OEM Procurement Intelligence — Gradient, Innova/Ephoca, and Midea
Purpose: Summarize what we know about three packaged-heat-pump OEMs spanning the market (a startup, an established niche player, and a global giant), and what their approach to pricing tells us about risk.
Scope note: This memo covers the OEM equipment layer only. Equipment is the largest single line in a clean drop-in, but for most NYC projects it is one input into a larger whole — the balance-of-system (enabling electrical work), complementary scopes, financing, and coordination all sit around it. Those layers, and any potential delivery model that ties them together not addressed here. Read the OEM analysis below as "the box," not "the project." -- though "the box" can be a large fraction of "the project"...
List: $3,500/HP for 1–100 units, $3,300/HP for 100+. Their volume curve is nearly flat (~$200/unit, ~6%).
Off-the-record: average customer actually pays ~$3,200 — i.e., real transaction price is ~9% below list and set by negotiation, not schedule.
Install labor ~$450/HP. With no panel/service/outlet work, equipment is the vast majority of installed cost; but once balance-of-system enters — e.g., dedicated outlets at $1,500–2,000/HP, let alone panel or building-service upgrades — equipment stays the largest single component but no longer dominates, and the project's economics shift to scopes this memo doesn't cover.
Terms: 50% on order / 50% on delivery. Negotiable levers: payment terms, lead time, warranty.
Key signal: Gradient could lower margin if a customer pays closer to COGS earlier and accepts delivery tied to actual manufacture-and-ship lead time
Feedback from one installer: Gradient is "a lot" easier to install than Midea — roughly $300/unit less install labor.
Legally Treau, Inc.; venture-backed — $18M Series A (Feb 2023) + $9M follow-on (Jun 2023) = $27.5M Series A; aggregators show ~$33–57M all-in.
No evidence of a conventional large debt facility.
Anchor customer (~$24M NYCHA Clean Heat for All); possible GE Appliances tie.
Read: capital expensive and constrained; customer payment timing plausibly matters to working capital (directionally consistent with public picture, not independently confirmed).
Innova/Ephoca — established niche OEM (schedules, as provided) #
Volume discount (in HPs), steeply front-loaded:
HP quantity
Discount
0–20
full price
21–100
24%
101–250
30%
251–500
31%
501–750
32%
751–2,000
33%
2,001+
35%
Lead-time discount: 3.7% at 9 mo, 5.5% at 12 mo (standard ~6 mo), decelerating after 9 months.
Premium product of the set (~$4,600 unit + accessories).
Equivalent window heat pump product to Gradient, sold through contractors at ~$2,400/HP flat, no discounts.
As a massive manufacturer, Midea has no working-capital problem → zero cheap-capital lever, even less than Innova. It also needs the contractor channel and is not trying to sell direct.
Harder to install than Gradient — roughly +$300/unit in labor (per installer feedback above).
Go-to-market weakness (important nuance). Anecdotally Midea's sales appear weaker than Gradient's, and Jordan at NYCHA says Midea "sucks at selling" — likely meaning selling to owners. This is probably structural, not just a Midea failing: a window heat pump primarily powers a discretionary, holistic whole-building retrofit, which is a much harder sale than a low-cost, obvious-payback swap like LED lighting or an end-of-life HVAC replacement. Contractors — Midea's channel — are good at the latter and poor at the former. The telling exception is MDG, a contractor doing huge NYCHA deals: it moves Midea volume because NYCHA's programmatic, non-discretionary demand removes the hard sell. So Midea's low price is real but comes attached to a channel that struggles to create demand for this product outside programmatic settings — a limitation that matters for how much its price advantage actually converts.
These three vendors span the full breadth of the supplier market — a huge global OEM (Midea), an established niche OEM (Innova/Ephoca), and a venture-stage startup (Gradient) — with correspondingly different cost structures, capital access, channel strategies, and pricing postures. Any procurement or selection logic has to hold across that whole range, not just one archetype.
Part 2 — The Cost-of-Capital Gap (the engine of any spread) #
Any procurement value we create arises from one arbitrage: money that funds a J-51-era retrofit is far cheaper than money an OEM uses to build and hold inventory.
