Normallab · Working Note · August 2026

Twenty million heat pumps.
One hundred gigawatts.

A plan to mass-manufacture battery-integrated heat pumps, engineer the labor out of installing them, and turn America's biggest heating goal into its largest distributed battery.

Internal · adjust any slider — every number on this page recalculates.

Distributed battery power
105 GW
across the deployed fleet
Heat pumps deployed
20M
by 2030
Energy storage
315 GWh
at 3h duration
Install savings unlocked
$117B
soft costs engineered out

The grid is being asked to do two enormous things at once: absorb record new load, and electrify heating for tens of millions of buildings. A coalition of 24 governors — representing over half the U.S. population — has pledged 20 million heat pump installations by 2030. The machines are not the bottleneck — the labor is. Soft costs of HVAC and electrical work routinely run two to three times the equipment itself, and every install competes for the same scarce electricians and panel capacity the grid buildout needs.

Our plan attacks the labor, not just the hardware. It runs in three steps, and each step compounds the one before it.

The counter-demand thesis, woven through everything below: data centers and re-industrialization are creating an enormous demand for power — and for the electricians, transformers, and switchgear that deliver it — that is not being served. Battery-powered heat pumps are the counterweight: they meet heating demand without competing for that constrained equipment and labor, and every install adds storage and flexibility that makes the grid more resilient for the new loads too.
The mandate & the math — reasoning
NERC now forecasts U.S. summer peaks growing 224 GW (+24%) in ten years — its largest jump on record — with winter peaks up 246 GW. New firm-ish capacity has to come from somewhere.N
24 governors — about 55% of the U.S. population — have already committed to 20 million heat pump installs by 2030. The demand signal exists; the delivery system doesn’t.CACA
The counter-demand is unserved and compounding: data center power demand grew ~22% in 2025 alone and is on track to nearly triple by 2030, while new manufacturing pushes national load forecasts up for a third straight year — all of it competing for the same transformers, switchgear, and electricians buildings need.SPGST
Both problems are gated by the same scarce resource: skilled field labor. A plan that only makes more machines doesn’t move either number — a plan that removes install labor moves both.SQ

Where the mandate lives

The 24 U.S. Climate Alliance governors behind the 20-million-unit pledge — about 55% of the U.S. population
AKMEVTNHWAIDMTNDMNWIMINYMARIORNVWYSDIAILINOHPANJCTCAUTCONEMOKYWVVAMDDEAZNMKSARTNNCSCGUOKLAMSALGAHITXFLPR
Climate Alliance pledge state / territory Not (yet) in the coalition

One platform, four markets

The same liquid-cooling, battery-integrated architecture ships into four segments. Each one is its own page — with its own reasoning and evidence:


Step 01

Engineer out the soft costs

We are building America's most integrated all-in-one liquid-cooling heat pump. No field-built refrigerant linesets, no on-site charge work, no bespoke ductwork engineering — the thermal loop is sealed and tested at the factory. The unit places like an appliance, not a construction project. When the machine absorbs the craft, install hours collapse, and soft costs collapse with them.

Install economics

What one install saves when the factory does the field's work
Saved per install
$5,830
34 crew-hours saved per install
Across the fleet
$117B
at 20M units
Labor freed up
680M hrs
crew-hours returned to the trades
Why step 1 makes sense — evidence
About 70% of a heat pump’s installed cost is labor and related soft costs, not the machine — so cost reduction has to happen in the field, and the cheapest field hour is the one the factory already did.SA
Permitting and admin alone run ~7% of project cost (≈$2,173 in parts of the Bay Area). A sealed, factory-charged appliance — no refrigerant work on site — collapses both hours and process.S
The HVAC trade can’t simply hire its way out: the workforce shortage now extends to the instructors who train new technicians. Engineering hours out of the install is the scalable path.AN

Where the install dollars go

One installed unit, conventional vs. ours — equipment held equal at $6,500, driven by the sliders on the left
Equipment Field labor Electrical upgrade (expected)
$0$5,000$10,000$15,000$20,000Conventional$13,050Normallab$7,220
View data table
EquipmentField laborElectrical upgrade (expected)Total
Conventional$6,500$4,800$1,750$13,050
Normallab$6,500$720$0$7,220
Step 02

Put a battery in every unit

The second bottleneck is electrical: panel upgrades, service upgrades, utility queues, and the electricians to do all of it. A battery-integrated heat pump sidesteps the blockage — the battery covers the surges while the unit draws a steady trickle, so it connects to the panel a building already has. No upgrade, no queue, no wait.

And the batteries don't just unblock installs. Deployed millions of times, in SKUs from 5 kW to 25 kW, they add up to utility-scale storage — placed exactly where the load is.

