Gravitational Upper Bounds on AGI/ASI from the Seidel–Bostrom Event Horizon Bound: Why Artificial Superintelligence Cannot Scale Indefinitely

The current form of the Seidel–Bostrom Event Horizon Bound (SBEB, Seidel 2025b; DOI:
10.5281/zenodo.17505947) demonstrates that the physical realization of polynomial-time
NP interventions inevitably leads, beyond a specific threshold, to the formation of an event
horizon that causally isolates the computation. This paper derives a direct consequence
for Artificial General Intelligence (AGI) and Artificial Superintelligence (ASI): Even in
distributed architectures, the coherence and synchronization required for global computation
enforce an effective localization of energy within the causal diamond of the final reduction.
Once the Schwarzschild condition is satisfied, an event horizon forms and isolates the
computation. Therefore, AGI/ASI systems are not indefinitely scalable; gravitation imposes
a hard, system-specific upper limit on computational growth. We formulate the assumptions,
sketch the derivation (local density condition, causal volumes, synchronization energy), and
discuss implications for both real and simulated universes.