Capital source
Approx. annual cost
Basis
Real estate — senior multifamily debt / refi
~mid-5% to ~6% strong borrowers; ~5–12.5% by product/credit
Profitable, Panasonic-backed; cheap capital, no acute WC problem
Gradient — venture debt (floor)
~8–15% + warrants
Market range for venture debt
Gradient — venture equity (ceiling, effective)
~30–50%+
VC return hurdle; opportunity cost of equity doing working-capital duty
Directional read: a building refinancing under J-51 accesses ~5–7% capital. Gradient's capital is dearer regardless of which source early payment displaces — venture debt (~8–15%+warrants) is the conservative floor of the gap; its equity (~30–50% effective) is the ceiling. The gap between ~6% real-estate money and Gradient's much dearer inventory financing, over the manufacture-and-ship window, is the raw material of a procurement spread.
Calibrations:
The gap accrues only over the working-capital cycle (weeks–months), so a large annual rate difference becomes low-single-digit percent of unit cost — still real on a $3,200 unit at volume.
Part of what OEMs discount for is demand certainty, not financing — a separate, often larger lever.
Critical distinction across the three OEMs. Gradient (startup) is genuinely capital-constrained, so the cheap-capital/early-payment lever is live there. Innova (established niche) is a profitable manufacturer with a strategic corporate parent (Panasonic, ~40%) throwing off ~€18.6M EBITDA — it does not have an acute working-capital problem; its lead-time discount is almost certainly pure production-planning/demand-visibility value, not a financing cry for help. Midea (global giant) has effectively no working-capital constraint at all and sells flat through contractors — no capital lever there whatsoever. So: lead with cheap early capital at Gradient; lead with forecastable, committed, patient demand at Innova; expect no capital lever at Midea. Don't over-weight the financing story anywhere but Gradient.
Part 3 — Appliance Simplicity Is Reshaping the Channel (threat and opening) #
The same simplicity that anchors our thesis is what's pushing these OEMs to sell direct to owners — and it cuts both ways.
Why it's happening. Traditional HVAC must sell through contractors/reps because it needs design, refrigerant handling, commissioning, service — structural channel dependence; the installed system is a designed-and-constructed assembly, not a product. Packaged room heat pumps engineer that away: plug-in, no line sets, no brazing, factory-tested, "ready out of the box." The simplicity that makes the product attractive is exactly what removes the OEM's need for a channel — hence the direct impulse from both Gradient and Innova.
We have already been disintermediated once — Related. Not hypothetical, and it defines our boundary exactly. Two conditions were true at once:
A true drop-in — existing electrical and existing PTAC sleeves meant essentially no balance-of-system scope for us to own; and
Related had deep in-house CM expertise from other projects plus prior diligence on Ephoca/Innova specifically, so they were comfortable with the technology and could self-manage the confidence and coordination layers.
With both the building and the owner capable, there was nothing left for a middleman to add. Related is our clearest real-world proof of where the moat ends.
Why it usually doesn't work — our opening. A direct-selling OEM can only sell its box. It can't own what actually gates most owners' "yes": the NYC electrical balance-of-system (our NEC edge, which Gradient's own proposal disclaims), the pathway/legitimacy narrative, financing and J-51 stacking, cross-technology comparison, coordination/accountability. For most owners, buying direct leaves every pre- and post-yes risk unmanaged — the OEM is a supplier, not a solution. And a direct-selling OEM (Gradient especially) is still financing its own inventory at its own high cost of capital, because scattered owners buying one building at a time provide no forecastable demand and no cheap early capital. Except where a customer looks like Related, we aren't fighting their direct channel — we're the better channel, supplying the aggregated, patient, pre-financed demand direct sales structurally can't.
The boundary is about willingness to carry risk, not sophistication. I initially framed the squeeze as "building simplicity AND owner sophistication both high." The Rudin counter-example shows that's wrong. Related and Rudin are both large, sophisticated owners — but offered the same "buy direct from the OEM" option, Related disintermediated and self-managed, while Rudin chose to pay the contractor markup specifically so warranty and accountability weren't their problem. Same capability, opposite choice. The real axis is not whether an owner can self-manage but whether they want to own post-yes accountability. So we're squeezed only where the building is a true drop-in AND the owner both can and actively wants to carry accountability themselves — the Related corner. That corner is smaller than "sophisticated owners," because many capable owners (Rudin) prefer to pay someone else to own the problem. Weaken either condition and we're back in the game: a simple drop-in owned by a board that won't self-manage is still our customer, and a complex building needs our balance-of-system edge regardless of owner type. Most of older NYC multifamily sits well outside the Related corner.
