Hardware-as-a-Service

What exactly is Hardware-as-a-Service?#

Hardware-as-a-Service is a business model in which a provider supplies continuing access to a physical device as part of an integrated recurring service. The offer usually includes embedded firmware, cloud software, connectivity, monitoring, maintenance, support, replacement, upgrades, or logistics. Pricing may be per active device, per site, per period, per unit of use, per unit of output, or a hybrid.

The useful boundary is not the label on the invoice. A HaaS provider normally retains three forms of responsibility:

1. Economic responsibility: the provider funds, or defers recovery of, hardware and deployment cost.
2. Operational responsibility: the provider must keep the device and supporting service usable, secure, and supported.
3. Lifecycle responsibility: returns, replacement, refresh, recovery, refurbishment, and end-of-life all affect the model.

Servitization is the wider concept. The 1988 article that named the shift described manufacturers adding value by adding services, then moving from selling the asset toward selling access, availability, use, or an outcome. Connected devices make that shift easier to meter and operate — and add cloud, privacy, data-access, and cybersecurity obligations that did not exist in 1988.

HaaS versus adjacent models#

These overlap. An agreement marketed as HaaS may be a lease for accounting purposes: under IFRS 16 the test is whether the customer controls the use of an identified asset for a period in exchange for consideration, and substantive supplier substitution rights can change that conclusion. The commercial name does not decide the accounting, and revenue recognition, sales tax, property tax, insurance, and balance-sheet treatment all depend on facts and jurisdiction.

Why does HaaS matter to founders?#

It removes the buyer's upfront adoption barrier#

A customer may value the device but resist a capital purchase, an uncertain payback, installation work, maintenance staffing, or obsolescence risk. A recurring offer converts a large purchase decision into an operating commitment and bundles deployment into one accountable service.

That can shorten the path to a pilot and widen the addressable market. It can also lengthen security, procurement, and accounting review, because the provider stays embedded in the customer's operations. A low upfront price moves the capital burden to the provider; it does not make the burden disappear.

It can create a continuing value and data loop#

Connected devices produce operating data that improves maintenance, workflow, safety, or product performance. The provider observes failures, ships software updates, refines service intervals, and expands applications without replacing the asset. This supports retention only when the data produces continuing customer value.

Samsara is the clearest public example of an integrated device-service model: a platform subscription bundles IoT data collection from a gateway, camera, or sensor with cellular connectivity, cloud applications and APIs, support, and warranty coverage, priced generally per asset and per application on multi-year initial terms. In fiscal 2026 it reported $1.62 billion of revenue and ended the year with $1.89 billion of ARR.

The lesson is not that every HaaS company should copy those terms. It is that a repeatable unit, an integrated product, continuing value, and contract duration have to work together. A device that unlocks a one-time job does not justify a recurring model.

Servitization can align price with uptime or output#

At the advanced end, a provider stops charging for possession of equipment and charges for availability or productive use. Rolls-Royce's TotalCare transfers engine time-on-wing and maintenance-cost risk to Rolls-Royce and charges dollars per engine flying hour. The customer gets predictable operating cost; the provider earns when the engine performs.

That alignment is powerful because the provider is often best placed to improve reliability. It also transfers variance. If the provider promises an outcome affected by customer misuse, site conditions, input quality, or third parties it cannot control, the contract becomes an uncapped insurance policy.

Recurring revenue can hide a financing business#

Hardware is paid for before or soon after deployment; service revenue arrives over years. Fast bookings growth therefore consumes cash. Inventory, non-cancelable component orders, freight, installation, activation delay, receivables, replacements, and device recovery all sit between a signed contract and cash payback.

Samsara's balance sheet makes the burden visible: at 31 January 2026 it carried $48.2 million of inventories and $440.1 million of current and non-current connected-device costs — roughly 27% of a full year's revenue tied up in devices already in the field. An early-stage founder should forecast two businesses at once: the customer economics of recurring value, retention, and expansion, and the asset economics of procurement, deployment, failure, recovery, and financing.

