The Cloud-Exit Math Federal Agencies Aren't Doing

Federal cloud strategy has spent a decade answering one question: how do we get into the cloud? The migrations, the authorizations, the lift-and-shift programs all point inward. The question that determines an agency's actual strategic position is the one almost nobody has run the numbers on: what would it cost to get out, or to move? Not because agencies are planning to leave the cloud — most are not and should not — but because the cost of leaving is the truest measure of how much leverage an agency has surrendered, and how much it is paying for the privilege. An agency that has never calculated its exit cost does not know its own negotiating position, and it is almost certainly weaker than the agency assumes.

The question the migration narrative skips

The federal cloud conversation has a directional bias baked into it. Every incentive — modernization mandates, the appeal of getting off aging data centers, vendor enthusiasm — points toward migration. The result is a decade of accumulated decisions optimized for getting in, with almost no attention paid to the reversibility of those decisions. Exit cost was never a design criterion, so it was never measured, so it grew unchecked.

This matters even for an agency with no intention of leaving, because exit cost is what sets the ceiling on the agency's leverage in every renewal, every renegotiation, every pricing conversation. A provider negotiating with a customer who cannot afford to leave is negotiating with a captive. The agency that has quantified its exit cost — and reduced it where it can — negotiates from a different position entirely, even if it never exercises the option. The math is leverage whether or not the agency ever moves.

"You don't calculate exit cost because you're planning to leave. You calculate it because it's the truest measure of how much leverage you've already given away — and it's almost always more than you think."

Why exit cost stays invisible

Exit cost is not a single number on an invoice. It is a diffuse, accumulated liability spread across architecture decisions, and that diffuseness is exactly why it stays invisible until the agency tries to move.

• It builds up silently. Each individual decision to use a provider-specific service is locally reasonable and individually cheap to reverse. The aggregate, after years of such decisions, is an architecture deeply entangled with one provider — and no single decision looks like the one that locked the agency in.
• It's nobody's job to track. No role owns 'exit cost' as a metric. Architects optimize for the workload in front of them; finance tracks spend, not reversibility. The liability has no owner, so it has no number.
• The provider has no reason to surface it. The economics of the relationship favor the provider when exit cost is high and invisible. There is no party in the arrangement with an incentive to make the number legible to the agency.

Where the lock-in actually accumulates

Lock-in is not mainly about the compute and storage that are genuinely portable. It accumulates in the layers above the commodity infrastructure, where the provider-specific value lives and where moving gets expensive.

• Data gravity. The more data lives in a provider's storage and the more services depend on it being there, the more expensive and disruptive it is to move. Data gravity is the heaviest single contributor to exit cost, and it grows every day the agency operates.
• Managed-service dependency. Provider-specific managed services — the ones that made development faster — are the ones with no clean equivalent elsewhere. Every managed service adopted is a strand of the rope, convenient going in and costly coming out.
• Operational muscle memory. Teams build skills, tooling, and runbooks around one provider's way of doing things. That accumulated operational capability is real value and a real switching cost; it doesn't transfer for free.
• Integration entanglement. The connections between systems get built using provider-native integration, so moving one system means untangling its connections to everything else. This is where the federal integration layer and the cloud-exit question meet.

Running the math agencies skip

Calculating exit cost is not exotic; it is an analysis most agencies simply have not commissioned. It asks what it would actually take to move the major workloads — re-architecting away from provider-specific services, the data egress and transfer, the re-integration, the re-skilling, the parallel-running during transition, and the time and risk of the move itself. The output is not a single figure but a profile: which workloads are genuinely portable, which are deeply entangled, and where the largest reducible costs sit.

The value of running it is rarely the decision to leave. It is the clarity. An agency that has run the math knows which of its dependencies are cheap optionality and which are expensive lock-in, and it can make deliberate choices about which to accept and which to reduce. It enters renewals knowing its real position. It designs new workloads with reversibility as a known cost rather than an ignored one. The analysis converts a diffuse, invisible liability into a managed, visible one — which is the whole point.

Optionality as the actual strategy

The strategic goal is not cloud exit. It is optionality — the deliberate preservation of the ability to move, used as leverage and as resilience, whether or not it is ever exercised. An agency that designs for optionality keeps the commodity layers portable, treats each provider-specific dependency as a priced decision rather than a default, and maintains a current understanding of its own exit cost. That agency negotiates from strength, adapts when the technology or the economics shift, and is not surprised by the bill when circumstances force a move. The agency that never does the math has made a strategic decision by default — to be captive — without ever deciding it. A decade of getting into the cloud was the easy half of the question. The agencies that will hold the stronger position over the next decade are the ones now doing the math on the harder half.^[2]