Selling Outcomes, Not Machines: Fleet Monitoring According to OEMs

On November 3, 2026, IIoT World published a summary of a panel at the Industrial AI Summit 2026 on the *outcome as a service* model in manufacturing. The speakers were Boris Scharinger (Siemens), Jan Pingel (Ingersoll Rand) and Karthikeyan Natarajan, CEO of Infinite Uptime, moderated by Rob Tiffany of IDC. One caveat is necessary: the article is marked "Sponsored by Infinite Uptime" (with the label "Editorially Independent"). What follows are the speakers' claims, not independently verified, and no data beyond their statements was provided.

Three stages: from selling a product to selling a result

The panel describes an evolution in three steps:

• Product sale: the customer owns the machine and carries the risk.
• Availability guarantee: the supplier manages the risk of downtime.
• Output as a service: the supplier owns the asset and all operating risk; the customer pays only for what it receives.

Two concrete examples

Siemens Energy sells small and medium-sized gas turbines to process industries. The customer grants a 10-day maintenance window every two years: all replacements must fit within that window, with no unplanned outages in between. The financial exposure was significant enough that Siemens' internal audit was asked to verify the prediction algorithms. According to the article, this is an early signal: internal audit will need to be able to examine predictive models, because a forecasting error becomes a contractual liability.

Ingersoll Rand, by contrast, offers "air over the fence": it installs compressors on adjacent land and pipes compressed air into the plant. The OEM owns the machines and pays for the energy; the customer pays for the quantity delivered. For each type of equipment (compressors, blowers, vacuum pumps), 50–100 insights are generated, amounting to thousands of predictions per week. They cover the probability of failure (over hours, days, weeks or months), service needs such as filter changes or clogged components, and capacity signals: if all machines run at 95%, the customer needs more units.

The real obstacle is the business model

Take an illustrative example: a $100,000 machine with roughly twice that amount in service revenue over 10 years. An outcome-based model cuts first-year revenue to about a third of the traditional figure. Startups can be born this way; established OEMs face a multi-year revenue decline. Many rely on financial companies to structure the agreements, and these, according to the panel, capture the largest share of the profit. In Europe, shifting capex to opex does not persuade customers: what matters is the transfer of risk to the supplier.

The data advantage

Pricing an outcome requires learning from hundreds of thousands of similar machines under varied conditions; a single customer's data is not enough. Customers often prohibit AI vendors from using production data to train models for others: one speaker cited 20 projects in three years without a single reusable product. OEMs with their own installed base do not need that permission. There is also a distinction in what to share: raw high-frequency vibration data lets the customer build its own diagnostics, while sharing the interpretation (for example, "bearing problem, 10–14 days before failure") preserves the supplier's advantage.

What to do with it in practice

For those selling or buying outcomes, the points to verify are:

• Monitoring and pipelines: reliable condition monitoring and robust prediction pipelines are the foundation of the contract.
• Auditability: models used in contract clauses must be documented, validated and verifiable.
• Fleet data architecture: data must be aggregated and normalized across many installations.
• Data clauses: ownership and sharing determine who can learn from the data, and therefore who can set a sustainable price.

Given the sponsored nature of the source and the basis of a single panel, these cases are best read as indications of direction rather than quantitative evidence.