Every few months, somewhere in Europe, a telecom operator has the same meeting.
The CEO wants cloud revenue. The CTO says 'we have the infrastructure — we just need a portal, billing, and automation.' Somebody pulls out a whiteboard. A year and a half later, they have a half-working portal, a billing system that disagrees with the invoice PDFs, and a provisioning workflow that still requires two engineers on Slack.
We've watched this cycle play out across multiple operators. The pattern is so consistent it's almost a physical law.
The Seven Systems Every Telco Rebuilds
Every cloud business needs the same translation layer between raw infrastructure and a customer who can click 'order.' Every team builds it from scratch. Every team underestimates it the same way.
1. Customer Portal
Three iterations: marketing landing, customer portal, admin console. Each gets outsourced to a different agency. None of them share auth. By month eight, someone proposes a 'unified design system sprint.'
2. Identity & Tenant Model
The team picks Keycloak because it's free, then spends months discovering that the free part is the binary, not the integration. Tenant hierarchy, MFA, RBAC, and service-to-service auth each get rewritten twice.
3. Product Catalog & Pricing
'It's just a product database' — said every architect in week one. By week twelve, pricing depends on region, tier, usage bracket, tenant class, and a spreadsheet that only the head of sales understands. The catalog becomes a bespoke rules engine.
4. Provisioning Orchestration
The most expensive line item, always. OpenStack, Kubernetes, DNS, certificates, IP allocation, notification, cleanup-on-failure. Every operator writes their own bash-scripts-behind-a-UI, then six months in, discovers they need sagas, retries, and compensation logic. The rewrite is the real cost.
5. Metering & Billing
The team picks a billing SaaS, signs a multi-year contract, then spends months writing the integration that maps their usage events to the SaaS's billing model. They discover invoice accuracy lives or dies on the metering layer — which the SaaS does not provide.
6. Support & Ticketing
Usually starts with a Jira project. Ends with a custom ticket portal because B2B customers refuse to log into Jira. Then integration with on-call, Slack, and the customer's email — each a two-week sprint that becomes two months.
7. Observability & Day-2 Ops
This is the silent killer. After launch, you need a team that understands every layer to keep the thing running. The first real outage teaches the CFO what Day-2 actually costs.
Why The Bill Looks The Same Every Time
Every telco has the same fleet: OpenStack or VMware clusters, a couple of datacenters, a DNS provider they've had since 2011, a billing system designed for voice minutes, and an OSS/BSS stack nobody fully understands. The cloud business layer needs to translate all of that into self-service products.
That translation layer — portal, auth, catalog, provisioning, billing, support, observability — is roughly 80% identical across operators. The remaining 20% is branding, pricing, and the specific quirks of their infrastructure.
But each team rebuilds 100% of it. Because the 20% that's unique gets used to justify not buying the 80% that isn't.
The Real Cost Isn't The Software
Here's the trap nobody warns you about: the dominant cost of the build path is not engineering. It's the months spent not selling.
While your team is in month eleven of debugging the provisioning saga, your competitor — who bought the platform — is in month eleven of onboarding customers. By the time your portal is stable enough to launch, theirs is on its second pricing iteration based on real customer feedback you don't have.
The financial gap between 'platform that ships in a quarter' and 'platform that ships in a year and a half' isn't measured in license fees. It's measured in cumulative ARR you never collected, customer relationships you never built, and product feedback loops that never started.
That's the gap that eats cloud initiatives.
'But Our Needs Are Unique'
No, they're not.
We've now connected PLATFORMA to OpenStack, OpenShift, VMware, Proxmox, and Hetzner — five different infrastructure stacks owned by operators who all believed their setup was uniquely complex. In every case, the differentiation lived in:
- Product packaging — tiers, regions, contracts
- Pricing model — flat, usage, hybrid, commit
- Brand & domain — white-label
- Support SLA & workflow
Not in the undifferentiated plumbing — the auth system, the Kafka topology, the saga orchestrator, the metering pipeline, the retry policies. That plumbing is identical, and rebuilding it is where months and budget vanish.
The Pattern In Every Postmortem
We keep notes from every integration. The recurring failure mode is the same across operators who tried the build path:
'We underestimated how many small systems need to work perfectly together for the customer to get their VM in under two minutes. Every one of them was a two-week project we didn't plan for.'
Multiply 'two-week project we didn't plan for' by forty, and you have your timeline.
What Actually Matters To Build Yourself
Not everything should be bought. Here's what we see successful operators build themselves — because it's where the actual business lives:
- Service catalog & pricing strategy — this is commercial, not technical
- Customer segmentation & sales motion
- Support playbooks specific to your customers
- Regulatory & sovereignty posture (especially for EU government)
- Partner integrations with your existing OSS, CRM, finance
Everything else — the portal, the provisioning engine, the event bus, the billing automation, the AI agent layer that runs Day-2 ops — is table stakes. It should be a platform you connect to, not a project you staff.
The Uncomfortable Conclusion
If your board has approved a cloud initiative and the plan involves hiring a large engineering team and a twelve-month timeline, you're not building a cloud business. You're building a tooling team that happens to have infrastructure.
The telcos winning in this market right now figured out early that the cloud business layer is not their moat. Their customers, their network, their pricing, and their support are.
Everything between the order button and the invoice can be bought.
Everything it takes to decide what's on that invoice, and why a customer should pay it, cannot.
Know the difference, and you'll ship in a quarter instead of a year and a half — with your runway intact.