Honeywell's $60M Refinery AI Claim Is Gross, Not Net

A number with its conditions attached

Most vendor benefit claims fail in the arithmetic. This one does not, and that is worth saying before the criticism. Honeywell describes the estimate as specific to Dangote's refinery at full-capacity throughput. The same release identifies the facility's current nameplate capacity as 650,000 barrels per day and separately discusses a planned expansion to 1.4 million. The arithmetic below uses that 650,000-barrel figure as an illustrative current-capacity basis.

At 650,000 barrels a day, 29 cents per barrel produces approximately $68.8 million over 365 days, or $62.2 million over 330 days. Both exceed Honeywell's stated $60 million threshold. The published figures are therefore arithmetically compatible, but the release does not disclose the throughput assumption or underlying calculation used to produce the estimate.

The problem is not the multiplication. It is what the product of that multiplication represents. Honeywell's first footnote says the benefit is calculated gross of implementation costs. Its second footnote repeats the point in full: gross operational benefit excludes the cost of implementing Forge Plant Cognition. A refinery reading this release learns what the platform might add before it learns what the platform costs, because the second number is not in the document at all. There is no licence price, no services figure and no contract value, for Dangote or for anyone else. Those omissions prevent readers from calculating a net return or independently assessing the projected benefit.

Preliminary, specific, and legally forward-looking

Three more words in the release do quiet work. The estimate rests on "preliminary data", which Honeywell does not publish. The launch materials contain no independent measurement or validation of the projected benefit from Dangote. The quoted Dangote executive discusses the platform selection, not a realized $60 million return. The estimate applies to "this specific facility", so it cannot be carried across to another refinery with a different configuration, crude slate or baseline. And in the release's own legal section, Honeywell lists the estimated gross benefit among its forward-looking statements, alongside Dangote's throughput and workforce goals, and states that such statements are not guarantees of future performance.

None of this makes the launch empty. Refineries are process plants where small yield and downtime improvements are worth real money, and Honeywell UOP has designed and licensed refinery processes for decades. A platform that combines live operating data with physics-based models and that vendor's process library is a plausible place to find 29 cents a barrel. Plausible is the correct word. The release describes a system designed to identify constraints, anticipate problems and recommend actions. It does not report a measured result from a plant already running it, because Dangote is the first site and the work starts now.

Keep the expansion out of this number

The same release carries a second, larger story, and the two are easy to merge by mistake. The release also discusses a planned expansion to 1.4 million barrels per day. That larger capacity must not be used to scale the $60 million estimate: Honeywell says the preliminary benefit is specific to the evaluated facility, and provides no corresponding estimate for the expanded operation.

What would settle the claim is straightforward, and none of it is in today's material: a published baseline, the implementation cost, and a measured per-barrel figure after a defined operating period, reported by Dangote rather than by the vendor. Until at least the first of those appears, the honest version of the headline is the one Honeywell itself wrote in small type: more than $60 million a year, gross, preliminary, for one refinery, not guaranteed. As Honeywell's own published estimate, it is a number the company can later be measured against. As a saving, it does not exist yet.

Sources

[1] Honeywell Technologies, "Honeywell Technologies Launches Forge Plant Cognition with Dangote, Built with NVIDIA AI and Dell Technologies Capabilities", press release, October 26, 2026 — the estimate, both footnotes, the forward-looking statements section, the platform components and the quotationsRead source
[2] Business Wire, identical release, October 26, 2026Read source
[3] Reuters coverage via Empire Magazine Africa, "Honeywell Deploys New Industrial AI Platform At Dangote Refinery With $60 Million Annual Benefit In View", October 26, 2026 — secondary coverage repeating the preliminary, gross and not guaranteed conditionsRead source

THE ALFW VERDICT

Honeywell's $60 million figure is not arithmetically implausible, but arithmetic is not evidence that the benefit has been achieved. The number is a preliminary estimate of annual gross operational benefit for a specific refinery, published by the company selling the platform. Implementation costs are excluded, the underlying baseline is unpublished and the outcome is not guaranteed. That does not make the estimate meaningless; it makes its evidentiary status clear. Operators evaluating the product need a measured per-barrel improvement over a defined period, implementation costs and a facility-specific net return before treating the projected benefit as realized savings. Quote the $60 million as Honeywell's estimate, not as money Dangote has already earned or is guaranteed to receive.