MTN’s freshly cloud‑native Enterprise Value Analytics platform, EVA 3.0, is more than a systems refresh — it is an aggressive bet that
telecom competition will increasingly be decided by data intelligence, not just radio coverage. The operator says EVA 3.0, rebuilt on Microsoft Azure and anchored on Azure Databricks, now processes roughly
22 billion records per day, orchestrates
800+ analytics workflows fed from
~1,700 data streams, and is protected by Microsoft Defender — figures the company and trade outlets have repeated as evidence of a production deployment at continental scale.
Background / Overview
MTN’s migration of its Enterprise Value Analytics (EVA) system to a lakehouse architecture on Azure is the culmination of a multi‑year strategic push with Microsoft that dates back to their 2022 strategic alliance and subsequent joint initiatives around cloud, skills and telco‑grade services. The corporate narrative is straightforward: move telemetry, OSS/BSS signals, billing and CRM data into a single governed platform so operations, product and analytics teams can act in near real time. The result, MTN says, is EVA 3.0 — a cloud‑native lakehouse built on
Azure Databricks with Delta Lake semantics and governance/controls layered by Microsoft Defender and Azure identity services. The company positions the South Africa rollout as a
blueprint for replication across other MTN markets in Africa and the Middle East. Independent trade coverage has echoed those same high‑level claims.
Why MTN chose Azure and Databricks
The technical logic
The telco problem is threefold: extreme ingestion velocity, a mix of streaming and batch processing needs, and strict regulatory and privacy constraints. A managed Spark platform (Databricks) over an object lake (ADLS/Delta Lake) solves those first two at scale: autoscaling compute, robust schema evolution, and ACID‑like guarantees for analytical tables. Azure’s identity, Defender and regional cloud footprint address parts of the governance and residency challenge. That architecture is now a mainstream path for large operators replatforming telemetry and analytics workloads.
Commercial and operational factors
Beyond raw engineering, Microsoft brings commercial incentives: integration with Azure data services, co‑engineering support, and the ability to show a telco‑grade reference in the region. For MTN, the trade‑off is speed and functionality against longer‑term dependency on a single hyperscaler and managed services such as Databricks. That vendor coupling simplifies day‑one delivery but raises governance and portability questions later.
What EVA 3.0 actually does for MTN
EVA 3.0 is framed as an active control plane for business decisions — not merely a historical data store. The platform’s stated capabilities include:
- Near‑real‑time detection of network and service issues by correlating telemetry from NOC/NMS with customer experience data.
- Automated workflows that can move insights into engineering fixes, customer offers, or fraud detection flows.
- A governed model lifecycle to support responsible AI: registries, staged deployments, lineage and monitoring to reduce risk in customer‑impacting models.
Operationally, MTN claims EVA 3.0 delivers materially faster processing and greater agility in testing and deploying models and queries — a shift from batch reporting to data as
infrastructure for daily operations. Multiple industry outlets repeat the primary performance claims.
The scale claim: what “22 billion records a day” implies — and what it doesn’t
Those headline numbers are important because they set expectations about throughput, cost and operational maturity. If sustained, 22 billion events/day implies:
- Massive parallelism and aggressive autoscaling policies for streaming and micro‑batch jobs.
- Rigorous partitioning, compaction and file‑management strategies to avoid the classic small‑file problem on cloud object storage.
- Mature SRE practices: p95 ingestion latencies, back‑pressure handling, failure‑isolation, and runbook automation.
- Sophisticated FinOps: job‑level cost attribution, retention tiering, and alerting on abnormal spend.
However, there is an important verification gap: these figures are
company‑reported operational metrics and — to date — have not been accompanied by independent third‑party benchmarks or a public engineering case study that discloses ingestion percentiles, job failure rates, or cost per TB‑month. Treat the numeric claims as credible scale indicators rather than independently audited throughput guarantees.
