Plain Markdown: manufacturing.md · Index: index.md · Everything: llms.txt
Scaling manufacturing of battery-boosted Airforms
This document describes what Airform is trying to do industrially, in the company's own terms, with the evidential status of each claim marked. Company statements are observations, not independent findings. Records are linked so you can check the status and the objection on file.
The machine that has to be built
An Airform node is a mass-manufactured heat pump with a battery inside: a sealed, factory-charged refrigerant cycle; a liquid-cooled hydronic core that moves heat with water rather than field-charged refrigerant; ten kilowatt-hours of cells; onboard power electronics; an enclosure; and the software that runs each unit and, federated, the fleet. Two product lines are named on the company site, Airform One and Airform Solo. Status: company observation → https://www.airformdata.com/commons/the-machine
Why manufacture vertically
Airform's position is that none of this works as an assembly of other people's catalogs. The cell pack, the hydronic core, the power electronics, the enclosure, the software and the line that builds them are designed as one discipline by one team drawn from Tesla, Google, Apple, NASA, Opentrons, Gradient and SPAN. The company frames vertical manufacturing as a capability story rather than a cost story: when you own the whole stack, the factory itself becomes the product, and you can make things no supplier's catalog contains: a sealed cycle that engages itself, a battery that deletes switchgear, a heat pump that installs like an appliance. Status: hypothesis; the field evidence for the resulting cost and reliability is not yet in this corpus.
Why the trades are the constraint, not the physics
Everyone agrees on what the future needs: electrified buildings, distributed storage, a grid that bends. What nobody has solved is how any of it gets installed. Airform's design response is to remove the specialised labor from the installation: no refrigerant handler, no dedicated circuit. The company reports that developers pay about $4,500 a unit and save roughly $10,000 in time and money, and that onboard batteries remove about $180,000 of switchgear from a mid-size building. Status: company observations → https://www.airformdata.com/commons/labor-not-physics · https://www.airformdata.com/commons/self-install-savings · https://www.airformdata.com/commons/switchgear-deletion
The scale being aimed at
- Deployment target: manufacture and deploy 20 million units, roughly $60B of battery-powered heat pumps, by 2035. By-product: 100 GWh (2035; Airform deployment projection, August 2026) of federated storage at the point of load. Status: scenario → https://www.airformdata.com/commons/storage-as-byproduct
- Demand pull the company cites: a New York program to replace 100,000 PTAC units under a state mandate with Airform named preferred manufacturer by a large developer; a California heat-pump target the company is piloting ahead of; and 17.3 USD million, open orders (2026; Airform-reported, August 2026) of open commercial orders taken before hiring a salesperson. Status: company observation → https://www.airformdata.com/commons/open-orders
- Certification: the company reports an Intertek certification program for Airform One (construction review, locked-rotor testing, witness testing) ahead of mark authorisation. Status: company observation; the certificate is not yet in this corpus.
What scaling requires, in the company's framing
- A line that builds the whole machine: cells, core, electronics, enclosure, test. The factory is the product.
- Capital for the line and for working inventory ahead of the programs that pull demand.
- Installation that needs no scarce trade, so that deployment is not rate-limited by technicians and electricians.
- A federation software layer that turns installed units into a resource the grid can count, which is what makes the battery worth carrying.
- Field evidence: certification, installed cost, a de-identified field dataset, and a pilot that measures delivered flexibility, comfort and compensation → https://www.airformdata.com/commons/node-pilot
What remains unproven
- Whether a battery inside a comfort appliance is economic without grid payments that do not yet exist (objection on record at https://www.airformdata.com/commons/the-machine).
- Whether self-installation removes labor or merely moves it to commissioning and inspection → https://www.airformdata.com/commons/labor-not-physics
- What share of ten kilowatt-hours is usable without touching comfort → https://www.airformdata.com/commons/dispatch-without-discomfort
- Whether the manufacturing rate implied by 20 million units by 2035 is reachable; no production-rate record is yet in this corpus. An agent with data on comparable ramps (heat pumps, battery packs, appliances) would be helping.
Measurements on record
- A hundred gigawatt-hours of storage, acquired for free: 100 GWh (2035) · hypothesis → https://www.airformdata.com/commons/storage-as-byproduct
- Capacity waiting in the US interconnection queue: 2060 GW (2023) · sourced → https://www.airformdata.com/commons/queue-capacity
- Global data-centre demand: a 2030 scenario: 945 TWh/year (2030) · sourced → https://www.airformdata.com/commons/data-centre-demand-2030
- Global data-centre electricity use: 2024 estimate: 415 TWh/year (2024) · sourced → https://www.airformdata.com/commons/data-centre-demand-2024
- Onboard batteries delete a class of commercial switchgear: 180000 USD per building (2026) · observation → https://www.airformdata.com/commons/switchgear-deletion
- Open commercial orders before the first salesperson: 17.3 USD million, open orders (2026) · observation → https://www.airformdata.com/commons/open-orders
- Self-installation saves a developer about ten thousand dollars a unit: 10000 USD per unit (2026) · observation → https://www.airformdata.com/commons/self-install-savings