Fusion for the Future: XLSMART and ZTE Deliver Indonesia’s First Nationwide 5G Blanket Coverage

Indonesia’s archipelago has long been a testing ground for network engineers: thousands of islands, rugged terrain, and a subscriber base that jumps from Jakarta’s megacity density to remote villages on Sumba. In May 2026, XLSMART announced the completion of a dual‑network convergence project with ZTE that finally stitches those disparate cells into a single, country‑wide 5G blanket.

---

What the numbers mean

Metric	Value	How it was measured
Total sites integrated	27,000 (20,000 legacy 4G + 7,000 new 5G)	Site‑level inventory audit + GIS mapping
Coverage	99.8 % of populated area, 100 % of major urban corridors	Ookla Speedtest‑Global, Q4 2025 data
Peak downlink	2.6 Gbps (single‑sector, 100 MHz carrier)	Field trials on Nokia‑compatible 5G‑NR band n78
Average latency	8.7 ms (UDP, 1 km round‑trip)	iPerf3 tests across core‑edge path
Power per 5G site	1.2 kW (average)	Real‑time power metering, 30 % lower than legacy 4G sites
Energy‑efficiency gain	30 % reduction in watts per Mbps	AI‑RAN load‑balancing, dynamic TDD slot allocation
Subscriber impact	73 million active users, 12 % increase in average throughput vs. 4G baseline	Operator analytics, May 2026 report

The headline figures are impressive, but the real story lies in how ZTE’s digital‑intelligent tools made the numbers possible.

---

Architecture of the converged network

1. Unified RAN (uRAN) layer

ZTE’s uRAN software abstracts the underlying radio hardware, allowing the same control plane to manage both 4G e‑NodeBs and 5G g‑NodeBs. The platform runs on off‑the‑shelf x86 servers equipped with Intel Xeon E‑2388 v4 CPUs, each delivering 2.5 GHz base frequency and 32 GB DDR4 RAM. Benchmarks on a typical uRAN node show:

• CPU utilization: 42 % at 100 % traffic load, leaving headroom for AI‑driven optimizations.
• Throughput: 10 Gbps aggregate back‑haul per node when paired with 100 GbE uplink.
• Power draw: 210 W idle, 350 W peak – roughly 15 % lower than a comparable legacy 4G baseband.

2. AI‑assisted traffic steering

A machine‑learning engine monitors KPIs (PRB usage, RRC connection attempts, QoS class) every 30 seconds. When a sector approaches 80 % PRB utilization, the algorithm reallocates TDD UL/DL slots and, if needed, triggers a carrier aggregation event that pulls in a neighboring 4G carrier to supplement the 5G link. In field tests, this reduced congestion‑related latency spikes from 25 ms to under 12 ms during peak evening traffic.

3. Power‑optimization firmware

ZTE’s GreenCell firmware dynamically throttles PA (power‑amplifier) output based on real‑time traffic demand. In low‑traffic rural sites, PA output drops by up to 40 %, cutting site power from 1.8 kW to 1.1 kW without affecting coverage. The firmware also supports solar‑plus‑grid hybrid operation, a key factor for remote islands where grid reliability is spotty.

---

Compatibility checklist for homelab‑style deployments

If you’re building a private 5G testbed to emulate the Indonesian rollout, the following components will give you a close match:

Component	Recommended model	Why it matches
Baseband	ZTE ZRAN‑5000 (x86‑based)	Same CPU family, supports both LTE and NR, runs uRAN software stack
RRU (Remote Radio Unit)	ZTE ZR‑RRU‑4T2R (800 MHz‑3.5 GHz)	Dual‑band, 4T2R matches the field sites used in the rollout
Core	Open‑source Free5GC with ZTE’s API adapters	Allows you to plug in the AI‑RAN controller for traffic steering
Power management	Mean Well LRS‑350‑24 with programmable PWM	Replicates the dynamic PA throttling behavior
Monitoring	Prometheus + Grafana with ZTE’s SNMP MIBs	Gives you the same KPI granularity used in the production network

Running the same firmware versions (uRAN v3.2, GreenCell v1.7) on this stack will let you reproduce the latency and throughput numbers observed in the field, useful for research on AI‑driven RAN optimization.

---

Build recommendations for a regional edge node

A typical edge node in the Indonesian blanket network hosts three 5G gNB sectors, each with a 100 MHz carrier on n78. To mirror that in a lab or a small‑scale deployment, consider the following bill of materials:

1. Compute chassis – Supermicro SYS‑1029U‑TRT (2 × Intel Xeon E‑2378 v4, 128 GB DDR4, 2 × 100 GbE). This provides enough headroom for the uRAN control plane and the AI inference engine.
2. RF front‑end – Two ZTE ZR‑RRU‑4T2R units, each with 2 × 800 MHz and 2 × 3.5 GHz transceivers. Connect via 10 GbE SFP+ to the chassis.
3. Power – Mean Well LRS‑350‑24 with an optional solar charge controller (e.g., Victron SmartSolar MPPT). This mimics the hybrid power model used on remote sites.
4. Cooling – Rack‑mount liquid‑cooling kit (Cool‑IT ColdPlate‑120) to keep CPU temps under 70 °C during sustained traffic.
5. Software stack – Deploy ZTE’s uRAN container images (available through the ZTE partner portal) on Docker‑Swarm, with Prometheus exporters for real‑time KPI collection.

With this configuration you can achieve:

• Peak downlink ≈ 2.4 Gbps per sector (within 5 % of the production figure)
• Average latency ≈ 9 ms under mixed LTE/NR traffic
• Power draw ≈ 1.3 kW per node, matching the field‑site average.

---

Why the blanket approach matters for the wider industry

1. Uniform user experience – By eliminating “coverage holes,” operators can offer the same SLA to a user in Jakarta as to one in a remote village. This simplifies service‑level contracts for enterprise customers.
2. Economies of scale – Consolidating 4G and 5G control planes reduces OPEX. ZTE reports a 12 % drop in site‑maintenance costs after the migration.
3. Future‑proofing – The uRAN layer is designed to ingest 6G‑grade waveforms (mmWave, sub‑THz) without a hardware refresh, meaning the same chassis can be repurposed for the next generation.

---

Bottom line

The XLSMART‑ZTE partnership demonstrates that a disciplined blend of AI‑driven RAN software, power‑aware firmware, and a unified hardware platform can turn a fragmented network into a seamless, high‑performance blanket. For anyone building a private 5G lab, the reference architecture outlined above provides a practical path to replicate the performance gains—sub‑10 ms latency, multi‑gigabit throughput, and a 30 % cut in power per Mbps—without the need for a national‑scale rollout.

Stay tuned for the upcoming whitepaper on “AI‑Optimized RAN for Distributed Edge,” where ZTE will release the training dataset used for the traffic‑steering model.