What Is GMP in Construction?
In construction, GMP stands for Guaranteed Maximum Price— a contract in which the contractor agrees to complete the project for no more than an agreed ceiling price, billing the owner for the actual cost of the work plus a fee while absorbing any cost above that ceiling.1 (In pharmaceuticals and manufacturing, "GMP" means Good Manufacturing Practice— an unrelated term. This article is about the construction contract.)
The American Institute of Architects puts the mechanism plainly: a GMP contract's basis of payment is "the Cost of the Work Plus a Fee with a Guaranteed Maximum Price."1 The contractor is reimbursed for actual costs plus a fee, but the total is capped— and if costs run past the cap, the contractor eats the difference, which shrinks the profit on the job.3
A few things worth keeping straight:
- GMP in one line: a price ceiling the contractor can't bill past without absorbing the overage.2
- What GMP stands for: Guaranteed Maximum Price.1
- Also called: a not-to-exceed price (NTE or NTX)— though a true GMP usually adds open-book accounting and a shared-savings clause that a bare not-to-exceed price may not.7 It also travels under "construction manager at risk contract."2
- Not to be confused with: Good Manufacturing Practice, the manufacturing and pharma "GMP."
You see GMPs on commercial, institutional, and public work— often paired with construction-manager-at-risk delivery or progressive design-build, which we'll get to. Knowing the term is one thing. Knowing how the contract actually behaves is what keeps you out of trouble.
How a GMP Contract Works
A GMP contract runs open-book. The contractor shares line-item budgets and expenses so the owner can see the actual cost of labor, materials, equipment, and subcontractors.2 The owner pays those actual costs plus a fee— up to the guaranteed maximum. Go over, and the overrun comes out of the contractor's profit;3 come in under, and the savings are usually split.2
What's inside a GMP estimate:
- Cost of the work— labor, materials, equipment, subcontractor packages.2
- General conditions— the cost of running the site: supervision, temporary facilities, cleanup.
- The contractor's fee or markup— negotiated, and visible to the owner under the open-book requirement.
- Contingency— a reserve the contractor builds in to absorb the uncertainty of pricing incomplete drawings and minor unforeseen conditions.2 Who can spend it, and on what, is set by negotiated rules in the contract. There's no standard percentage; it varies by project, scope maturity, and risk.
- Allowances— placeholder amounts for scope that isn't fully designed yet (the lobby flooring, say), trued up later.
Two more moving parts matter. Change orders: scope changes go through a formal change-order process— and that process is also one of the main ways the "guarantee" grows over the life of a job. The shared-savings clause: if the project finishes below the guaranteed maximum price, the difference is divided between owner and contractor under a negotiated formula. A 50/50 split is common, but the ratio is negotiable;2 absent a shared-savings clause, the underrun defaults to the owner.
So the owner gets cost visibility and a ceiling. The contractor gets reimbursed for real costs plus a fee— and carries the risk that the real costs blow through the cap.3 All of that hinges on one question most explainers rush past: when, exactly, does the number get set?
When the GMP Gets Set— and Why That's the Whole Ballgame
The GMP is set before the design is finished. In private work it's commonly fixed when the construction documents are roughly 60 to 80 percent complete;3 on U.S. federal projects, the final estimated cost of the work— and therefore the GMP— generally can't be established earlier than 75 percent of construction documents, and should be set before 100 percent.4 That timing is the point: the contractor is guaranteeing a price based on drawings that aren't done.
Private projects: a GMP is commonly set at roughly 60–80% construction documents. Federal projects: generally no earlier than 75% construction documents, and before 100% (48 CFR § 536.7105-2).4
Setting the number early is a feature. It lets construction start well before lump-sum-grade design is finished— lump-sum bidding typically needs the drawings essentially complete first.3 Crews mobilize, long-lead items get ordered, the schedule compresses.
It's also the risk. When you price incomplete drawings, every scope gap and design assumption is the contractor's exposure— which is exactly what the contingency exists to absorb.5 And the "guarantee" in a guaranteed maximum price has asterisks. It erodes through allowances, qualifications and exclusions, owner-caused change orders, and a contingency that can be padded.5 Late design changes are a leading way a GMP grows.
