Is Skynet Coming in 25 Years?

Yesterday President Donald J Trump invited a room full of the top AI tech leaders to dinner. I’m still wondering why? Myself and thousands of others today are wondering why he included Bill Gates in the mix. President Trump has often talked about Bill Gates in terms of him being an evil person. And there were others in the group that I personally don’t believe have the best intentions towards our country or humans in general. (Watch the Dinner Meeting)

I asked my “friendly” ChatGPT assistant to give me its thoughts of what our future might look like in twenty five years. I told it to report the vision as an imaginary reporter. Below is the Special Report by Briana from Oklahoma City with a vision of 2050.

If tech titans get their way, the American landscape of 2050 may swap cattle pastures for cooling towers. With AI data centers demanding a mind-boggling amount of energy, power plants, especially nuclear and renewables, could start popping up across the countryside. Some analysts believe AI could consume up to 20–25% of the nation’s electricity, sparking a national conversation about who gets the juice: humans or hyperscalers?

OKLAHOMA CITY, 2025 — In a sweeping rural landscape, your future knowing correspondent stands ready to guide you through a pressing story: how America's data centers,especially those fueling AI innovations, are becoming power-hungry superstars reshaping our electricity grid.

America’s electricity demand is on the rise, and AI-powered data centers are a major reason why. In 2023, U.S. data centers consumed about 4.4% of all electricity—roughly 176 terawatt-hours (Source). But that's just the beginning.

These numbers are expected to climb steeply. Projections suggest data center energy use in the U.S. could swell to up to 12% of total electricity by 2028, particularly if AI expansion continues its current trajectory. (Source and Source) Globally, power demand from data centers is expected to more than double by 2030—with AI as the major. (Source and Source)

This growth is pushing infrastructure to its limits. A study by Deloitte estimates that by 2035, U.S. AI data centers could demand about 123 gigawatts, a more than thirty-fold increase from today’s roughly 4 gigawatts. (Source) That translates to monstrous energy needs—some individual AI data centers are being designed to pull up to 2 gigawatts, on par with the output of large power plants. (Source)

With all that power being siphoned toward hyperscale AI servers, what does that mean for you? Well, right now, AI accounts for about 20% of U.S. data center electricity use, while traditional (or "legacy") uses still dominate the rest—roughly 80% in 2025. (Source) By 2035, that split may even out to 50–50.

But the bottom line is: this isn’t just a tech industry problem, it’s a community one. As power demand surges, household electricity bills are already creeping upward in regions dense with data centers, like Northern Virginia and upstate New York. (Source) And alongside energy strain, local communities face environmental and social pressures—from water usage to air pollution. (Source and Source)

Share

Summary Notes:
Using today’s best public forecasts as anchors, a reasonable central estimate is that US AI data centers could require ~1.5 petawatt-hours (PWh) of electricity in 2050—about 23% of projected national generation. Plausible bounds: ~1.0 PWh (≈15%) if growth plateaus after 2035, up to ~2.0 PWh (≈30%) under faster expansion. These ranges hinge on build-out of new generation, transmission, and efficiency gains.

Where we start (grounded facts):

• Deloitte estimates US AI data centers could reach 123 GW of power demand by 2035 (from ~4 GW in 2024). (Source and Source)

• The IEA says global data-center electricity use could double to ~945 TWh by 2030, with AI the dominant driver. (Source)

• The EIA’s AEO 2025 projects total US electricity generation around 6,000–6,650 TWh by 2050 (≈50% above 2025). (Source and Source)

Key conversion we use:
For steady 24/7 workloads, 1 GW ≈ 8.76 TWh/year at 100% utilization. AI data centers typically run very high duty cycles; using a 90% capacity factor gives ~7.884 TWh/GW-yr.

1. Anchor on 2035:
   123 GW × 7.884 TWh/GW = ~970 TWh in 2035 for US AI data centers. Deloitte

2. Project 2035 → 2050 (15 years):
   Because nobody has perfect visibility, we bracket three trajectories:

• Constrained/Flat build (Grid, permitting, efficiency gains dominate):
  Hold ~970 TWh through 2050.

• Central case (growth slows to ~3% CAGR**):**
  970 TWh × (1.03)^15 ≈ 1,510 TWh (~1.5 PWh) in 2050.

• High case (continued rapid expansion, ~5% CAGR**):**
  970 TWh × (1.05)^15 ≈ 2,015 TWh (~2.0 PWh) in 2050.

3. Compare to national power in 2050:
   Using 6,000–6,650 TWh as the EIA reference envelope:

• Flat case: ~15% (970 / 6,650).

• Central case: ~23% (1,510 / 6,650).

• High case: 30%+ (2,015 / 6,650). (Source and Source)

• Supply: Hitting the central/high cases requires massive new generation (nuclear SMRs, gas with CCS, wind/solar + storage) plus new transmission. The 2030s demand surge flagged by IEA and banks like Goldman Sachs is the near-term setup. (Source and Source)

• Efficiency: If model efficiency, better chips, and scheduling tame power per unit compute, we drift toward the flat/low case. If AI use cases explode faster than efficiency gains, we move toward the high case. (Source)

• Power requirement (AI data centers, 2050): ~1.5 PWh (central), with a range ~1.0–2.0 PWh.

• Share of US electricity (2050): ~23% (central), with a range ~15%–30%+.
  These are scenario estimates, not certainties—but they’re anchored to credible 2030–2035 waypoints and the EIA’s 2050 system size.

A portion of this content may be AI generated

Stay smart, keep your eyes and ears open — Briana