Load patterns
Four patterns recur across the category. The list is not exhaustive — each estate’s actual profile is some subset plus additions specific to the family — but the framework applies generally, and the engineering response to each pattern is similar enough to design against.
Working spaces. The garage where the principal restores cars, the maker’s space for design and fabrication, the estate winery, the commercial-scale kitchen on estates that host seriously. Industrial equipment dominates the peaks: welders 8–16 kW under load with significant strike inrush, plasma cutters 12–25 kW, CNC mills and lathes 5–30 kW depending on size, air compressors 5–15 kW continuous during recharge cycles. Three-phase service is often warranted, dedicated subpanels always. Refrigeration dominates the continuous baselines where present: winery cold storage and walk-in coolers constant year-round, with seasonal peaks of 30–50 kW during harvest crush or major event hosting. Maker’s spaces sit between the two profiles — lower peaks (laser cutters 1–3 kW, bench electronics in the hundreds of watts, large-format 3D printers 0.5–1 kW) but longer duty cycles, prints running for hours or days.
Creative and scientific installations. The observatory with its telescope, cryocoolers (1–2 kW continuous), and dome mechanics. The recording studio whose climate control for acoustic isolation often draws more than the equipment itself (3–8 kW continuous). The art conservation studio with precision humidity. The home laboratory with fume hoods and analytical equipment. The principal’s serious avocations outside their day occupation. Aggregate draws are modest by working-space standards; what these installations share is stringent power-quality and reliability requirements that affect classification rather than service size — these become priority loads in the microgrid’s load classification, addressed below.
On-premises compute. The principal’s GPU cluster for personal model training, the local-LLM environment that hosts a distilled model the household uses without cloud dependency, the rendering or simulation cluster for serious technical projects. A workstation-class GPU draws 400–700 W under load; a 4-GPU server runs 2–4 kW continuous when training; a rack of inference servers for serious local AI work draws 5–15 kW continuous. Cooling overhead adds 30–50% to the electrical load year-round. The cumulative impact is a continuous multi-kW load with architectural footprint reserved in the equipment-room program from the start — not retrofitted.
Land-extended operations. On estates with substantial property. A working horse barn at 3–15 kW depending on equipment (climate, lighting, water heating, treadmills, hydrotherapy, veterinary equipment for serious operations). Irrigation pumping — often the single largest grounds-related load at 5–30 kW during operation, depending on water source and pressure requirements. Well pumps on private water. Greenhouse climate. Working orchards or vineyards adding processing peaks during harvest. The land-extended loads scale with property size and operational ambition.
The substrate compute distinction
The on-premises compute above is the household’s own — chosen by the principal because they are the kind of person who chooses it. Adjacent to it but structurally different is the substrate compute that runs the residence’s operational systems: EstateAI’s reasoning layer, the digital twin’s database servers, the operations console, the local inference that processes security feeds, the edge compute the substrate requires to function.
Similar electrical profile; different operational classification. Household compute is discretionary — valuable, often substantial, but shed-able in the worst-case islanded scenario where the family chooses household function over personal pursuits. Substrate compute is essential. It stays powered through islanded operation at the same priority as life-safety and security. The two often share equipment-room space and cooling infrastructure but belong to different categories. Instrument them separately at the panel so the microgrid controller can shed household compute during extended islanded scenarios while maintaining substrate compute indefinitely.
The decisions that matter
Four decisions in handling this load category have consequences worth surfacing at the design level.
Honest characterization at design. The family describes pursuits specifically, not abstractly — not “we’ll have a workshop,” but the actual equipment, the patterns of use, the simultaneity assumptions, the realistic projection of how the pursuits will evolve. Vague descriptions produce undersized installations that surprise the household within the first year of operation. Treat the family’s pursuits as architectural input data on the same footing as the room program.
Load classification. The microgrid classifies every load as essential, priority, or discretionary. Classifications are set deliberately for each load based on the family’s actual values — the observatory cryocooler may be priority because loss damages equipment; the maker’s space may be discretionary because it can be paused; winery refrigeration is essential during the periods when inventory is at risk and priority otherwise. Panels are instrumented so the microgrid controller can act on the classifications during islanded operation or peak shaving.
Dedicated electrical infrastructure. Subpanels, dedicated runs from the main equipment building, three-phase service where the equipment warrants, equipment-room space for the compute racks and their cooling. The commitment is made at design rather than retrofitted. The cost difference at design is modest; retrofitted, it is an order of magnitude larger.
Future-proofing for the pursuits that haven’t arrived. The household’s load profile in 2026 is not the profile of 2031. Electrical service headroom and conduit pathways for additions are sized the same way the main service is sized for fleet expansion. These loads tend to grow over the residence’s life; the energy system should be ready for the growth.
EstateOps
The household’s pursuits become part of the estate’s operational life rather than electrical surprises the system has to accommodate — instrumented through the substrate, modeled in the digital twin, dispatched by EstateAI against the household’s actual patterns, shed by load classification when the islanded scenario demands it. The estate’s energy system that handles only the household and the fleet handles only what is generic about the residence. The working and pursuit loads are what make the estate the place this particular family actually lives.