Part 4 — What Actually Dominates: Inter-OEM Price Dispersion and the Accountability Layer #
Two data points reorder the whole thesis: the procurement spread is real but small, and it is dwarfed by two larger things we are better positioned to own.
Inter-OEM price dispersion swamps the procurement spread. The cost-of-capital spread nets out to low-single-digit percent of unit cost. But the price between OEMs is far larger. On an all-in (equipment + install) basis:
The install-difficulty gap narrows what looked like an $800–1,100 equipment chasm to a ~$500/unit all-in gap — but ~$500/unit across a 500-apartment building (~1,250 HPs) is still ~$625K, an order of magnitude more than the procurement spread. The highest-leverage "procurement" move is not financing an OEM's inventory — it's neutral cross-OEM selection: matching the right unit to the building on all-in cost and cold-climate performance and installability. That is squarely our edge — we already track the post-retrofit performance data owners can't accumulate on their own.
The contractor markup is the market price of accountability — a revenue target, not a cost to strip out. Innova frames the traditional ~10% contractor markup as payment for warranty handling, coordination, and standing behind the outcome — not 10% of contractor margin. Rudin's choice proves owners will pay it even when they don't need to: a sophisticated owner that could self-manage chose to pay the markup so post-yes risk wasn't theirs. Harder-to-install equipment (Midea) makes that accountability more valuable, not less. So the markup isn't waste we help owners skip by going direct — it's a layer we are competing to provide, and one owners demonstrably pay for. The question is whether we can deliver warranty/coordination/accountability better than a contractor and capture part of that ~10%.
Net: our defensible value is (a) neutral cross-OEM selection on true all-in cost and performance, and (b) owning the accountability layer owners like Rudin pay ~10% for. The procurement spread is a marginal funding mechanism on top of these — not the product.
Volume discount is table stakes, not our edge — and correcting the units makes this stronger, not weaker. Tiers are in HPs, and at ~2.5 HPs/apartment, the 21-HP threshold is reached at ~8–9 apartments and the 101-HP (30%) tier at ~40 apartments. So essentially any real multifamily building we'd target already clears the first tier alone, and a ~40-apartment building already reaches 30% solo; a 500-apartment complex already sits at 32–33%. Aggregating across buildings to cross tiers buys almost nothing at either end — small buildings are already on the curve, large ones already near the top. Volume pricing should mostly pass through to owners; it is not where our spread lives.
Our edge is selection and accountability first, procurement spread second. The primary levers are neutral cross-OEM selection (Part 4) and owning the accountability layer owners pay ~10% for. Below those, the procurement spread comes from supplying what each OEM discounts for — cheap early capital (Gradient) and forecastable, committed, patient demand (Innova) — funded by assets an individual owner can't supply but we can: cheap J-51-era real-estate capital, and pipeline visibility. Treat the spread as a funding mechanism, not the headline.
Frame it to owners as optimization, never a tradeoff they manage. The owner should never hear "wait 12 months, save 5.5%." Buildings already face long timelines — J-51, refi, board approvals, resident communication, site and electrical verification, construction planning — and we use that unavoidable planning period to optimize procurement behind the scenes. Customer-facing message:
"Because this project is already moving through a capital-planning, financing, and incentive process, we can use that lead time to lock better equipment pricing, reduce OEM working-capital burden, and improve total project economics."
Not "delay and save." Absorbing the lead-time and payment mechanics ourselves is what makes the spread ours to earn rather than theirs to capture.
The first "buy through Momentum" is an equipment reservation, not an installed project. This is the concrete Stage-1 wedge, and it does not require becoming a GC. Rather than buying a fully installed construction project, the owner reserves a verified equipment package: price, quantity, lead time, payment terms, warranty, cancellation rights, delivery schedule. That's a bounded, non-construction, sellable product — the procurement spread realized before we ever touch construction risk.
