The Age of Super Intelligence: Why It Runs on Power

by | Oct 7, 2026 | Energy | 0 comments

By Jason Charles Guck, Business Development, Delta Edge CI

Every so often a technology outgrows its name. We still say horsepower about engines that have never seen a horse. We still dial phones that have no dial. Usually the old word sticks around long after it stops being accurate.

Occasionally the moment arrives to retire the label and choose a better one. That moment may be here, and for anyone who buys electricity it matters more than a naming argument usually would.

What actually happened

On September 29, 2026, the President signed Executive Order 14434, titled Inaugurating The Era Of Super Intelligence. It directs federal agencies to use the terms Super Intelligence and SI in place of Artificial Intelligence and AI across official correspondence, communications, websites, reports and policy documents.

The order’s own reasoning is worth quoting. It states that contemporary systems far exceed what was envisioned when the term Artificial Intelligence first came into use, and that they increasingly amplify human ingenuity rather than merely imitate or automate discrete aspects of human intelligence.

The same day, the President hosted a luncheon with many of the most influential people in the industry, several of whom signed a voluntary safety accord. What followed is the part worth noticing: the leaders of these companies began using the word themselves. Jensen Huang of NVIDIA described data centres as super intelligence factories. Elon Musk corrected himself mid-sentence at the press event. Mark Zuckerberg had used the term the day before.

Not everyone is convinced the label will stick. Some researchers argue superintelligence should describe systems more advanced than anything running today. That is a fair objection and worth taking seriously.

Why the old word stopped fitting

Artificial intelligence was coined in the 1950s for a research programme aimed at imitating aspects of human thinking. For decades the description held. Early systems were brittle, narrow and obviously imitative.

Artificial suggests a copy of the real thing. In specific domains these systems are no longer copying human performance. They are exceeding it.

The companies building them already name themselves accordingly. Meta’s frontier research division is Meta Superintelligence Labs. Ilya Sutskever, an OpenAI co-founder, runs Safe Superintelligence Inc. Sam Altman has written that OpenAI is, above all, a superintelligence research company. When the builders name their organisations after where the technology is going, the rest of us can reasonably use the same word.

Already superhuman, in narrow domains

The objection that superintelligence is premature deserves a direct answer, and the useful distinction is between narrow and general. A system that only has to drive, and drives better than a person, is superhuman at that task.

On the road. Waymo reports more than 270 million fully autonomous miles through June 2026 across five metropolitan areas. Compared with human drivers in the same areas, it reports 95% fewer crashes causing serious injury or worse, and 82% fewer injury-causing crashes, estimating that at least 841 people avoided injury as a result. Those are the company’s own figures and should be read as such, but the direction is hard to dismiss.

In the laboratory. The 2024 Nobel Prize in Chemistry went in part to Demis Hassabis and John Jumper for AlphaFold2, credited with solving a fifty-year-old problem by predicting the structure of virtually all 200 million proteins researchers have identified.

In mathematics. In July 2025 an advanced version of Gemini Deep Think achieved a gold-medal standard at the International Mathematical Olympiad, solving five of six problems, graded by the competition’s own coordinators.

Superhuman driving safety, Nobel-recognised science and Olympiad-level mathematics are not imitations of intelligence. Whatever you call the category, the word artificial is doing it no favours.

Where this turns into an electricity problem

Huang’s phrase deserves to be taken literally. These facilities take in electricity, chips, cooling and capital, and produce computation at industrial scale. Like the steel mills and auto plants of earlier eras they need land, construction crews, electricians, transmission lines and enormous quantities of reliable power.

The forecasts all point the same way.

  • Lawrence Berkeley National Laboratory, in its 2025 update for the Department of Energy, estimates data centres could use about 11.8% of total US electricity by 2030, within a range of 9.5% to 15.3%.
  • The International Energy Agency projects global data centre electricity consumption reaching roughly 945 TWh by 2030, and has noted that in the US data centres account for close to half of electricity demand growth through 2030.
  • PJM, the largest grid operator in the country, forecasts a summer peak of 222,106 MW in 2036, a ten-year increase of 65,733 MW. Data centres are the main driver.

Those forecasts are already colliding with reality. PJM proposed a one-time Reliability Backstop Procurement to secure roughly 6.8 GW for the 2028/29 delivery year. On September 29 FERC accepted parts of it and suspended it until February 28, 2027, citing cost allocation, transmission-owner exit rules and collateral requirements. We covered that decision and what it means in a separate piece.

Meanwhile the largest buyers are contracting directly. Amazon and Constellation signed a 20-year agreement covering 690 MW at Calvert Cliffs. Meta announced agreements for up to 6.6 GW of nuclear capacity. Microsoft’s agreement supports restarting the 835 MW Crane plant. The companies building this are locking in firm, around-the-clock power for decades, and doing it now. The state-level rules reshaping that are covered in our piece on large-load tariffs.

The honest caveats

The safety accord signed at that luncheon is voluntary. It commits the companies to internal controls, external auditors and board-level oversight, and critics reasonably note that it lets firms define safety on their own terms.

There is also a community dimension. Polling this month found a clear majority of registered voters opposed to data centres being built in their own areas, with power costs a significant part of that concern. That is not a fringe view and it will shape what gets approved.

None of which changes the demand curve. The buildings are going up and the power has to come from somewhere.

What a commercial energy buyer should do

  • Know your capacity exposure. In PJM and other tight markets, capacity and transmission are a growing share of the bill. Find your capacity tag, understand how it is set, and model what it could cost over three years.
  • Revisit your contract horizon. The largest buyers are signing twenty-year agreements. You almost certainly do not need twenty years, but a habit of one-year renewals leaves you fully exposed every twelve months.
  • Manage your peaks. Demand response, load shifting, storage and on-site generation all reduce peak-driven charges, and each becomes more valuable as capacity tightens.
  • Look at firm long-term options. Nuclear, uprates and other around-the-clock resources are back on the table, and mid-sized buyers can sometimes participate through aggregation or retail structures.
  • Follow the calendar. The FERC and PJM proceeding runs to February 2027, and the next base capacity auction follows. Those decisions determine who pays for new capacity.
  • Budget for 2027 now. Do not wait for a renewal notice to discover what the market has done.

The bottom line

Names matter because they shape how people plan. Artificial intelligence made this sound like a clever imitation of something human. Whatever label wins, the technology is already superhuman in narrow domains, is attracting hundreds of billions of dollars a year, and is reshaping the American power grid.

You can embrace the new term or roll your eyes at it. The demand curve behind it is real either way, the factories are being built, and someone has to pay for the power.

If you want a clear view of your capacity exposure and contract structure before that reaches your bill, that is work we do at Delta Edge CI.

Primary sources

Pin It on Pinterest

Share This