The $10B VA–Oracle EHRM Is Out of Money and Time — and Your Oncology Data Is Stuck in the Middle

The most consequential oncology health-IT story this week isn't an FDA clearance or a flashy AI trial. It's a procurement document.

On Tuesday the Department of Veterans Affairs proposed expanding its Oracle Cerner electronic health record modernization contract for one blunt reason: it is running out of money and time. The ceiling already sits near $10 billion. VA concedes it will exhaust what's left by the first quarter of fiscal year 2027 and needs an increase. Contract ordering currently runs only through May 2028; VA wants three additional one-year firm-fixed-price periods, pushing performance to May 16, 2031.

Read the sourced rationale and it reads like a confession: "unanticipated complexity of the effort," "extensive site-specific customizations," technical glitches, training failures, clinical functionality gaps "with potential impact on patient safety." As of July 2026, VA had deployed to 14 of 156 sites. Thirty-six are scheduled to go live between August 2026 and January 2028. The remaining ~120 sites cannot be completed within the existing contract's ceiling or period of performance.

This isn't an EHR story. It's the story of where the nation's largest integrated oncology delivery network will be computing for the next decade.

The Federal Oncology Data Spine Is Stuck in a Half-Finished Migration

Here's what the health-IT trade press misses: the VA National Oncology Program is one of the largest cancer-treatment enterprises in the country. Millions of veterans receive cancer care through it, and its reach now extends into community care referrals and clinical trial infrastructure. Every day those sites run on a mixed, half-migrated environment — legacy VistA/CPRS alongside Oracle Cerner at the 14 deployed hospitals.

For an oncologist, that's not an administrative nuisance. It's a data-integrity problem with clinical consequences:

• Therapy continuity breaks. A veteran treated at a deployed Oracle site transferred to a legacy CPRS hospital doesn't move cleanly across the seam. Chemotherapy histories, staged imaging, and mutation panels live on two incompatible systems.
• Trial matching degrades. Precision oncology depends on a complete longitudinal record. Fragmented records mean a patient with an actionable alteration can be missed by trial-alerting and eligibility screens that query a partial datasource.
• Community care coordination suffers. VA's own secretary frames the pain point: executives calling to "move up on the list" for deployment, precisely because the split environment cripples internal work and community care handoffs.

The FY2027 House Military Construction–VA appropriations bill would pour $3.4 billion into this effort. More money into a system that, per VA's own documents, was mis-costed from the start.

The Market Meme Is Wrong: Epic Isn't Coming to Save Us

Here's the contrarian angle that matters most. The dominant 2026 narrative is that Oracle Health is collapsing and Epic will consolidate the entire market — one study this week put 71% of health systems on an Epic-first AI strategy. If that story were true, the VA crisis would end with VA ripping out Oracle and going Epic.

It didn't. Last month VA ran a formal market search naming Epic, Veradigm, Athenahealth, and Meditech — and concluded no other source than Oracle could meet VA's requirements for an EHR system and associated support without "negatively impact[ing] VA's ability to timely provide veteran care."

Make no mistake about what that conclusion buys: at least three more years (to 2031) of Oracle Cerner as the federal oncology backbone, customizations and all. The half-migrated, mixed-CPRS/Oracle reality isn't a transitional state to be suffered through — it is the operating environment oncology informatics teams will build around for years. Vendor-switch fantasy doesn't fix that; interoperability engineering does.

What This Means for Oncology Leaders

Draw the practical readouts:

The federal market is locked, not contested. Stop waiting for "the federal win" that rewrites vendor leverage. The VA decision signals that switching costs — real, patient-safety-bearing switching costs — lock incumbent EHR footprints in place, even when the incumbent is flagging.

Interoperability is the only durable hedge. The mixed environment will persist. Oncology programs that invest in FHIR-based aggregation, vendor-neutral archives, and longitudinal data spines — not the next EHR loyalty bet — are the ones that keep their data portable regardless of which giant's logo sits on the screen.

Clinical-safety flags are a governance signal, not noise. VA documents explicitly cite "potential impact on patient safety" from clinical functionality gaps. When the country's largest public cancer enterprise is running on a system that averted clinical gaps, community oncology practices with far fewer informatics resources should treat workload-safety validation as a first-class requirement for every AI and EHR decision they make.

The takeaway isn't "Oracle failed." It's that no vendor "wins" the federal oncology backbone by being good — they win by being impossible to remove. That's the real economics of oncology health IT, and it's the one the hype cycle keeps refusing to confront.

Your data strategy will outlast every vendor's contract. Build it like the contract is already out of money.