q3-2026.pdf meta_title: “Your Plan’s $21,000 Power Gap” meta_title_source: “derived_from_title” meta_description: “Manhattan Associates’ cloud fees passed its services revenue last quarter, by $21,000. On a $306.2 million book, that gap is a rounding error, and it is also…” meta_description_source: “derived_from_lead”

Manhattan Associates’ cloud fees passed its services revenue last quarter, by $21,000. On a $306.2 million book, that gap is a rounding error, and it is also a first. A week ago, Gartner said the thing that now decides whether a supply chain can grow is whether the site can get power. Put the two together and your plan stops at the substation: the money is moving off the work that fits a model to one network, just as the network’s binding limit became a local grid connection.

At a glance

  • The shift: Software revenue crossed from setup work to cloud access as Gartner named energy the binding limit.
  • The number: $21,000 — cloud fees’ margin over services at one vendor last quarter.
  • The risk: Setup work gets deferred first, so the plan loses its site-specific model.

The Optional Line Goes First

The order of events is the argument. Manhattan’s annual filing for 2025, out in February, showed its services line shrinking, and put the fall down to buyers who pushed services work into later periods. Nine months on, the September quarter showed where the deferral ends: cloud fees grew 28.2% to $134.4 million, while services grew 1.1% [2]. Nine days after those results, on 5 November, Gartner published research saying that energy supply now limits supply chain growth, and that grid capacity shortfalls and delays in getting a connection can hold a new facility shut for years.

The two lines are not swaps in the vendor’s model. They are swaps in the buyer’s budget, and that is what makes the crossover a planning fact rather than a finance one. The same filing spells out what the services line buys: implementation, training and support, including “system planning, design, configuration, testing, and other software implementation support”. It then adds that those services “are typically optional and distinct from our software”.

That work is where a standard forecast model learns the local facts — which plant, which lane, which supplier lead time, which site cannot draw the power it needs. A cloud fee runs the model. It does not know the site.

I read the filing twice before I accepted what it was saying. Gartner’s limits are physical and local: a grid connection, a shortfall, a queue. The work that would put those limits into a plan is billed by the hour and scoped per project, and the vendor calls it optional. So the buyer keeps the engine and drops the fitting, and that choice is not foolish. It is the one line in the stack a budget committee can move.

Gartner puts the same point in the language of the plan itself. These limits, its analysts write, are “increasingly shaping network design decisions, capacity planning and facility expansion strategies” — the three inputs a yearly plan takes as given. A planning cycle can absorb a bad forecast, because next month’s review corrects it. It cannot absorb a plant that will not open, because the plan holds no second site and the queue does not move on a monthly clock.

By the numbers

  • $22.5 million: What Manhattan’s services line lost in 2025 as buyers deferred setup work [3].
  • $503 million: What that line still billed in 2025, nearly half of company revenue [3].
  • 20% to 30% higher costs: Gartner’s 2029 forecast if a maker lacks strong energy management [1].

What the Next Filings Will Show

The test is cheap to run, which is why I expect it to be run badly. Most listed supply chain software vendors report services as their own line, so two more quarters will show whether last year’s deferral became a permanent mix or a one-year pause. A services line that stays flat while cloud fees compound would mean the industry stopped building the staff who fit a model to a site, just as the site’s limits turned physical.

On the buyer’s side, the tell sits in what a renewal now covers. The cloud fee is where the commitment lives, and the planning contract now outlives the planning team that knew how the model was fitted. Some planning groups now ask for a list of hard limits as a deliverable: which sites have firm power, which sit behind a queue, which face a price nobody can fix. That list is what turns a forecast into a plan someone can run. Whether the vendor builds it or the buyer’s own engineers do is a choice with a cost on both sides.

The Catch

My reading, held loosely: the crossover may say less about planning than it looks. A shrinking services line can also mean the work got easier — cloud delivery takes out plenty of the on-site setup that implementation once billed, so a vendor can sell less of it and thin nothing. Manhattan’s filing blames buyer budgets, not efficiency, but a budget pause and a simpler product can look the same for a year. Gartner’s 20% to 30% is a forecast for 2029, not a measure, and one vendor’s mix is not the market. What survives all three caveats is narrower, and still awkward: the setup work is optional, and the limit is not.

Related reading

Sources