Resiliency and Operations: Overview
How the Kronos AI stack keeps the breeder (Hyperion) and burner (Aegis / MetroVolt) running safely when components, sensors, or models fail.
What resiliency means here
Resiliency is the property that lets a plant survive faults without violating a safety limit and without losing more availability than necessary. In the Kronos stack it is not a single module but a discipline that cuts across control (L1), the digital twin (L3), diagnostics, and fleet coordination (L7). This category documents the timing model that bounds how fast the stack must react, the operating narratives that describe a shot and a campaign, the failure-mode analysis that enumerates what can go wrong, the failover and graceful-degradation logic that keeps power flowing, and the availability model that turns all of this into a number.
Two very different machines share the framework. The breeder (Hyperion) is a D-T spherical tokamak: Q_sci 3.076, 85.0 MW fusion power, 9.66 MA plasma current, 16.84 T peak field (8 T on-axis), negative triangularity delta -0.30, tritium breeding ratio treated as a design lever across 1.1/1.5/1.8, producing a ~4 kg/yr tritium class output, ~1.97 kg/yr helium-3, and 14 MeV neutrons. The burner (Aegis / MetroVolt) is a D-3He tandem-mirror generator with a 26.49 T plug, 17 T throat, 5.44% neutron fraction and direct energy conversion. A tokamak lives shot-to-shot; a mirror aims at continuous export. Their resiliency problems are therefore different in kind, not just degree.
The honest availability gate
The central design-and-simulation finding for the burner is that modelled availability sits at roughly 0.86-0.995, while a hyperscale Tier III data-center target is 0.99982. That is a 30-100x shortfall in tolerable downtime. We state this plainly across the category: a single MetroVolt unit is not a sole Tier-III source today. It informs redundancy, spares, and fleet strategy rather than being hidden. See The Availability Gate.
Honest framing
These machines are design and simulation studies. Breeder construction begins Q2 2027 with first-of-a-kind (FOAK) first tritium targeted ~2030; NOAK and BOAK follow. No hardware net-gain is claimed before FOAK. Every resiliency claim here is exercised against the digital twin and fault-injection simulation, with a defined path to instrumented validation once units exist.