Lifecycle obligations continue long after the sale#

Connected hardware is not finished when it ships. NIST's IR 8259 Revision 1, published 20 April 2026, says manufacturers should identify customer cybersecurity needs and plan cybersecurity capabilities, maintenance, support, and end-of-life communication before products are sold. Market rules are moving the same way: the EU Data Act has applied since 12 September 2025 and gives users of connected products access to raw and pre-processed data generated through their use, including for rented or leased products. The EU Cyber Resilience Act's vulnerability and severe-incident reporting duties begin 11 September 2026, with its main obligations applying from 11 December 2027.

Support period, patchability, data portability, and end-of-life communication are product-cost inputs, not fine print added after launch.

The device-service flywheel: six decisions#

A weak link invalidates the whole model.

1. Outcome — what recurring job is the customer buying? Start with the operating job, not the device. "A sensor for $165 per month" is weak; "continuous temperature evidence and alerts for each cold-storage room" is a job; "verified temperature compliance during covered operating hours" is stronger still, but only when the provider controls the drivers. State the next-best alternative — buy equipment, rent it, hire people, run a manual process, or accept the loss — and quantify the avoided downtime, labor, waste, risk, or capital.

2. Unit — what activates, bills, and deactivates? Define one auditable commercial unit that works in five systems: the contract identifies it, operations can provision it, telemetry can measure it, billing can reconcile it, and the customer can audit it. Avoid charging per shipped device when value starts only at activation. Separate ordered, shipped, installed, connected, active, billable, suspended, returned, refurbishing, redeployed, lost, and retired states.

3. Promise — which obligations are bundled? Write it as a service catalog: device specification and approved environment; installation and customer prerequisites; connectivity, cloud, data retention, and APIs; support hours and response targets; maintenance, repair, and replacement; uptime measurement and exclusions; firmware and security-update policy; refresh eligibility; title, loss, damage, and insurance; and return, data deletion, and end-of-life process. Product architecture, contract promises, and accounting cannot be designed independently — Samsara, for instance, treats its platform and device access as one combined performance obligation recognized over the subscription period.

4. Risk — who controls each source of variance? Build a risk register before pricing.

Keep catastrophic and unbounded risk out of a flat fee. A service-credit cap is not a substitute for product safety, but it stops a routine SLA miss from becoming unlimited commercial liability.

5. Cash — does each activation earn back all cash before risk catches up? Build monthly cash flow by activation cohort, not one blended annual margin:

Unit NPV = U - D - CAC + Σ[S_t × (P_t - V_t) - F_t] / (1 + r_m)^t + S_T × R_T / (1 + r_m)^T

where U is upfront cash collected, D is landed device plus provisioning, freight, and installation cash, CAC is acquisition cash for the activation, S_t is the probability the unit is still active and paying in month t, P_t is recurring and usage revenue, V_t is direct cost while active, F_t is cost that does not scale cleanly with activity (repairs, truck rolls, loss, credits, storage), R_T is recoverable value net of reverse logistics and refurbishment, and r_m is the monthly risk-adjusted discount rate.

Run it over distributions, not averages. Device failure, churn, and field-service cost are skewed: a small tail of bad installations can consume the apparent margin from the median device. Track three views and never silently substitute one for another — cash contribution (revenue less direct cash cost while active), economic contribution (cash contribution less device consumption and expected lifecycle cost), and accounting gross margin (revenue and cost under the applicable policy).

6. Loop — can operations improve reliability, recovery, and value? Servitization is most defensible when lifecycle responsibility creates learning. Rolls-Royce uses engine-health monitoring and fleet-wide experience to plan maintenance while a fixed fee per flying hour rewards engine performance. For a startup the loop is simpler: fewer truck rolls, longer battery life, faster remote diagnosis, a common replacement part, higher first-time install success, reliable redeployment. If field data cannot change cost or value, the provider has taken lifecycle risk without gaining a learning advantage.

Pricing and packaging choices#

Use value-based pricing to find the customer's economic ceiling, then use device-service NPV to find the provider's floor. A competitive monthly price below that floor is evidence to redesign hardware, scope, term, financing, or segment — not evidence that lifecycle costs can be ignored.

Worked example: 100 monitoring devices#

An industrial-monitoring device with cloud analytics and replacement coverage, on a 36-month minimum term. All figures are hypothetical.

Per activated device — upfront cash. Activation fee collected $250; device manufacturing cost $900; freight and provisioning $70; standard installation $130; allocated CAC $300.