Discipline: the quoteability gate. Equipment commitments must follow verification, never precede it. The worst outcome is locking equipment pricing before confirming electrical, outlet, panel, service, window/facade, access, and financing constraints. This is exactly where our NEC methodology and site process earn their keep — fast, credible verification is what lets us open the reservation product safely. Once a project is sufficiently verified, an equipment reservation / price-lock is one of the first scalable ways to turn market visibility into measurable owner savings and OEM value.
Magnitude is unknowable from public data — it must be discovered. No gross-margin data exists for either OEM; Innova's 35% top tier only sets a floor (margin > 35% at list). The only reliable read is a term-sheet request asking each OEM to separate and stack the three levers — volume, extended lead time, and prepayment — and quote all-in pricing beyond published tiers. That tells us whether the spread is ~2 points or double digits.
The three OEMs span the market — a global giant (Midea, ~$2,400 flat through contractors), an established niche player (Innova, premium, Panasonic-backed), and a venture-stage startup (Gradient, ~$3,200 negotiated) — with different cost structures, capital access, and channel strategies. Against that spread, our defensible value is not the procurement spread and not volume aggregation (both curves are nearly flat, and at roughly 2.5 HPs/apartment even small buildings are already on them). It is (a) neutral cross-OEM selection on true all-in cost and cold-climate performance — where inter-OEM price dispersion (roughly $500/unit all-in Midea vs. Gradient, ~$625K on a large building) dwarfs the low-single-digit procurement spread — and (b) owning the accountability layer that owners like Rudin pay the ~10% contractor markup for even when they could self-manage. The procurement spread (cheap capital at Gradient, demand certainty at Innova, nothing at Midea) is a marginal funding mechanism on top. Appliance simplicity lets OEMs attempt direct sales — Related already did it, at the intersection of a true drop-in and an owner that wanted to carry accountability itself — but that corner is smaller than "sophisticated owners," and for most owners the same move traps the OEM in the expensive-capital/lumpy-demand problem we solve. The first scalable product might be a verified equipment reservation behind a quoteability gate, not an installed project — a procurement spread with no GC risk.
Strategy Notebook
OEM Procurement Intelligence — Gradient, Innova/Ephoca, and Midea
Purpose: Summarize what we know about three packaged-heat-pump OEMs spanning the market (a startup, an established niche player, and a global giant), and what their approach to pricing tells us about risk.
Part 1 — Just the Facts #
Gradient — venture-stage startup OEM (direct conversation) #
Gradient (directional, public) #
Innova/Ephoca — established niche OEM (schedules, as provided) #
Volume discount (in HPs), steeply front-loaded:
Innova/Ephoca (directional, public) #
Midea — huge global OEM (direct conversation) #
Market breadth #
These three vendors span the full breadth of the supplier market — a huge global OEM (Midea), an established niche OEM (Innova/Ephoca), and a venture-stage startup (Gradient) — with correspondingly different cost structures, capital access, channel strategies, and pricing postures. Any procurement or selection logic has to hold across that whole range, not just one archetype.
Part 2 — The Cost-of-Capital Gap (the engine of any spread) #
Any procurement value we create arises from one arbitrage: money that funds a J-51-era retrofit is far cheaper than money an OEM uses to build and hold inventory.
Directional read: a building refinancing under J-51 accesses ~5–7% capital. Gradient's capital is dearer regardless of which source early payment displaces — venture debt (~8–15%+warrants) is the conservative floor of the gap; its equity (~30–50% effective) is the ceiling. The gap between ~6% real-estate money and Gradient's much dearer inventory financing, over the manufacture-and-ship window, is the raw material of a procurement spread.
Calibrations:
Critical distinction across the three OEMs. Gradient (startup) is genuinely capital-constrained, so the cheap-capital/early-payment lever is live there. Innova (established niche) is a profitable manufacturer with a strategic corporate parent (Panasonic, ~40%) throwing off ~€18.6M EBITDA — it does not have an acute working-capital problem; its lead-time discount is almost certainly pure production-planning/demand-visibility value, not a financing cry for help. Midea (global giant) has effectively no working-capital constraint at all and sells flat through contractors — no capital lever there whatsoever. So: lead with cheap early capital at Gradient; lead with forecastable, committed, patient demand at Innova; expect no capital lever at Midea. Don't over-weight the financing story anywhere but Gradient.
