Tuned mass dampers are old news — Taipei 101's 660-ton pendulum sways to cancel wind loads. But a pendulum is reactive and sloppy. A control moment gyroscope (CMG) is active and surgical: spin a massive rotor, then tilt its axis, and the resulting precession torque can push back against ground motion in any direction. The ISS uses four 100-kg CMGs to point itself without burning fuel. What if we scaled one up to defend an entire city block?

Picture a 50-meter-diameter steel rotor buried in a vacuum chamber beneath downtown Tokyo, spinning on magnetic bearings.

How much angular momentum do we need?

A magnitude 7 earthquake delivers peak ground accelerations near 0.5 g over ~20 seconds. For a 200,000-ton skyscraper, the lateral inertial force during a 0.3-g jolt is roughly F = m·a = 2×10⁸ kg × 3 m/s² = 6×10⁸ N. If the building's center of mass sits 100 m above ground, the overturning torque is τ ≈ 6×10¹⁰ N·m.

A CMG produces a precession torque τ = H × ω_gimbal, where H is the rotor's angular momentum and ω_gimbal is how fast we tilt the spin axis. Gimbal at a brisk 1 rad/s, and we need H = 6×10¹⁰ N·m·s.

Sizing the rotor

For a solid steel disk of radius R, thickness t, density 7850 kg/m³: H = ½ M R² · Ω. A 25-m-radius, 2-m-thick disk weighs 30,800 tons. Spin it at 600 rpm (Ω ≈ 63 rad/s) — well below steel's burst limit around 900 m/s rim speed (we'd hit 1,570 m/s, so we'd need maraging steel or carbon-fiber overwrap, or drop to 250 rpm and double the mass). Either way:

H = 0.5 × 3.08×10⁷ kg × 625 m² × 63 rad/s ≈ 6×10¹¹ N·m·s

That's 10× our requirement — enough headroom to stabilize a whole block of mid-rise buildings tied to a shared foundation raft.

The energy and the catches

• Stored kinetic energy: E = ½ I Ω² ≈ 1.9×10¹³ J — about 5 GWh, or a Hiroshima-bomb's worth. Containment failure is non-negotiable: this thing needs a kilometer of reinforced concrete shielding, or better, a deep shaft that vents radially into rock.
• Spin-up time: At a generous 100 MW input, charging takes ~50 hours. You keep it spinning continuously; magnetic bearings in vacuum lose maybe 10 kW to eddy currents.
• Gimbal actuators must apply τ_gimbal = H × ω_earthquake. For typical seismic frequencies (1 Hz, ω ≈ 6 rad/s), that's 3.6×10¹² N·m on the gimbal bearings — utterly absurd. The trick: use multiple smaller CMGs in a scissored pair configuration, where reaction torques cancel internally and only the desired output reaches the foundation.
• Saturation: A CMG can't push forever in one direction without its gimbal hitting stops. After a 20-second quake, it needs to "desaturate" against a secondary system — likely deep rock anchors with hydraulic rams.

For comparison: 30,800 tons of steel is ~$30M raw, but the vacuum chamber, magnetic bearings, gimbal structure, and seismic isolation push the bill toward $2–5 billion per installation. That protects maybe $50B of real estate. The math nearly works — but base isolation (rubber bearings under each building) does 80% of the job for 1% of the cost.