Fleet capacity

Set the deployment, the SKU mix, and the battery sizing — see what the fleet becomes
Fleet battery power
105 GW
≈ 14% of U.S. peak demand · ≈ 1.9× today’s utility-scale fleet
Fleet energy
315 GWh
≈ 53% of projected U.S. storage by 2030
Avg battery per unit
5.3 kW
avg unit rating 10.5 kW
SKU mix shares are normalized automatically. U.S. peak demand ≈ 745 GW.
Why step 2 makes sense — evidence
Panel and service upgrades cost $2,000–$4,500 and stall projects — yet LBNL found 96% of homes with 100-amp panels could add a heat pump without upsizing if peaks are managed. A built-in battery is that management, in hardware.S
There is no electrician surplus coming: BLS projects ~81,000 openings a year while the trade loses ~10,000 and gains ~7,000 annually. Every panel upgrade we design out is an electrician-day the grid buildout gets back.Q
The pattern is proven at small scale: battery-equipped 120V induction stoves already ship precisely to dodge panel upgrades. We apply the same physics to the biggest load in the building.C
The U.S. utility-scale battery fleet reached ~52 GW by mid-2026 after growing ~70%/yr — proof of demand for storage. Heat pump deployment at 20M units piggybacks another ~100 GW on installs that were happening anyway, sited exactly where load lives.EBS

Where the gigawatts come from

Fleet battery power contributed by each SKU size
012.52537.550GW205 kW3010 kW2315 kW2020 kW1325 kW
View data table
SKUShareBattery power
5 kW40%20.0 GW
10 kW30%30.0 GW
15 kW15%22.5 GW
20 kW10%20.0 GW
25 kW5%12.5 GW
Total105.0 GW

The gigawatt context

Our fleet at your settings, against the grid numbers that define the decade — all in GW
0 GW63 GW125 GW188 GW250 GWOur fleet (your settings)105 GWU.S. utility-scale battery fleet, mid-202652 GWDOE virtual power plant need by 203080–160 GWNERC 10-yr summer peak demand growth224 GW
View data table
BenchmarkGW
Our fleet (your settings)105
U.S. utility-scale battery fleet, mid-2026 (EIA)52
DOE virtual power plant need by 2030 (DOE)80–160
NERC 10-yr summer peak demand growth (NERC)224
Step 03

Turn a cost into an asset

A heat pump today is bought as a cost. Fleet-managed, it becomes an asset. Every unit is a thermal battery and an electric battery in one box. We coordinate them at the building level — shaping a whole building's thermal-electric load — and at the grid level, shifting millions of small loads in concert to open capacity on the grid in a distributed way. That flexibility is capacity utilities pay for, year after year.

Grid value of the fleet

What coordinated flexibility earns annually
Annual grid value
$11B/yr
recurring, fleet-wide
Per unit, per year
$525
the payment that flips cost to asset
Over 10 years
$105B
before energy-bill savings
Why step 3 makes sense — evidence
DOE’s Commercial Liftoff analysis calls for 80–160 GW of virtual power plants by 2030 — saving ~$10B a year and covering 10–20% of peak. At default settings, our fleet alone lands inside that national target range.D
Distributed capacity is the cheapest capacity: DOE finds residential VPP capacity ~40% cheaper than a utility-scale battery and ~60% cheaper than a gas peaker; Brattle puts VPP net cost at 40–60% of alternatives.DB
Brattle: just 60 GW of VPPs would save utilities $15–35B over a decade, plus $20B+ in societal benefits. Those savings are the revenue stream that flips a heat pump from a cost into an asset.B

The ramp to 2030

Cumulative units deployed — an S-curve as factory lines come online in parallel
06.312.518.825Millions of units2026202720282029203020M
View data table
YearAddedCumulative
20260.4M0.4M
20271.6M2.0M
20284.0M6.0M
20296.4M12.4M
20307.6M20.0M

Why we can do this

Vertically integrated, in America

We are a vertically integrated U.S. manufacturer — design, manufacturing, and supply chain under one roof. That is not a slogan; it is the mechanism. When the team that designs the thermal loop sits next to the line that builds it, the install-hour problem becomes an engineering problem — and engineering problems get solved on factory timelines, not construction timelines.

Integration is also why we can run multiple SKUs in parallel, from 5 kW to 25 kW, sharing a common battery, controls, and liquid-cooling architecture. One platform, many sizes, one fleet. Re-industrializing this supply chain is how the cost curve keeps bending — and it is the moat.