Part 3 — Appliance Simplicity Is Reshaping the Channel (threat and opening) #
The same simplicity that anchors our thesis is what's pushing these OEMs to sell direct to owners — and it cuts both ways.
Why it's happening. Traditional HVAC must sell through contractors/reps because it needs design, refrigerant handling, commissioning, service — structural channel dependence; the installed system is a designed-and-constructed assembly, not a product. Packaged room heat pumps engineer that away: plug-in, no line sets, no brazing, factory-tested, "ready out of the box." The simplicity that makes the product attractive is exactly what removes the OEM's need for a channel — hence the direct impulse from both Gradient and Innova.
We have already been disintermediated once — Related. Not hypothetical, and it defines our boundary exactly. Two conditions were true at once:
With both the building and the owner capable, there was nothing left for a middleman to add. Related is our clearest real-world proof of where the moat ends.
Why it usually doesn't work — our opening. A direct-selling OEM can only sell its box. It can't own what actually gates most owners' "yes": the NYC electrical balance-of-system (our NEC edge, which Gradient's own proposal disclaims), the pathway/legitimacy narrative, financing and J-51 stacking, cross-technology comparison, coordination/accountability. For most owners, buying direct leaves every pre- and post-yes risk unmanaged — the OEM is a supplier, not a solution. And a direct-selling OEM (Gradient especially) is still financing its own inventory at its own high cost of capital, because scattered owners buying one building at a time provide no forecastable demand and no cheap early capital. Except where a customer looks like Related, we aren't fighting their direct channel — we're the better channel, supplying the aggregated, patient, pre-financed demand direct sales structurally can't.
The boundary is about willingness to carry risk, not sophistication. I initially framed the squeeze as "building simplicity AND owner sophistication both high." The Rudin counter-example shows that's wrong. Related and Rudin are both large, sophisticated owners — but offered the same "buy direct from the OEM" option, Related disintermediated and self-managed, while Rudin chose to pay the contractor markup specifically so warranty and accountability weren't their problem. Same capability, opposite choice. The real axis is not whether an owner can self-manage but whether they want to own post-yes accountability. So we're squeezed only where the building is a true drop-in AND the owner both can and actively wants to carry accountability themselves — the Related corner. That corner is smaller than "sophisticated owners," because many capable owners (Rudin) prefer to pay someone else to own the problem. Weaken either condition and we're back in the game: a simple drop-in owned by a board that won't self-manage is still our customer, and a complex building needs our balance-of-system edge regardless of owner type. Most of older NYC multifamily sits well outside the Related corner.
Part 4 — What Actually Dominates: Inter-OEM Price Dispersion and the Accountability Layer #
Two data points reorder the whole thesis: the procurement spread is real but small, and it is dwarfed by two larger things we are better positioned to own.
Inter-OEM price dispersion swamps the procurement spread. The cost-of-capital spread nets out to low-single-digit percent of unit cost. But the price between OEMs is far larger. On an all-in (equipment + install) basis:
The install-difficulty gap narrows what looked like an $800–1,100 equipment chasm to a ~$500/unit all-in gap — but ~$500/unit across a 500-apartment building (~1,250 HPs) is still ~$625K, an order of magnitude more than the procurement spread. The highest-leverage "procurement" move is not financing an OEM's inventory — it's neutral cross-OEM selection: matching the right unit to the building on all-in cost and cold-climate performance and installability. That is squarely our edge — we already track the post-retrofit performance data owners can't accumulate on their own.
The contractor markup is the market price of accountability — a revenue target, not a cost to strip out. Innova frames the traditional ~10% contractor markup as payment for warranty handling, coordination, and standing behind the outcome — not 10% of contractor margin. Rudin's choice proves owners will pay it even when they don't need to: a sophisticated owner that could self-manage chose to pay the markup so post-yes risk wasn't theirs. Harder-to-install equipment (Midea) makes that accountability more valuable, not less. So the markup isn't waste we help owners skip by going direct — it's a layer we are competing to provide, and one owners demonstrably pay for. The question is whether we can deliver warranty/coordination/accountability better than a contractor and capture part of that ~10%.