None of that makes a GMP a bad instrument. It makes it an honest one— a number that's only as solid as the design thinking standing behind it when it gets locked. Before we follow that thread, it helps to see where a GMP sits among the other ways to price a job.
GMP vs. Lump Sum vs. Cost-Plus vs. T&M
Here's the short version. A lump-sum contract is one fixed price the owner pays no matter what— and gets no savings if the job comes in low.8 A cost-plus contract reimburses actual cost plus a fee with no ceiling.7 Time-and-materials bills hourly labor rates plus materials, usually for small or undefined-scope work. And a GMP is cost-plus with a ceiling, open books, and shared savings— more transparent than a lump sum, more cost-certain than open-ended cost-plus.8
| Contract type | Who sets the price & when | Who carries cost-overrun risk | Cost transparency | Savings if under budget | Best fit |
|---|---|---|---|---|---|
| Lump sum / stipulated sum | Owner accepts one fixed price, bid on ~95–100% complete design | Contractor, within the fixed price | Low— costs stay private | None— owner keeps the fixed price, no share8 | Fully designed scope, simple risk allocation |
| Cost-plus (no cap) | Open-ended; owner pays actual cost + fee | Owner— no ceiling | High— actual costs billed | N/A— no target to beat | Fast-track work where scope is genuinely unknown |
| Time & materials | Hourly rates + materials, no fixed total | Owner— hours are what they are | Medium— rates and quantities visible | N/A | Small jobs, repairs, undefined scope |
| GMP (Guaranteed Maximum Price) | Contractor commits a ceiling, often on ~60–80% design | Contractor— absorbs anything above the GMP3 | High— open-book accounting2 | Underrun usually shared (50/50 common, negotiable)2 | Larger projects where work starts before design is fully done |
How to read it: a lump sum gives you the cleanest single number, but only after full design and with no upside if it comes in low. Cost-plus gives maximum flexibility and minimum cost certainty. A GMP is the middle path— and the quality of that middle path depends on how complete the design is when the number gets set. Two of these— GMP especially— usually arrive packaged with a particular delivery method and a particular contract form.
GMP, CM-at-Risk, and the Preconstruction Phase (AIA A102 & A133)
A GMP almost always shows up inside a construction-manager-at-risk (CMc / CMAR) or progressive design-build delivery, governed in U.S. practice by AIA Document A102 (Owner/Contractor) or AIA Document A133 (Owner/Construction Manager as Constructor).16 A133 splits the construction manager's work into a preconstruction phase and a construction phase; once the owner executes the Guaranteed Maximum Price Amendment (Exhibit A), the construction manager is contractually bound to provide labor and materials and to complete construction at or below that price.6
The two forms in a line each:
- A102— Owner/Contractor, where the basis of payment is the cost of the work plus a fee with a GMP.1
- A133— Owner/Construction Manager as Constructor, the same payment basis plus preconstruction services; the GMP is fixed through a GMP Amendment.6
The preconstruction phase is where the interesting work happens. While the design is still developing, the construction manager works alongside the architect to advise on the cost of design options, materials, labor availability, and scheduling.2 This is where value engineering lives— and, as Procore notes, that immediate pricing feedback "enables an iterative design approach that leads to a more cost-effective project."2
Which is the setup for the part most explainers skip: this preconstruction collaboration is exactly where design-development speed starts to matter. Which brings us back to the junior designer and her slider.
Why Design-Iteration Speed Is Upstream of the GMP
The GMP is committed on incomplete drawings, so every unresolved design decision becomes priced risk and a likely future change order.5 That's why design-iteration speed sits upstream of the number. The faster a team can explore design options— and freeze the good ones— before the Guaranteed Maximum Price Amendment is signed, the tighter and less-padded the scope it hands the estimator, and the fewer surprises after the price is set.
Quick definition, because the term gets thrown around: a Grasshopper definition is the script. Change one input— a floor-to-floor height, a panel module, a curtain-wall angle— and every downstream component recalculates instantly,10 so a designer tests dozens of variations in the time manual modeling would take for one. Grasshopper itself is a visual programming environment inside Rhino 3D, built by David Rutten at Robert McNeel & Associates: you assemble a program by dragging components onto a canvas and wiring their outputs into other components' inputs, and the saved network is the "definition."9
The chain is simple to state:
- Explore— generate many design options fast, instead of one a day.