Security, governance and the “responsible AI” promise
Baseline controls stated by MTN
EVA 3.0 reportedly uses Azure identity (Entra/Azure AD), Unity Catalog‑style metadata and lineage, Microsoft Defender for workload protection, and model registries for lifecycle control. These are the right tool types for a telco operating in regulated markets and with sensitive PII and location telemetry.
Where toolsets aren’t enough
Tools are necessary but not sufficient. Turning Defender or Unity Catalog into a comprehensive governance program requires:
- Active policy‑as‑code, continuous entitlement reviews and CIEM (cloud identity governance) to manage privilege creep.
- Transparent data residency mapping and contractual SLAs for law‑enforcement requests and cross‑border transfers.
- Regular red/blue exercises, pen tests and SOC playbooks that ingest cloud alerts into a SIEM for threat hunting (for example, Microsoft Sentinel).
MTN positions responsible AI as a program goal, but concrete governance artifacts (published policies, audit attestations, or independent audits) are not yet widely available. That absence is precisely the gap regulators and enterprise customers will scrutinize as EVA expands beyond South Africa.
The business case: from MTTR to new revenue streams
EVA 3.0 aims to shift several telco KPIs:
- Reduce mean time to repair (MTTR) through earlier fault detection and automated remediation.
- Improve retention and ARPU via targeted, behaviour‑driven offers that combine network quality signals with billing and CRM.
- Create new B2B products: anonymised network‑insights and analytics services for enterprises and public sector customers.
These are familiar claims for a governed lakehouse: the architecture enables them, but realizing the commercial upside requires
measurement and linkage between platform metrics and product KPIs. MTN says it plans to use EVA 3.0 as a blueprint that can be rolled out to other markets to achieve group‑level scale benefits.
Risks, trade‑offs and the hard operational work ahead
No telco cloud transformation at this scale is risk‑free. The practical trade‑offs include:
- Vendor lock‑in: heavy use of managed Azure services and Databricks accelerates delivery but increases exit complexity. Exportable formats (Delta, Parquet) reduce friction, but not commercial dependence.
- Cost governance: poorly tuned Spark jobs, long retention windows, and unchecked autoscaling can create runaway monthly bills. FinOps discipline is essential.
- Regulatory complexity: MTN spans markets with vastly different data‑protection and lawful‑access regimes; centralising telemetry can collide with local residency rules unless tenancy and contractual controls are in place.
- SRE and operational maturity: hundreds of interdependent workflows demand site reliability engineering, synthetic testing, chaos engineering, error budgets and robust runbooks — skills that certification alone does not guarantee.
- AI governance risks: models can embed bias or make high‑impact decisions (pricing, credit, churn offers). A model registry is only the first step — continuous bias testing, explainability tooling and human‑in‑the‑loop gates are necessary.
Industry commentary and technical assessments repeatedly stress that the platform design choices MTN has made are sensible
if accompanied by disciplined governance, cost control and operational processes — not as standalone guarantees of long‑term value.
How credible is the “largest telco cloud in the Middle East and Africa” claim?
MTN and Microsoft describe EVA 3.0 as the largest telco cloud implementation in the Middle East and Africa; several news outlets repeated the phrase after MTN’s announcement. The claim is competitive and directional, and is supported by internal throughput figures and the scale of ingestion MTN reports. However, independent cloud throughput rankings for telcos in the region are not centrally published; verifying the assertion requires defined comparators (peak sustained ingest, number of workflows in production, or audited cost and SLAs). For procurement or partner evaluation, independent binning of evidence — telemetry extracts, audited throughput tests, or an engineering case study — is the only reliable way to verify the superlative.
Practical checklist for telcos and IT leaders evaluating MTN’s blueprint
If your organization is considering a similar migration, insist on proof and prepare for the operational realities. Key actions:
- Request proof‑of‑value metrics:
- Ingest latency percentiles (median, p95), job failure rates, and representative pipeline telemetry.