Net: our defensible value is (a) neutral cross-OEM selection on true all-in cost and performance, and (b) owning the accountability layer owners like Rudin pay ~10% for. The procurement spread is a marginal funding mechanism on top of these — not the product.
Part 5 — Implications for Our Model #
Volume discount is table stakes, not our edge — and correcting the units makes this stronger, not weaker. Tiers are in HPs, and at ~2.5 HPs/apartment, the 21-HP threshold is reached at ~8–9 apartments and the 101-HP (30%) tier at ~40 apartments. So essentially any real multifamily building we'd target already clears the first tier alone, and a ~40-apartment building already reaches 30% solo; a 500-apartment complex already sits at 32–33%. Aggregating across buildings to cross tiers buys almost nothing at either end — small buildings are already on the curve, large ones already near the top. Volume pricing should mostly pass through to owners; it is not where our spread lives.
Our edge is selection and accountability first, procurement spread second. The primary levers are neutral cross-OEM selection (Part 4) and owning the accountability layer owners pay ~10% for. Below those, the procurement spread comes from supplying what each OEM discounts for — cheap early capital (Gradient) and forecastable, committed, patient demand (Innova) — funded by assets an individual owner can't supply but we can: cheap J-51-era real-estate capital, and pipeline visibility. Treat the spread as a funding mechanism, not the headline.
Frame it to owners as optimization, never a tradeoff they manage. The owner should never hear "wait 12 months, save 5.5%." Buildings already face long timelines — J-51, refi, board approvals, resident communication, site and electrical verification, construction planning — and we use that unavoidable planning period to optimize procurement behind the scenes. Customer-facing message:
Not "delay and save." Absorbing the lead-time and payment mechanics ourselves is what makes the spread ours to earn rather than theirs to capture.
The first "buy through Momentum" is an equipment reservation, not an installed project. This is the concrete Stage-1 wedge, and it does not require becoming a GC. Rather than buying a fully installed construction project, the owner reserves a verified equipment package: price, quantity, lead time, payment terms, warranty, cancellation rights, delivery schedule. That's a bounded, non-construction, sellable product — the procurement spread realized before we ever touch construction risk.
Discipline: the quoteability gate. Equipment commitments must follow verification, never precede it. The worst outcome is locking equipment pricing before confirming electrical, outlet, panel, service, window/facade, access, and financing constraints. This is exactly where our NEC methodology and site process earn their keep — fast, credible verification is what lets us open the reservation product safely. Once a project is sufficiently verified, an equipment reservation / price-lock is one of the first scalable ways to turn market visibility into measurable owner savings and OEM value.
Magnitude is unknowable from public data — it must be discovered. No gross-margin data exists for either OEM; Innova's 35% top tier only sets a floor (margin > 35% at list). The only reliable read is a term-sheet request asking each OEM to separate and stack the three levers — volume, extended lead time, and prepayment — and quote all-in pricing beyond published tiers. That tells us whether the spread is ~2 points or double digits.
Bottom Line #
The three OEMs span the market — a global giant (Midea, ~$2,400 flat through contractors), an established niche player (Innova, premium, Panasonic-backed), and a venture-stage startup (Gradient, ~$3,200 negotiated) — with different cost structures, capital access, and channel strategies. Against that spread, our defensible value is not the procurement spread and not volume aggregation (both curves are nearly flat, and at roughly 2.5 HPs/apartment even small buildings are already on them). It is (a) neutral cross-OEM selection on true all-in cost and cold-climate performance — where inter-OEM price dispersion (roughly $500/unit all-in Midea vs. Gradient, ~$625K on a large building) dwarfs the low-single-digit procurement spread — and (b) owning the accountability layer that owners like Rudin pay the ~10% contractor markup for even when they could self-manage. The procurement spread (cheap capital at Gradient, demand certainty at Innova, nothing at Midea) is a marginal funding mechanism on top. Appliance simplicity lets OEMs attempt direct sales — Related already did it, at the intersection of a true drop-in and an owner that wanted to carry accountability itself — but that corner is smaller than "sophisticated owners," and for most owners the same move traps the OEM in the expensive-capital/lumpy-demand problem we solve. The first scalable product might be a verified equipment reservation behind a quoteability gate, not an installed project — a procurement spread with no GC risk.