- Freeze— lock the ones that work, so more of the design is resolved earlier.
- Price— hand the construction manager a tighter scope when the GMP gets built.2
In practical terms: the more of the design that's resolved when the estimator prices it, the less unknown the contingency has to carry. Now the honest part. Faster iteration of form and geometry doesn't automatically produce better cost numbers. Pricing still depends on subcontractor quotes, market escalation, and the estimator's judgment— none of which a slider replaces. What the slider does is narrower and real: it removes a bottleneck in design development, so more of the design is resolved and frozen before the number is locked. That's the lever. AI amplifies what the designer can do; it doesn't do the designer's job, and it doesn't do the estimator's. You don't need a clever prompt here— you need design judgment, encoded in a definition anyone on the team can run. Speed and rigor. Both are true. All of it matters.
Where does AEC stand on this? Adoption is partial and uneven, and the surveys disagree on how much. A 2025 industry survey reported by the American Society of Civil Engineers found about 27 percent of firms currently use AI in their operations, with roughly 94 percent of those adopters planning to increase that investment within the year.11 A separate 2025 survey of architecture professionals put the figure nearer 46 percent, with another quarter or so planning to start soon.12 Different populations, different samples— so read it as a range, somewhere between a quarter and half of firms, and rising. If you're deciding where AI investment actually pays off first, preconstruction is a strong candidate: a tighter design earlier protects the fee on every GMP.
But here's the part that decides whether any of this compounds: what happens to that Grasshopper definition after the project ships.
From One Clever Script to a Firm-Wide Capability
A junior designer's clever Grasshopper definition is worth something on one project and worth almost nothing across a firm— unless someone makes it durable. Undocumented, unmaintained, quietly broken by the next Rhino update, it leaves when the designer does. The firms that actually compound the advantage are the ones that turn the hero hack into a maintained, documented, reusable capability— a tool the team reaches for on every GMP, not a trick one person knows.
That's a different job than building the automation in the first place. It's version-controlled definitions, naming conventions, a place the team looks first, someone who owns keeping it working. It's the same move a federal grant-writing consultant we work with made when he stopped handing out one-off ChatGPT GPTs "one at a time or through links"— clunky, no quality control— and put a maintained suite of tools on a single platform he actually ran. One-off to system. The one-off automation is the prompt, not the destination.
And it's a margin issue. Better-resolved design earlier means tighter GMPs, which means fewer change orders eating the contingency, which means protected fee. AEC adoption of this kind of work is still uneven— somewhere between a quarter and half of firms, depending on who's surveyed1112— which means the firms that operationalize it now compound an advantage while it's still rare.
If turning a one-off automation into a repeatable system sounds like a different job than building it, that's because it is. An implementation partner can help map the right tooling to your workflows— the actual preconstruction workflows your teams run, not a generic AI rollout— and that's the work Dan Cumberland Labs does with AEC firms.
The point isn't the script. It's making a tool a team habit rather than one person's trick.
And it's worth measuring whether the investment is paying off— in change orders avoided, contingency preserved, fee protected. That's the number behind the number.
A few questions that come up whenever GMP enters the conversation, answered straight:
FAQ— GMP in Construction
What does GMP mean in construction?
GMP means Guaranteed Maximum Price: a contract in which the contractor finishes the project for no more than a set ceiling price and absorbs any cost above it, billing the owner for the actual cost of the work plus a fee.12 (In manufacturing and pharmaceuticals, "GMP" means Good Manufacturing Practice— a different, unrelated term.)
Is a GMP contract the same as a not-to-exceed contract?
They're close— both cap the price— but a true GMP usually adds open-book accounting and a shared-savings clause splitting any underrun, which a plain not-to-exceed price may not include.7 A GMP also travels under the name "construction manager at risk contract."2
Who carries the risk in a GMP contract?
