Congress Subpoenas Oracle Health as VA Medical-Records Costs Nearly Triple

The US House Veterans’ Affairs Committee voted 19–0 to subpoena Oracle Health executives — chairman and CTO Larry Ellison and CEO Mike Sicilia — over the escalating cost of the Department of Veterans Affairs’ electronic health record modernisation.

A unanimous vote on a subpoena is unusual. Committees divide on most things; they divided on nothing here.

The numbers

The contract was signed with Cerner in 2018 at $10 billion. The current ceiling stands at $27 billion — roughly a 170% increase — with completion pushed to 2031.

Oracle acquired Cerner in June 2022 for $28.3 billion and inherited the deal. Following the acquisition the VA paused the rollout to renegotiate. Deployment has since resumed at a fraction of planned sites, with more than 160 VA medical centres still awaiting deployment before 2031.

What actually angered the committee

Cost overruns on large federal IT programmes are, unfortunately, ordinary. Three specific things turned this one into a subpoena.

First, the 2022 testimony. Sicilia told Congress that Oracle would absorb costs beyond the original $10 billion ceiling. The current $27 billion figure is difficult to reconcile with that commitment, and members treated the gap as a credibility question rather than a budgeting one.

Second, how they found out. Committee members learned of the increase through news coverage rather than from Oracle or the VA. For a body whose function is oversight of taxpayer spending, discovering a $17 billion escalation in the press is a direct challenge to its role.

Third, Oracle declined to attend, citing timing constraints. “Oracle’s leaders must appear before this committee and answer for this contract,&rdquo> said Rep. Maxine Dexter. Declining an invitation is what converts a hearing into a compelled appearance.

Why VA record modernisation is genuinely hard

The technical difficulty here is real, and worth stating before assigning blame entirely to a vendor.

The VA operates one of the largest integrated health systems in the world, and it was running VistA — a homegrown electronic record developed in-house over decades, deeply customised at individual facilities, and in some respects ahead of commercial products when it was built. Replacing it means migrating decades of clinical data, retraining an enormous workforce, and reconciling local variations that grew up over many years, all without interrupting care for millions of veterans.

There is also the Department of Defense dimension: a core rationale for the programme is a shared record between DoD and VA so a service member’s history follows them into veteran status. Interoperability across two vast federal departments is a large part of the complexity, and a large part of the value if it works.

What has gone wrong in practice

Deployment has proceeded at a small number of sites, and the pace tells its own story. A programme confident in its product does not take seven years to reach a modest fraction of facilities. Rollout was paused after the Oracle acquisition and resumed cautiously, with more than 160 medical centres still to come before the 2031 deadline — a schedule requiring a dramatically faster rate than has been achieved so far.

The Defense Department did the same migration

The most useful comparison is not another vendor but another federal customer. The Department of Defense deployed the same underlying Cerner platform across its military treatment facilities as MHS Genesis, and completed its rollout while the VA’s remained at a fraction of planned sites.

The DoD experience was not smooth — early sites reported significant usability and productivity problems, and the programme drew sustained criticism. But it finished. That undercuts the simplest defence available to any party here, which is that the task was impossible. It was difficult, and it was completed once at comparable scale.

The differences are real: the VA is larger, its facilities more heterogeneous, and VistA more deeply customised locally than the systems DoD replaced. Still, the existence of a completed federal deployment of the same product changes the nature of the questions the committee can ask.

Why this matters beyond the VA

The programme has become the reference case for large-scale health IT replacement, and the lesson is not primarily about one vendor.

Health record migrations are systematically underestimated because the visible product — software — is the smallest part of the job. The bulk is data migration, workflow redesign, training, and reconciling how differently clinicians at different sites actually work. Fixed-price contracts signed against that reality tend to become cost-plus contracts in practice, and the incentive to disclose escalation early is weak on all sides.

What happens next

Oracle’s compliance with the subpoena remains to be seen, and the company had not issued a statement at the time of reporting. The immediate stakes are political and reputational; the practical stakes are whether the remaining rollout accelerates, and at what further cost.

For veterans, the consequences are mundane and real: care delivered on a mix of old and new systems for years longer than planned, with the fragmentation that implies. Business and policy news, not investment advice.