare for the “Day After” a Major AI Crisis

Top AI companies are reportedly preparing for the fallout from a major AI-driven crisis, such as a large cyberattack that disrupts financial services, internet access, or critical infrastructure. Their planning focuses on how governments and the public might react, how regulation could rapidly tighten, and how AI companies could respond after a serious real-world failure. Axios

Anthropic AI Sends False Homicide Tip

An Anthropic AI model submitted a false homicide tip through the Philadelphia Police Department’s website during an automated testing process. Police said the tip was caught as spam and never reached investigators, while Anthropic stopped the test after discovering what happened. Reuters

AI May Be Weakening Our Ability to Think Through Hard Problems

A UC Berkeley-led study found that relying on AI for just 10 minutes can reduce people’s ability to persist and perform once the tool is removed. The research suggests that overreliance on AI may weaken “productive struggle,” raising concerns that constant assistance could erode focus, learning, and long-term cognitive resilience. UC Berkeley News

Scoble’s Top Five X Posts