Procurement budget forecasts for 2026
An early read on federal, state, and enterprise procurement budgets for 2026. Where the RFP dollars are moving, what categories are expanding, and what that implies for the proposal teams covering them.
This is the research team’s first long read of 2026. The question: where are procurement dollars moving in the year ahead, and what does that imply for the proposal teams covering each segment?
The post synthesizes three public data sources: federal obligations data from USASpending and FPDS, state-level procurement data aggregated through NASPO, and the publicly excerpted portions of Gartner and Forrester IT-spending forecasts. Enterprise-side numbers are harder to ground in public data, so we have been more cautious there and flagged the assumptions.
Federal — flat top line, reallocated mix
Federal total procurement obligations in FY2026 came in at a level that was roughly flat against FY2025, within the noise band that continuing-resolution cycles produce. The flat top line hides a meaningful mix shift.
The categories expanding: cybersecurity services, cloud migration, AI assurance and evaluation services, and health IT modernization. The categories compressing: legacy on-premise infrastructure refresh, traditional staff-augmentation contracts, and some of the custom-software-development vehicles that were being re-platformed anyway.
For a proposal team covering federal civilian agencies, the practical implication is that the RFPs landing in Q1 and Q2 are weighted toward services contracts with technical-assurance deliverables. That changes the evaluation surface. A 2025 RFP for cloud migration would have asked for platform certifications and case studies. A 2026 RFP is asking for those plus a model-assurance narrative — who owns the AI controls, what the audit trail looks like, what happens when the model is wrong.
The defense-side picture is separate from the civilian-side picture and worth calling out explicitly. Defense obligations remain elevated relative to the civilian baseline, weighted toward software-defined capabilities, unmanned systems integration, and the categories of AI that fall under autonomy and decision-support rather than document processing. For proposal teams covering DoD, the Q1 queue is heavier than the civilian queue and the evaluation bars are different — the compliance matrices are longer, the security review is deeper, and the past-performance requirements are stricter. Teams that cover both civilian and defense from the same proposal function need to budget the effort split carefully; the two segments do not share templates well.
The continuing-resolution dynamic is worth naming. FY2027 is likely to spend a meaningful portion of Q1 and Q2 under a CR rather than a full appropriation, based on the pattern of recent years. A CR compresses discretionary-start new procurement while leaving existing vehicles (IDIQs, BPAs, GWACs) as the primary procurement surface. For proposal teams, that argues for prioritizing capture work against existing vehicles the company is already on and de-prioritizing pursuit of new-start programs until the appropriation resolves.
State — surge in AI-adjacent procurement
State-level procurement data is noisier because the collection pipelines are inconsistent across states, but the signal from the NASPO-aggregated data is clear: states are starting AI-procurement programs in volume. A sample of 12 states we pulled data on shows 24 new AI-adjacent procurement vehicles announced in Q4 2025, against five in Q4 2025.
The “AI-adjacent” label covers several distinct things. Some are genuine AI-system procurements (a state DMV buying a document-processing pipeline, for example). Others are traditional services procurements with AI-disclosure and model-governance clauses bolted on. The second kind is the larger category by count and will be the dominant pattern in Q1 and Q2.
For proposal teams responding to state RFPs, the practical change is that the compliance matrix for a Q1 2026 response needs an AI-disclosure column even when the procurement is not obviously an AI procurement. The 8-stage pipeline we wrote about last year has a new line item at the compliance stage. We will update that post to reflect it.
The state picture is not homogeneous. A cluster of states — California, New York, Washington, Massachusetts, and Colorado in our sample — are issuing AI-specific procurement vehicles and AI-disclosure clauses at a much higher rate than the sample median. A second cluster, primarily in the Southeast and Mountain regions, has issued very few AI-specific vehicles but is starting to include AI-disclosure language in general services procurements. For a vendor active across multiple states, the practical implication is that the compliance language needs to be per-state for the cluster-one states and can be templated for the cluster-two states, at least for the next two quarters. Expect the cluster-two states to catch up through 2026 and 2028.
State-level contract values are also worth calling out. The median state AI-adjacent procurement in our sample was meaningfully smaller than a comparable federal vehicle, typically in the low-seven-figure range rather than the mid-seven or eight-figure range. That changes the economics of pursuit. A state AI procurement is often a worthwhile bid for mid-market vendors who would not win a competitive federal equivalent. For large vendors, the state pipeline is valuable as volume but individual bids do not warrant the pursuit cost a federal bid does. Size your capture investment accordingly.
The vehicle-vs-new-start shift
Cutting across federal and state, a secondary pattern in the data is worth calling out. The share of procurement dollars flowing through pre-existing contract vehicles — IDIQs, GWACs, BPAs, state-cooperative vehicles — relative to new-start competitive procurements is rising. In our federal sample the ratio of vehicle-flowed dollars to new-start dollars increased by roughly 6 percentage points from FY2024 to FY2026. The direction is not new; the rate of change is.
