Nobody has a budget for your robot. They have a budget for the thing it replaces.

Every hardware company eventually runs the same experiment. It builds a machine that is genuinely better than what the customer uses today, proves it on site, wins the engineer, and then watches the deal sit for two quarters while somebody looks for money that was never budgeted. The machine was never the problem. The purchase had no home in a plan that was written before the seller arrived. The practice behind this note lives at GTM consulting for deep tech and hardware.

The largest robotics commitment of the year was not a robotics purchase.

Put those two sentences next to each other and the commercial structure is obvious. HII was already growing an outsourced-work line by 30 percent. Two robotics companies were priced into that line as suppliers of shipbuilding work rather than as vendors of welding and sanding equipment. Nobody at HII had to open a new capital line for physical AI, argue it against a submarine program, and defend it to a board. The line was already open and already growing, and the only question was who fills it.

A buyer with no budget line for your category is not a slow buyer. They are a buyer you have not found yet.

Three companies, three lines that already existed.

HII: the outsourced labour line. More than 2.5 million hours in 2026, up 30 percent. The robotics companies are paid against milestones and against cost, schedule and quality performance, which is how a shipbuilder already pays a subcontractor. The instrument is familiar to the buyer even though the technology is not.

Monumental: the subcontractor line. The Amsterdam company raised a $32 million Series B led by Khosla Ventures (company announcement, July 2026) and does not sell bricklaying robots at all. Contractors pay for completed wall. That is the line a general contractor has been buying from masonry subcontractors for a century. The robot never appears in the budget, because the budget was never for robots.

Antora: the energy line. Antora closed a $550 million Series C on July 30, 2026, and the deployment that makes the case is at POET's Big Stone City bioprocessing facility: a 5 gigawatt-hour multi-day thermal storage system, more than 200 thermal batteries, delivered from initial construction to commissioning in under twelve months, all figures from Antora's own announcement. An industrial site does not have a thermal-battery line. It has an energy line, and it has a schedule. Antora sold against both.

None of the three asked a buyer to invent a category inside their own plan. Each found an operating line the buyer already defends every year, and sold in that line's units: hours, finished wall, energy on a date.

Why the new-line sale is the one that dies.

It dies at a handoff, not at a demo. Median pilot-to-production conversion across hardware is 12 percent (IDC 2025, in Hardware GTM Benchmarks 2026). The failure is rarely technical. A pilot funded from an innovation or R&D budget has to find a production budget owner who was not in the room and who inherits an operating commitment they did not choose. A pilot funded from the operating line it is meant to reduce has no such handoff, because the person who owns the line is already the person paying. The full anatomy is in pilot to production.

It dies on the calendar. The manufacturing baseline deal runs $48K on a 124-day cycle at a 19 percent win rate (Digital Bloom), and capital-equipment purchases run far longer than that because they attach to an annual capital cycle rather than to a quarter. A new capital line has a filing date, and if a seller arrives after it, the earliest possible close is next year no matter how good the pilot went. An operating line has no such gate. It has a manager with discretion.

It dies in a room the seller never covered. The modal B2B buying committee runs 6.3 to 6.8 people (Gartner and 6sense composite), and a new-category capital request adds seats rather than removing them: finance, who has no comparable to price it against; operations, who inherits the maintenance; and whoever loses budget to fund it. Selling into an existing line shrinks the room, because the line already has an owner and a precedent. The seats are mapped in the hardware buying committee.

This is the general case of the argument in why hardware doesn't sell like SaaS. Software found the seat licence, a unit every buyer already understood, and priced into it. Hardware keeps inventing units nobody budgets in.

Four moves to find the line that already exists.

One: ask what the work costs today, not what the problem costs. Problem framing produces an impressive number nobody owns. Cost-of-work framing produces a smaller number attached to a named manager, a code, and a plan. HII's 2.5 million hours is a cost-of-work number. A seller who had opened with the total cost of the skilled welder shortage would have been talking to public affairs.

Two: price in the buyer's units. Finished wall. Hours. Energy delivered on a date. Parts qualified. When the unit matches the line, the comparison is arithmetic the buyer can do themselves, and the purchase stops being a technology decision. When it does not match, the buyer has to build the translation, and buyers do not do unpaid work for vendors.

Three: put the pilot on the budget that would fund the rollout. This makes the pilot harder to start and it is still the right trade. An innovation-funded pilot converts at the rate above. A pilot paid from the operating line has already answered the only question that matters at conversion, which is whose money this becomes at scale.

Four: decide who carries the working capital, before the first proposal. Outcome pricing moves the purchase into an operating line and shortens the approval path. It also means someone finances the fleet until the outcomes pay it back, and that someone is usually the seller. That is a balance sheet decision made at the founder level, not a pricing page decision made by a sales lead. Getting it backwards is how a company wins its first ten deals and runs out of cash on the eleventh.

Three questions, answered straight.

Which budget line pays for a hardware purchase?

Almost never a new one. In practice the money comes out of a line the buyer is already defending in this year's plan: outsourced labour hours, contract services, maintenance and repair, energy, quality rework, or an existing capital replacement schedule. HII's commitment of up to $900 million to Path Robotics and GrayMatter Robotics across seven years sits against outsourced shipbuilding work, a line HII said it plans to grow beyond 2.5 million hours in 2026, a 30 percent increase over 2025. The seller's job is to find which existing line the buyer would rather spend, then price into that line's units rather than into a capital request nobody has filed. The firm-level version of that engagement is described on the GTM consulting page; the earlier-stage version is go-to-market consulting for hardware startups.

Why do hardware pilots stall even when the technology works?

Because a successful pilot funded from an innovation or R&D budget still has to find a production budget owner who was not in the room, and that handoff is where median pilot-to-production conversion falls to 12 percent (IDC 2025). A pilot paid for out of the operating line it is meant to reduce does not have that handoff, because the person who owns the line is already the person paying. The structural fix is to run the pilot on the budget that would fund the rollout, even if that makes the pilot harder to start.

Should hardware be priced as capex or as an outcome?

It depends on who is being asked to sign and what they already buy. Outcome pricing moves the purchase into an operating line and shortens the approval path, which is why Monumental sells completed wall rather than bricklaying robots. It also moves the financing burden onto the seller's balance sheet, because someone has to fund the fleet until the outcomes pay it back. Capital pricing keeps the seller's cash cycle clean and pushes the buyer into a committee of 6.3 to 6.8 people (Gartner and 6sense composite) and a capital calendar the seller does not control. Neither is right in general. The question is which line the buyer already defends, and whether the seller can carry the working capital the other structure requires.

Keep reading.

This note sits inside a published method. Start with the page that ties it together, or go straight to the piece that names your stall point.