Why us — reasoning
Every constraint above — install hours, refrigerant handling, panel limits, coordination — is fixed at design time, not install time. Only a manufacturer that owns design, production, and supply chain can iterate on all of them at once.SQ
The timeline demands it: the pledge window closes in 2030 and NERC’s load growth arrives on the same clock. Parallel SKUs from one platform is how 5 kW through 25 kW ship on factory timelines, not construction timelines.CAN
Segment · Residential

The appliance install

Homes are where the 20-million-unit pledge lives, and where soft costs bite hardest: on a $13,000 conventional install, the machine is a minority of the bill. Our 5–10 kW residential SKUs are engineered as appliances — factory-sealed liquid-cooling loop, no field refrigerant work, and a battery that lets the unit run on the panel and service the home already has.

That last part is the unlock. The residential bottlenecks — panel upgrades, service upgrades, electrician scheduling — are exactly the resources data centers and new factories are now bidding away. A home install that touches none of them is a home install that can happen this month, not next year.

Counter-demand: every residential install that skips a panel upgrade returns an electrician-day and a piece of distribution capacity to the grid buildout — while adding ~2.5–5 kW of storage behind the meter that rides through outages and shaves the evening peak.
Why residential works — evidence
LBNL: 96% of homes with 100-amp panels can add a heat pump without upsizing when peaks are managed — the battery is that management, built in. Avoided upgrades run $2,000–$4,500 each.S
The electrician shortage is structural — ~81,000 openings a year against a shrinking pipeline — and residential retrofits are the lowest-priority work when data centers pay overtime. Installs that need no electrician win by default.QSP
Battery-buffered 120V appliances are already a shipping product category (induction ranges) — the pattern is validated; we apply it to the largest load in the home.C
At residential scale the fleet is the VPP: DOE counts household devices as the core of the 80–160 GW of virtual power plants the grid needs by 2030.D
Homes that need no panel upsize
96%
of 100-amp homes, with managed peaks (LBNL)
Avoided upgrade cost
$2–4.5K
per home, plus the utility queue
Install labor
6 hrs
vs ~40 conventional — tune it in The Plan
Segment · Commercial

Shrink the switchgear, free the building

New commercial buildings are stalling for a reason that has nothing to do with concrete: electrical equipment. Data center demand has swallowed the world's supply of transformers and switchgear — medium-voltage gear now quotes at 52–80 weeks, substation transformers at up to 110. A building that needs a bigger service is a building in a queue.

HVAC is the reason services are sized big: space conditioning approaches half of a commercial building's energy, and its afternoon peak sets the demand the gear must carry. When our units take on that load with storage attached, the fleet controls the building's entire thermal-electric footprint — batteries carry the HVAC peak, the service is sized to the average, and the building drops one or two whole switchgear classes. Smaller, standard gear is gear you can actually buy.

Counter-demand: the same purchase order a data center places for 2,500 A metal-clad gear is the one an office tower is waiting behind. A building that peak-shaves its way down to standard distribution equipment exits that queue entirely — and hands the capacity it isn't using back to the utility as dispatchable flexibility.

Load footprint & switchgear calculator

One commercial building — what fleet-controlled heat pumps with batteries do to its electrical service
Service before → after
1,200 A → 800 A
1,069 A → 770 A actual · 480 V, 3φ, 0.9 PF
Peak shaved
224 kW
28% off the building’s peak
Procurement
−1 gear class
standard 16-wk gear vs 66-wk custom queue
Time to power-on
−3.2 months
powered on at month 18.0 vs 21.2
Returned to the grid
5.6 GW
across 25,000 buildings · ≈$560M/yr as VPP capacity
Why commercial works — evidence
Medium-voltage switchgear quotes at 52–80 weeks and substation transformers at 75–110 — vendors themselves warn that pre-pandemic “24–36 week” assumptions are fiction. Equipment, not construction, is the critical path.T
That queue exists because data center demand overwhelmed manufacturing capacity — S&P has data center power use nearly tripling by 2030 — and every commercial project now bids against hyperscalers for the same gear.TSP
HVAC is the size driver: space heating is ≈32% of commercial building energy, ventilation ~10% more. Own that load with storage attached and you own the building’s peak — and the service rating it forces.E
The shaved peak isn’t just avoided gear — aggregated across buildings it is dispatchable capacity worth 40–60% less to buy from us than from a peaker, per DOE and Brattle.DB

