What Should Evergy Pay a Homeowner to Host an AI Node?
Start the bill credit at about $112 a month, half the peak capacity each node keeps off the grid. Then put noise, heat and service in writing.
A reader of my Playbook for Evergy asked the question the rider skipped. Anyone asked to put equipment on their house wants to know two things: what’s in it for them, and what’s the hassle. The rider promised hosts a bill credit and never put a number on it. This piece gives one, shows where it comes from, and treats the hassle as seriously as the money.
Start from the value the house creates
The case for flexible home compute rests on one figure. A node that drops from 12.5 kW to 1 kW when Evergy calls keeps 11.5 kW off the system peak. At about $234 per kW-year for new peaking capacity, that is worth about $2,691 a year per node.
That value exists because of the house. The node sits on an existing service, behind a panel that can throttle it, on a street where the transformer has room. Without the host there is no fast, flexible megawatt. So I would set the credit as a share of that value:
credit = s × (P_node − P_curtail) × C_cap
With s = 0.5, the host earns 0.5 × 11.5 kW × $234 ≈ $1,346 a year, about $112 a month, shown as a line on the Evergy bill.
A share of avoided capacity beats a flat rent for three reasons. Evergy can verify it from node telemetry. It rises if new capacity gets more expensive. And it never comes from other ratepayers: the aggregator funds it through the rider, and Evergy only passes it through.
Check it against three other numbers
The aggregator can afford it. A node with 16 GPUs at $1.80 per GPU-hour and 60% utilization grosses about $151,000 a year, so the credit is 0.9% of revenue. Five years of credits come to about $6,700, roughly 6% of the $111,000 per node that getting power in six months instead of three years is worth.
Homeowners already accept similar deals for less. Massachusetts’ ConnectedSolutions program pays $275 per kW for a home battery’s help during summer peaks, up to $1,375 a year for a 5 kW battery, across as many as 60 events. A battery owner gives up a few afternoons. A node host gives up a corner of the yard all year. The node credit should not come in below the battery check, and half the avoided capacity lands within $30 of it.
The vendor has hinted at the ceiling. SPAN’s chief revenue officer told pv magazine USA that hosts in the best-value areas could get discounted electricity “up to and including free electricity.” A typical Evergy residential bill runs about $112 to $141 a month depending on territory: $112 in Missouri Metro on EIA data, $136 in Kansas Metro and $141 in Kansas Central on the Kansas consumer advocate’s July 2026 tally. So the starting credit covers 80% to 100% of a typical bill, which puts it right where SPAN’s own pitch already sits.
The hassle needs rules, not more money
No credit buys back a bad night’s sleep. The rider should write the hassle into its terms, so hosts and their neighbors know what they are agreeing to.
Energy. The node’s power must never land on the host’s bill. SPAN describes the node’s load as sitting behind the home’s meter, so the rider should require a revenue-grade submeter and bill that energy to the aggregator. The house also comes first: when the host’s own load climbs, the panel throttles the node, never the air conditioner.
Noise. SPAN’s partners describe liquid-cooled, fanless parts inside the server, but the heat still has to leave the cabinet outdoors, and no decibel figure has been published. Kansas City, Missouri’s code (Sec. 46–4) sets outdoor residential limits of 60 dB(A) by day and 55 dB(A) at night, measured at the property line. The city’s enforcement test is loose for this purpose: a violation needs sound more than 5 dB above ambient, measured at least 50 feet from the source. The rider should drop those conditions and hold every node to 55 dB(A) at the property line around the clock, across all of Evergy’s territory. A node over the limit throttles until it complies, and a neighbor’s complaint counts the same as the host’s.
Heat. A 12.5 kW node turns about 300 kWh a day into heat, as much heat as a 3.5-ton central air conditioner pulls out of a house at full load, except that it never cycles off. The house can’t use that much heat outside deep winter, which is why the cabinet sits outdoors. Exhaust should point away from windows, decks and the neighbor’s house, with the same setbacks used for condensers.
Service. The aggregator owns, insures and services the cabinet. Visits stay outdoors, with 48 hours’ notice for scheduled work and a published response time for faults. Each node carries 16 RTX Pro 6000 Blackwell GPUs, which Nvidia listed at $13,250 apiece in June 2026, so a cabinet holds about $212,000 of GPUs. That makes it a theft target, so that risk sits with the operator, not the homeowner’s policy.
Exit. The host can leave with 90 days’ notice, and on a sale the buyer decides whether the node stays. Removal and site repair are the aggregator’s cost, backed by a bond Evergy holds, so a failed operator cannot leave a dead cabinet beside someone’s house.
The credit keeps flowing during curtailment and during outages the host did not cause. Compute risk belongs to the operator.
Where I’d start
$112 a month, written into the tariff as half of each node’s avoided peak capacity, paid on the Evergy bill and funded by the aggregator, with those five rules attached. The 10 MW pilots in Kansas and Missouri, about 800 homes each, are the place to test it. If homes sign up at that number, it is enough. If they don’t, exit surveys will show whether money or hassle held them back, and Evergy will want that answer before anyone scales.
Thanks to the reader who asked. The full rider, the economics and the Python model are in A Playbook for Evergy.