Net deployment cash outlay = $900 + $70 + $130 + $300 - $250 = $1,150

Monthly economics while active. Subscription revenue $165; connectivity and cloud $15; routine support and field-service reserve $12; warranty and replacement reserve $8; billing, insurance, and fleet admin $5.

Direct operating cash cost = $40 Monthly cash contribution = $165 - $40 = $125 Cash contribution margin before device recovery = $125 / $165 = 75.8%

Cash payback.

$1,150 / $125 = 9.2 months

For 100 activations the provider needs $115,000 of net deployment cash before shared headcount, safety stock, receivable timing, taxes, or manufacturing deposits.

Assume the unit stays active all 36 months, a 12% annual discount rate expressed as an equivalent monthly rate of about 0.949%, and $100 of net residual value after recovery and refurbishment:

Undiscounted unit cash = -$1,150 + (36 × $125) + $100 = $3,450 Unit NPV ≈ $2,718 100-device cohort NPV ≈ $271,800

Two caveats before that number goes in a deck. The cohort figure assumes 100 simultaneous activations with no survival decay — real cohorts churn, and S_t belongs in the model. And the 75.8% cash contribution margin is not an accounting gross margin; it treats device consumption as free. Spreading the $1,100 landed-and-installed cost over the 36-month contract gives monthly device cost of $1,100 / 36 = $30.56 and an economic contribution margin of:

($165 - $40 - $30.56) / $165 = 57.2%

Downside test. Now let competitive pressure cut the monthly price to $135, direct operating cash cost rise to $60, the customer exit after 18 months, and recovery produce a $100 net exit cost rather than value. Monthly cash contribution falls to $75, and at the same discount rate:

Downside unit NPV ≈ $1

The entire risk-adjusted return has disappeared. One extra truck roll, payment delay, lost unit, or failed recovery makes the cohort negative. The conclusion is not "raise price" by default — there are six levers: collect more deployment cash upfront, lengthen the minimum term or add a termination buyout, lower device and installation cost, lower recurring service and failure cost, improve recovery and redeployment yield, or move to a segment with greater measurable value.

Related concepts#

• Subscription Model: recurring access, term design, retention, and billing cadence.
• Pricing Metric and Value Metric: decide whether price follows devices, sites, usage, uptime, or output.
• Hybrid Pricing: combine a base device-service fee with measured usage or overage.
• Usage-Based Pricing: design auditable consumption meters and manage variable revenue.
• Outcome and Performance-Based Pricing: structure attribution and risk sharing before promising results.
• Managed Services: distinguish device lifecycle responsibility from operating a customer function.
• Everything-as-a-Service (XaaS): place HaaS within the wider as-a-service family.
• API-as-a-Product: device APIs, integrations, metering, and external developer contracts.
• Value-Based Pricing: quantify avoided cost rather than pricing from device cost alone.
• Bootstrapping vs. Venture Capital: match asset intensity and growth cash needs to financing strategy.

Sources#

1. Samsara Inc., Fiscal 2026 Form 10-K
2. Samsara Inc., Fourth Quarter and Full Fiscal Year 2026 Financial Results (March 5, 2026)
3. Rolls-Royce, TotalCare
4. Rolls-Royce, TotalCare as a circular business model
5. NIST IR 8259 Rev. 1, Foundational Cybersecurity Activities for IoT Product Manufacturers (April 2026)
6. NIST, IoT Device Cybersecurity Capability Core Baseline and technical catalog
7. Regulation (EU) 2023/2854 (Data Act) and European Commission, Data Act explained
8. Regulation (EU) 2024/2847 (Cyber Resilience Act) and European Commission, CRA summary
9. IFRS Interpretations Committee, Definition of a Lease — Substitution Rights (IFRS 16)
10. Vandermerwe, S. and Rada, J., "Servitization of Business: Adding Value by Adding Services," European Management Journal 6(4), 1988

---

This page is an educational and operating explanation, not accounting, legal, or tax advice. Lease classification, revenue recognition, product-safety duties, and data and cybersecurity obligations vary by jurisdiction and by the facts of each contract. Have qualified advisers review material agreements before you standardize them.