The operational implication for proposal teams is that pursuit strategy in 2026 should over-weight work to get onto the vehicles that matter for the company’s target segments. A company that is on the right vehicle will see a steady flow of task-order RFPs; a company that is not on the vehicle will miss most of the opportunity flow regardless of how strong the response capability is. This is a capture-planning question more than a proposal-execution question, which is why it is easy to miss when reviewing procurement data from a proposal-function vantage.
Enterprise — procurement centralization is accelerating
Enterprise-side numbers are the most assumption-laden portion of this post, and we have flagged the methodology note at the bottom accordingly. Public forecasts from Gartner and Forrester (the excerpted portions only — we do not have access to the paid research) point at IT-spending growth in the mid-single digits for 2026, weighted heavily toward platform consolidation and AI tooling.
The procurement implication we are seeing in our customer base, which we can observe directly but cannot generalize from, is that procurement functions are centralizing decisions that used to sit with line-of-business buyers. An RFP that would have been issued by a single department two years ago is now being issued by a central procurement function with an evaluation committee that includes IT, security, compliance, and the original business owner.
That is not a new trend — it has been building for five years — but the 2025 to 2026 step-up is larger than previous years, and the implication for a vendor’s proposal response is that the executive summary is being read by four or five distinct personas instead of one or two. Writing for a multi-persona evaluation committee is a different craft than writing for a single business-owner reader. Sarah has a post on that coming later in January.
What this implies for proposal teams
Three adjustments to the Q1 operating plan.
Rebuild the compliance-matrix template. Add an AI-disclosure column and a model-governance column. Both will be asked for, in some form, in more than half of the federal and state RFPs landing in Q1. The work of adding them to the template once is an hour. The work of adding them to every live response in the first week of January is a full week.
Update the past-performance record. Past-performance records that describe 2025 and 2025 awards need an AI-related paragraph, even if the awarded work did not involve AI. Evaluators are reading those sections for a sense of how the vendor operates in an AI-inflected environment. A past-performance writeup with no mention of AI in 2026 reads as stale, fairly or not.
Budget time for buyer-side Q&A in AI-specific questions. Our sample of Q4 2025 RFPs shows that buyer-side Q&A rounds now include AI-specific clarifications in roughly 60% of responses — questions about training data, model provenance, human-in-the-loop, and refusal behavior. Those are not questions most proposal teams’ SME roster is ready to answer on a two-day turnaround. Identify the internal owner now.
Rework the pricing narrative for multi-year AI-infused services. Multi-year procurements that include AI capabilities are being structured with contingent pricing more often than before — base-year pricing that is firm, out-year pricing that has adjustment clauses tied to model-provider cost changes, and optional modules for additional AI capabilities added in out-years. For the proposal team, that means the pricing narrative now has to explain not just what the customer pays this year but what the pricing mechanism is for year two and three. Vendors whose pricing narratives are single-year snapshots will read as unprepared.
What to watch next
Three signals we are watching closely in Q1 to confirm or refute the above.
The first is the FY2027 appropriation timing. A full-year appropriation passed before Q2 is a meaningfully different environment for federal procurement than a series of continuing resolutions extending into Q3. The difference for proposal teams is large enough that Q1 staffing plans should flex against both scenarios rather than committing to one.
The second is the Q1 state AI-procurement volume. If states issue 30+ new AI-adjacent vehicles in Q1 2026 (against the 24-in-Q4 baseline), the surge is durable and teams should staff for it. If the Q1 number comes in at 15 or below, the Q4 number was a one-quarter anomaly and the staffing plan can be more conservative.
The third is the enterprise-side centralization signal. We cannot measure this from public data; we can measure it indirectly from the composition of the evaluation committees we observe in customer responses. A committee that includes a named AI-governance role is a committee that will be asking AI-specific questions throughout the evaluation. Watch the committee composition of the first two or three enterprise RFPs you respond to in Q1.
The caveats
Three caveats on the numbers.
The federal obligations data is one-quarter lagging; FY2027 started in October and the cleanest view we have is still partial. The state-level aggregation is a 12-state sample, not a census. The enterprise-side numbers lean on public excerpts of paid research that we cannot fully replicate. Treat the federal and state numbers as directional and the enterprise numbers as informed guesses.
We will revisit this post in April with Q1 actuals. That is when the FY2027 federal number starts to resolve against the forecast, and when we will know whether the state AI-adjacent surge is a Q4 ripple or a durable trend.
One note on what is not in this post. We have not written about healthcare-sector procurement here — that is a separate research note, landing in a few days, with the level of detail the segment deserves. We have also not covered international procurement, which has its own dynamics tied to DORA in the EU and the UK’s evolving AI-procurement guidance. Those are on the Q1 research calendar and will be standalone reads rather than sections inside a broader post.
The purpose of this post is to give proposal teams a head start on Q1 planning. Treat the directions as hypotheses to test against your own pipeline in the first two weeks of the quarter; the actual data your pipeline generates will be a better signal than any external forecast. Research reads a quarter ahead; operators read last week.