S18.5 · Government, Defense & Public Sector
IT infrastructure and modernization services sold to agencies, under budget pressure even as cloud and AI spend grows.
This segment covers IT infrastructure and modernization services sold to civilian and defense agencies. The combined federal IT market is commonly cited at roughly $100-135B annually across civilian and defense spending (OMB/GovWin estimates), with federal cloud spend specifically at about $8.3B in fiscal 2025 (Data Center Dynamics, 2025). Two things are true at once and pull in opposite directions: continuing resolutions and government-efficiency-driven contract reviews are flattening topline spend, while cloud and AI modernization keep growing at double digits off a small base. Which side of that divide a vendor's revenue sits on is most of the analysis.
Growth is flattening under fiscal 2025-26 continuing resolutions and the contract reviews reported since February 2025 (Nextgov); cloud and AI-modernization spending still grows at a double-digit rate, though from a small starting point relative to the overall IT budget. FedRAMP authorization defines addressability, and the contractor base concentrates in Washington DC and Northern Virginia. The two layers have very different structures. Infrastructure is highly concentrated among three hyperscalers — AWS GovCloud, Azure Government and Google — which have captured the underlying compute and storage layer outright; the systems-integration and modernization layer above them fragments among Leidos, GDIT, SAIC, Accenture Federal and Booz Allen Hamilton.
Value flows from hyperscaler infrastructure downstream to the systems integrators delivering agency-specific implementation, and the billing models diverge accordingly: integration work is still mostly time-and-materials or cost-plus, while infrastructure spend increasingly runs as a fixed subscription. Revenue quality is the sector's weakest here — this is the clearest example of annually appropriated, competitively rebid revenue anywhere in it, with task orders recompeted individually against multi-award vehicles such as Alliant and CIO-SP4 and no backlog protection at the contract level. Entry requires FedRAMP or StateRAMP authorization, IL5/IL6 authorization for Defense Department cloud workloads, and cleared staff embedded at the agency.
Adjacent territory includes government professional-services consulting, commercial cloud and software-as-a-service vendors entering government through FedRAMP, and the cloud infrastructure supporting the intelligence community. Incumbent integrators are holding position on existing multi-award vehicles; commercial software vendors earn their way in by clearing the FedRAMP authorization bar.
Agentic AI attacks the time-and-materials billable hour directly — the model underneath help-desk support and legacy-code modernization. Agentic coding tools substitute for the large cleared development benches that systems integrators have historically staffed and billed by the hour, and the moat eroding is bench size itself — the ability to mobilize hundreds of cleared developers — rather than any proprietary technology. That opens the door for small, AI-native, FedRAMP-authorized government software vendors to compete directly against large-headcount integrators on task orders that once required scale to win, at cost structures the incumbents cannot match without shrinking their own billable base. FedRAMP authorization of AI tools has proliferated since 2024 and continues through 2026; full disruption of the underlying labor model sits within two to five years.