Lead times become schedule

Same building, two electrical paths — the switchgear order sets the power-on date
Design & permits Construction Switchgear procurement Energize & commission
mo 0mo 6mo 12mo 18mo 24Custom-gear buildingon at mo 21.2Normallab-controlled buildingon at mo 18.03.2 months sooner
View data table
ScenarioPhaseStartEnd
Custom-gear buildingDesign & permitsmo 0.0mo 6.0
Custom-gear buildingConstructionmo 6.0mo 16.0
Custom-gear buildingEnergize & commissionmo 19.2mo 21.2
Custom-gear buildingSwitchgear procurement (1,200 A custom · 66 wks)mo 4.0mo 19.2
Normallab-controlled buildingDesign & permitsmo 0.0mo 6.0
Normallab-controlled buildingConstructionmo 6.0mo 16.0
Normallab-controlled buildingEnergize & commissionmo 16.0mo 18.0
Normallab-controlled buildingSwitchgear procurement (800 A standard · 16 wks)mo 4.0mo 7.7

Peak control & the virtual power plant, in one picture

A design day — batteries carry the HVAC peak, refill overnight, and the shaved band becomes VPP capacity the grid can dispatch
With Normallab fleet control Uncontrolled building Shaved peak → VPP capacity
02505007501000kW12am6am12pm6pm11pm→ VPP800 kW peak576 kW
View data table
HourUncontrolledControlledTo VPP
0:00264 kW342 kW0 kW
3:00264 kW343 kW0 kW
6:00378 kW378 kW0 kW
9:00535 kW497 kW39 kW
12:00686 kW542 kW144 kW
15:00800 kW576 kW224 kW
18:00686 kW542 kW144 kW
21:00319 kW281 kW39 kW
Segment · Industrial

Process heat without the power fight

Re-industrialization is real — and it runs on heat. Process heat is up to 51% of on-site industrial energy, and less than 5% of it is electric today. Modern industrial heat pumps already reach 160 °C, cutting process-heat energy by up to a third with paybacks under two years in the right applications.

The blocker is the same one everywhere else: new factories are told the grid can't serve them, because data centers got in line first. Our answer is arrays of 25 kW battery-integrated units ganged into modular thermal plants: they draw steadily, store when power is cheap and available, deliver heat when production needs it, and never force the facility into a bigger interconnection than the wires can give it.

Counter-demand: a factory that electrifies heat through storage-buffered heat pumps adds manufacturing capacity without adding peak demand — and its battery fleet becomes reserve capacity the local grid can call on, hardening the same system the factory depends on.
Why industrial works — evidence
Process heat is up to 51% of on-site industrial energy and less than 5% electric — the largest un-electrified load in the economy, and most of its low-temperature share sits inside today’s 160 °C heat pump envelope.AC
Industrial heat pumps cut process-heat energy up to one-third with sub-2-year paybacks in the right applications — the economics work before any grid value is counted.AC
New factories face the same interconnection wall as data centers — load growth from manufacturing is a leading driver of the national forecast surge. Storage-buffered heat arrays let production grow inside the existing service.GSN
Of on-site industrial energy
51%
is process heat (ACEEE)
Electrified today
<5%
of process heat — the whitespace
Energy cut
by industrial heat pumps, paybacks <2 yrs
Segment · Agrivoltaic Land

Thermal loads at the grid edge

Agrivoltaics has quietly become real infrastructure: about 600 sites, 62,000 acres, and 10 GW of solar already share land with crops, grazing, and greenhouses. What that land mostly lacks is a productive local load — the power is exported into rural feeders that were never built for it.

Our units are that load. Greenhouses, dairies, poultry barns, and food processing all run on exactly the low-temperature heat our solar-paired SKUs deliver — and their batteries soak up midday solar the feeder can't carry, then serve heat and power through the night. Generation, storage, and thermal demand land on the same acre, so the energy never has to travel.

Counter-demand: rural feeders are the grid's weakest links and the last to be upgraded — the equipment went to the data centers. A farm that consumes its own solar through battery heat pumps needs less wire, not more, and keeps running when the feeder doesn't. That is resilience delivered from the edge in, not the substation out.
Why agrivoltaic land works — evidence
Agrivoltaics is scaling fast — 4.5 GW in 2020 to 10 GW across ~600 sites by late 2024 — but that generation feeds rural circuits with little local load to absorb it.NR
NREL reports communities adopt agrivoltaics seeking “energy resilience through solar plus battery storage” — precisely the package a battery heat pump completes: consume the solar on-site as useful heat, bank the rest.NR
Distributed capacity at the grid edge is the cheapest resilience there is — 40–60% below conventional alternatives — and it is the only kind that keeps a farm running when the feeder fails.BD
U.S. agrivoltaic solar today
10 GW
across ~600 sites, 62,000 acres (NREL)
Grew from
4.5 GW
in 2020 — more than doubled in four years
What farms buy first
Resilience
solar + storage cited as a leading community goal