- Cost attribution: compute hours, storage TB‑months, egress and cost per million events for a defined period.
- Start with a narrow pilot that exercises edge cases: burst ingestion, low‑latency alerting and automated remediation.
- Embed FinOps from day one: job tags, quotas, automated cluster shutdown and monthly cost reviews aligned to product KPIs.
- Design data residency and lawful access from the outset: region‑aware tenancy, Key Vaults, and contractually bound SLAs for cross‑border requests.
- Invest in SRE: synthetic tests, chaos drills, runbooks for portal/identity/control‑plane failures and clearly defined escalation paths.
- Operationalize responsible AI: model registries, bias testing, explainability, canary deployments and human‑in‑the‑loop approvals for high‑impact actions.
Where MTN’s human capital investment matters
Technical tooling alone rarely wins large migrations. MTN’s narrative emphasises a concentrated upskilling program and a Cloud Centre of Excellence to coordinate deployment, training and reuse of templates across markets. MTN has reported more than
1,250–1,350 Azure certifications among its engineers in 2025 — a material indicator of investment in capability. Certifications reduce the “unknowns” in operating new services, but they must be paired with structured rotations, on‑call experience and documented runbooks to build durable institutional knowledge.
Broader implications for the African cloud and telecom ecosystem
MTN’s EVA 3.0 is a signal to three constituencies:
- Hyperscalers: a regional win for Microsoft and Databricks that demonstrates the commercial viability of telco‑grade, cloud‑native analytics in Africa.
- Competitor telcos and vendors: a rising bar for analytics capability and a template for modernization, which can accelerate competitive offers but also strengthen hyperscaler leverage.
- Regulators and enterprises: a reminder that cloud modernization shifts the policy conversation from “can we store data in the cloud?” to “how do we audit, govern and hold cloud operators accountable at scale?” Governments, regulators and enterprise customers will press for artefacts — data‑residency maps, audit trails, and independent attestations — as EVA expands to other markets.
What to watch next
- Publication of an MTN or Microsoft technical case study that discloses representative ingest latencies, pipeline SLAs and cost attribution. This would materially strengthen the public record behind the headline numbers.
- The first production EVA 3.0 rollout into a non‑South African MTN operating company — evidence of the blueprint’s replicate‑and‑adapt capabilities.
- Any third‑party audits or compliance attestations confirming runtime protections and model‑governance claims.
- Evidence of measurable business outcomes (reduced MTTR, improved ARPU, new B2B revenues) tied to EVA metrics in public disclosures or investor materials.
Verdict — significant progress, but proof will be operational
EVA 3.0 is a credible and consequential move: the technical architecture (Databricks + Delta Lake on Azure) is well matched to telco telemetry patterns, the partner ecosystem provides deep integration and commercial incentives, and MTN’s investment in people reduces transition risk. If MTN can translate the platform into measurable reductions in MTTR, demonstrable revenue uplift and robust cross‑market compliance, EVA 3.0 will stand as a genuine blueprint for telco modernization in Africa. At the same time, the most important caveats are operational and contractual. The headline scale numbers — 22 billion records a day, 800 workflows, 1,700 feeds — are plausible and consistent across MTN and trade reports, but they remain
company‑reported metrics until substantiated by independent benchmarks or published telemetry extracts. Observability, FinOps discipline, SRE maturity and clear legal artifacts governing data residency and access will determine whether EVA 3.0 is a durable competitive advantage or an expensive, tightly coupled experiment.
EVA 3.0 reframes the telco data conversation in Africa: it treats data as an active instrument of decision‑making rather than an archive. For MTN and for the industry, the next phase is less about launching architectures and more about proving operational outcomes under audit — and showing that cloud‑scale analytics can be governed, affordable and resilient across the diverse legal and technical landscapes of the continent.
Source: TechTrendsKE
MTN Takes Its Data to the Cloud, and the Rules of Control Start to Change