Captive offshore centers — wholly-owned subsidiaries operating engineering teams in lower-cost geographies — won the 2010s for a specific reason: at scale, they produce lower per-engineer cost than any other delivery model. The capex amortizes; the institutional knowledge accumulates; the long-run unit economics are unbeatable when the engagement runs 5+ years at 100+ engineers.
The 2020s have complicated the picture. Talent geography is more fluid. Capital structures favor opex over capex for many companies. Engagement horizons are shorter. Workload is more variable. None of this makes captives obsolete; it does mean the model fits fewer engagements than it did.
This piece walks through the capex-vs-opex framing, when each is right, and why the practical answer for most enterprises is hybrid — captive plus a VDC layer for the workloads the captive isn't sized for.
The capex-heavy model
Captive offshore is capex-heavy across five categories:
- Setup capex. Facilities lease, IT infrastructure, regulatory registration. $500K–$2M depending on scale.
- Recruiting capex. $30K–$50K per engineer hired in offshore markets. Multiplied by team size.
- Onboarding capex. 8–12 weeks of subscale productivity per new engineer. Visible as opex but functions as capex (recoverable over engineer's tenure).
- Operational capex. Cap-ex for ongoing facilities maintenance, IT refresh, compliance certifications.
- Knowledge capex. The institutional knowledge accumulating in the team's heads. This is real value that doesn't appear on a balance sheet but is part of why captives win at long horizons.
The first four amortize over engagement duration. The fifth amortizes over engineer tenure (mid-tenure employees carry more knowledge value than recent hires).
If your engagement runs 5+ years at scale, this capex pays back through lower per-engineer cost than any opex-only alternative. Below 24 months, the capex doesn't recover.
The opex-only model
VDCs invert the capex shape entirely:
- No setup capex. No facilities to lease, no IT to provision, no regulatory registration. The platform handles all of this at the platform layer.
- No recruiting capex. Pre-vetted bench eliminates per-engineer recruiting cost.
- Minimal onboarding capex. 5–10 day pod ramp vs 8–12 week individual ramp. Pod-level onboarding amortizes over the engagement, not per engineer.
- No operational capex. Platform absorbs facilities, IT, compliance.
- Distributed knowledge capex. Knowledge accumulates at the pod level (preserved across rotations) and in the codebase (transferred at commit). Less concentrated than captive's single-team knowledge but more transferable.
Cost flows match value flows. There's no capex to recover before delivery starts paying back.
The financial-structure question
Beyond the operational fit, the choice is also a finance-team decision. Three considerations:
Cash-flow profile
Captive: large upfront cash outflow during build phase, declining over operate phase as the team scales. Some companies have constrained cash that won't support this profile.
VDC: even cash outflow per milestone. Linear with delivery, not lumpy.
Capital allocation strategy
Some companies prefer capitalizing engineering investment — capex shows on balance sheet, depreciates over time. Others prefer opex — current-period expense, simpler accounting.
Captive favors the first preference. VDC fits the second.
Risk tolerance
Captive carries reversibility risk: closing or downsizing means severance, facility unwind, and capex write-offs. Some boards have low tolerance for this.
VDC has no reversibility cost — exit at any milestone.
The hybrid: captive + VDC layer
The realistic answer for enterprises with substantial engineering investment:
- Captive handles steady-state, high-volume, long-running work. Maintenance, regulated batch, embedded ops. Capex amortizes over 5–10 year horizons.
- VDC layer handles new builds, modernization, burst capacity. Anything where the captive's capex shouldn't have to absorb the variability. VDC pods spin up for transformation programs, ship, then spin down.
- Governance is unified at the program level. Both report into the same engineering leadership; both follow the same milestone discipline; both contribute to the same codebase.
This pattern works because each model serves the workload it's structurally suited for. Pure captive misses the burst-capacity case. Pure VDC misses the deep-knowledge accumulation case.
When NOT to add a VDC layer
If your captive is healthy and your workload is genuinely steady-state:
- Engineering velocity is consistent across quarters.
- Roadmap doesn't include new builds or modernization in the next 18 months.
- Capacity needs aren't burst-shaped.
- The captive has talent depth across all needed specialisms.
In this scenario, a VDC layer can we waited. The captive handles everything. Adding a VDC creates coordination overhead without solving an actual problem.
For most enterprises, this scenario is rare — roadmaps rarely stay this stable for 18+ months — but where it exists, leave the captive alone.
Frequently asked questions
How is "captive offshore" different from "ODC"?
Mostly synonymous. Some firms use "ODC" for vendor-operated offshore centers and "captive" for wholly-owned ones; others use the terms interchangeably. The economic shape is similar regardless of naming.
What about the talent-retention argument for captives?
It's real but overstated. Captive teams have employee turnover too — typically 12–18% annually in major offshore tech hubs. The retention advantage exists but isn't the order-of-magnitude difference some captives claim.
Can a VDC produce captive-level institutional knowledge?
Different shape, similar effect. VDC pods retain knowledge at the pod level (preserved across rotations) and in the codebase (documented during work). The knowledge isn't in specific employees' heads; it's in pod-level operating patterns and the code itself. For most enterprises, this is sufficient. For deep strategic IP, captives still win.
What about the geopolitical risk of offshore captives?
Real and increasing. Concentrating operations in one geography creates exposure to local regulatory, currency, and political shifts. VDC's distributed talent model spreads this risk; captive concentrates it. Worth factoring into multi-year scenarios.
Where to start
If you have an existing captive that's healthy, evaluate adding a VDC layer for the workloads the captive isn't sized for. The hybrid pattern is well-tested and additive.
If you're considering standing up a captive from scratch, run the math at your specific scale and horizon. For engagements under 15–20 engineers or 24-month horizons, VDC almost always wins financially. For longer/larger engagements, captive economics may win.
For a workload-fit and cost-modeling conversation, schedule a 30-minute call. For comparative model context, see VDC vs GCC, VDC vs ODC, and VDC vs BOT.