The contractor carries the cost-overrun risk— anything above the GMP comes out of their fee.3 The owner gets price certainty but should still watch quality and the contract's allowances, qualifications and exclusions, and change-order terms, since those are where the "guarantee" tends to erode.5
When is the GMP established?
Before the design is complete— commonly at roughly 60 to 80 percent of construction documents in private work.3 U.S. federal projects generally require at least 75 percent of construction documents (and the GMP set before 100 percent) before the final cost of the work is fixed (48 CFR § 536.7105-2).4 That timing is what lets construction start earlier than lump-sum bidding— and also why scope gaps are the contractor's main risk.5
What is a GMP contingency?
A reserve the contractor builds into the GMP to cover the uncertainty of pricing incomplete drawings and minor unforeseen conditions.2 How it can be used, and by whom, is negotiated in the contract. There's no standard percentage— it varies by project, scope maturity, and risk.
GMP vs. lump sum— which is better for the owner?
A lump sum gives one clean fixed number, but only after the design is essentially complete (around 95 to 100 percent) and with no shared savings if the job comes in low.83 A GMP lets work start earlier on partial drawings, opens the contractor's books, and shares any underrun— but the strength of the "guarantee" depends on how complete and well-defined the design is when the number is set.5
A guaranteed maximum price is a number. It's also a bet on how well the design was thought through before the number got locked.
References
- American Institute of Architects (AIA Contract Documents), "Guaranteed Maximum Price (GMP) Contracts: A Complete Guide" (2023) — https://learn.aiacontracts.com/articles/understanding-guaranteed-maximum-price-contracts/
- Procore Technologies, "Guide to Guaranteed Maximum Price (GMP) Contracts in Construction" (2023) — https://www.procore.com/library/gmp-contracts-construction
- Oracle NetSuite, "How Do Guaranteed Maximum Price (GMP) Contracts Work?" (2023) — https://www.netsuite.com/portal/resource/articles/accounting/guaranteed-maximum-price-contract.shtml
- Legal Information Institute, Cornell Law School, "48 CFR § 536.7105-2 — Guaranteed Maximum Price" (reproducing the GSAM) (2024) — https://www.law.cornell.edu/cfr/text/48/536.7105-2
- El-Khoury Law, "The GMP Contract Trap: 7 Provisions That Erode the Guarantee Before a Shovel Hits the Ground" (2024) — https://www.elkhoury.law/construction-law-blog/the-gmp-trap-7-provisions-that-erode-the-guarantee-before-a-shovel-hits-the-ground
- AIA Contract Documents Help Center, "Summary: A133–2019, Standard Form of Agreement Between Owner and Construction Manager as Constructor… with a Guaranteed Maximum Price" (2019) — https://help.aiacontracts.com/hc/en-us/articles/1500009303082-Summary-A133-2019-Standard-Form-of-Agreement-Between-Owner-and-Construction-Manager-as-Constructor-where-the-basis-of-payment-is-the-Cost-of-the-Work-Plus-a-Fee-with-a-Guaranteed-Maximum-Price
- Wikipedia, "Guaranteed maximum price" (2025) — https://en.wikipedia.org/wiki/Guaranteed_maximum_price
- Built Technologies, "Construction Contract Types: Lump Sum, T&M, Cost-Plus, and GMP" (2023) — https://getbuilt.com/blog/understanding-the-4-common-construction-contracts/
- Wikipedia, "Grasshopper 3D" (2025) — https://en.wikipedia.org/wiki/Grasshopper_3D
- Novatr (with Robert McNeel & Associates community documentation), "How Rhino Grasshopper Is Used in Parametric Designs & Modelling" (2024) — https://www.novatr.com/blog/rhino-grasshopper-in-parametric-design-modelling
- American Society of Civil Engineers (Civil Engineering Source), "Architecture, engineering, construction sector slow to adopt AI, survey shows" (2025-12-18) — https://www.asce.org/publications-and-news/civil-engineering-source/article/2025/12/18/architecture-engineering-construction-sector-slow-to-adapt-ai-survey-shows
- ArchiLabs, "Artificial Intelligence (AI) in Architecture & Engineering" (2025) — https://archilabs.ai/posts/ai-in-architecture