Build vs Buy · Total Cost of Ownership

Building your own Kindo costs more than it looks.
A lot more.

A realistic total cost of ownership for replicating an enterprise AI agent platform in your own cloud account: the platform engineering, the GPU inference for a Deep Hat-class security model, and above all the specialist team you would have to hire, retain, and keep on call.

~$0M
Three-year total cost of ownership
~$0M
Spent before the first user ever logs in
0 mo
To a production-credible v1, best case 15
0 FTE
Specialist hires, in the most contested talent market in tech
The scope

What “build it yourself” actually means

This is not a wrapper around a model API. A Kindo-class platform is a product in its own right, and every piece below is table stakes, not gold plating.

~660,000 lines across 30 backend services

Durable agent orchestration, a model gateway with routing and failover, DLP/PII redaction, policy-based authorization, RAG (vector store, embeddings, indexing), code-execution sandboxing, SSO/SCIM, audit, and metering.

An integrations platform

OAuth broker, a custom provider catalog, and a fleet of MCP servers that must track every vendor's API changes, forever.

Deep Hat self-hosting

H100-class GPU serving with HA and burst capacity, model upgrades, eval harnesses, red-teaming, and guardrail engineering. GPUs alone run $300–700k per year in production posture.

Perpetual upstream maintenance

Every third-party component is pinned, patched, and rebuilt. CVE and breaking-change churn across the stack never stops, and neither can the team that absorbs it.

The team

Nine hires you have to win, then keep

ML-infrastructure and platform engineers are bid on by AI labs and AI-native startups. Expect 4–6+ month searches, agency fees of 20–25% of first-year base, and 15–20% annual attrition in exactly the hardest seats.

RoleFTEFully loaded / yr
Tech lead / engineering manager1$350k
Senior backend / platform engineers3$285k each
ML / inference engineers GPU serving, evals, red-teaming1.5$380k each
Frontend engineers1.5$255k each
DevOps / SRE1$285k
Security engineer DLP, authorization, sandbox review1$300k
Build-phase team9~$2.7M / yr
The money

Three years, eleven million dollars

Year one is pure spend: nothing ships until month 15–18. And it never drops below ~$3M a year after that.

Three-year cost of ownership, $M per year
Hover a segment for detail
Payroll (fully loaded) Recruiting + backfill Cloud: GPU + platform Compliance + tooling
View as table
Year 1Year 2Year 3Total
Payroll (fully loaded)$4.0M$2.2M$2.2M$8.4M
Recruiting + backfill$0.5M$0.15M$0.15M$0.8M
Cloud: GPU inference + platform$0.4M$0.6M$0.6M$1.6M
Compliance + tooling$0.1M$0.15M$0.15M$0.4M
Total~$5.0M~$3.1M~$3.1M~$11.2M
82% of it is people.

This is a people problem, not a cloud-bill problem. Four of every five dollars go to hiring and retaining a team whose market clears against you. The lean, everything-goes-right floor is ~$7M; the realistic ceiling is ~$14M.

The long tail

The costs that never end

DIY estimates always price the build. They almost never price what comes after.

5–6 FTE

The on-call floor

A sane 24/7 rotation needs five to six engineers who each understand the whole system. You cannot staff half a rotation: ~$1.7–2M/yr just to keep the lights on, before a single new feature.

~3 mo

Model half-life

New model generations land every few months. Someone must eval, red-team, quantize, and migrate, or your security AI is stale within a year.

$50–150k/yr

Compliance you now own

The platform carries its own SOC 2, pentest, and audit burden. Evidence collection, tooling, and auditor fees land on your team, every year.

The bottom line

DIY costs roughly $5M and 18 months before the first user logs in, then $3M+ every year after, contingent on winning a hiring market you are structurally disadvantaged in, to own an undifferentiated, perpetually-behind copy of something you could deploy this week.