O-RAN vs Traditional RAN: Total Cost of Ownership in 2026

By 2026 there is enough operator data to stop arguing about O-RAN TCO from first principles. Vodafone, Rakuten, AT&T, DISH, NTT DOCOMO, Deutsche Telekom, and BT have all published outcomes — some glowing, some sobering, all instructive.

Here is what the numbers actually say.

What the brochures get right and wrong

The vendor pitch: O-RAN is 30-40% cheaper because of commodity hardware and multi-vendor competition. The reality: capex savings are real, opex savings are conditional, and integration costs are usually underestimated by a factor of two.

The headline 30% savings is usually a steady-state number — year 4 or 5. Year 1-2 is often more expensive than traditional RAN.

Capex breakdown

Hardware

O-RAN RUs from Mavenir, NEC, Fujitsu, Samsung, and emerging Indian/Korean vendors are 15-30% cheaper than equivalent traditional vendor RUs at like-for-like specs in 2026. Massive MIMO 64T64R parity has largely been reached, though traditional vendors still hold a slight edge on highest-tier 5G-Advanced massive MIMO products.

DU/CU on COTS servers (Dell, HPE, Supermicro, Lenovo with Intel Xeon or AMD EPYC, plus Marvell/Nvidia accelerators for L1) costs roughly 60-70% of equivalent vendor proprietary baseband. The accelerator matters — software-only DUs on x86 CPUs are still inefficient at high cell counts.

Software

This is where the picture gets messy. Traditional vendors bundle. O-RAN forces unbundling: separate licenses for CU, DU, near-RT RIC, SMO, possibly each xApp.

The sum can exceed the bundle. Operators who negotiate hard and use open-source components where possible (O-RAN SC RIC, open-source SMO components) come out ahead. Operators who buy commercial everything sometimes pay more than for traditional RAN software.

Opex: where the dragons live

Integration cost

This is the line item operators consistently underestimate.

Multi-vendor by definition means no single throat to choke. When the RU from vendor A and the DU from vendor B do not interwork after a software update, someone has to coordinate the fix. Operators handle this in three ways:

1. Build an internal SI team. 30-100 engineers depending on scale. Real cost, real value, slow to staff. Vodafone, AT&T, BT.
2. Hire a system integrator. Wind River, VMware/Broadcom, Capgemini, TCS, Infosys, Accenture. Cheaper to start, harder to extract from later.
3. Use a Managed Open RAN partner — Mavenir, Rakuten Symphony — that delivers a multi-vendor stack as a single deliverable. Lower risk, less differentiation.

DISH spent ~$10B building its O-RAN network and a substantial portion was integration. Greenfield operators face the steepest curve; brownfield operators can stage migration.

Operations

Steady-state opex is where O-RAN claims its biggest wins. Reported operator data:

• Site visits: Down 15-30%. Software-defined RUs and remote configuration reduce truck rolls.
• Spare parts: Down 20-40%. Commodity hardware, broader supplier base.
• Energy: Down 5-10%. AI-driven sleep modes (rApps for energy saving) outperform traditional vendor schemes by a measurable but modest margin.
• Software upgrades: Mixed. Cloud-native upgrades are faster but require mature CI/CD; manual upgrades on poorly cloudified stacks are slower than traditional.

Vodafone Italy's published 2025-2026 figures cite ~25% opex savings on their O-RAN footprint versus comparable traditional sites, but only after the third operating year.

Operator outcomes summary

Rakuten Mobile (Japan)

The original cloud-native O-RAN greenfield. Built fast, hit major coverage gaps in early years, ARPU lower than incumbents. Capex per pop reported as substantially lower than incumbents. Symphony spun off as a vendor, which is the part that has been profitable. Lessons: greenfield O-RAN works technically; the business case depends on subscriber acquisition, not just network cost.

Vodafone

UK and Italy O-RAN deployments at scale by 2026. Hard requirement: 30% of network O-RAN by 2030. Published TCO advantage of 10-25% at scale, mostly in opex. Ericsson and Nokia shifted strategy in response — both now offer O-RAN-compliant products.

AT&T

The Ericsson partnership announced in late 2023 reframed AT&T's O-RAN strategy: Ericsson as the prime, with O-RAN-compliant interfaces enabling future multi-vendor RU sourcing. Capex and opex outcomes per the 2025 disclosures are favorable but not transformative — closer to 10% TCO savings versus pure traditional, with the option value of multi-vendor RUs not yet exercised.

DISH (now EchoStar)

Greenfield O-RAN at national scale. Cost overruns, coverage struggles, but technical proof of concept is delivered. Total spend ~$10-12B. Lesson: building a national O-RAN network from zero is harder than it looks even with vendor partners.

Deutsche Telekom

Germany O-RAN town deployment — small scale, technical learning. Conservative deployment. Concluded multi-vendor works but operational maturity lags traditional by ~2 years.

When O-RAN wins on TCO

O-RAN's TCO case is strongest when:

1. Greenfield or major refresh. The cost of swapping out working traditional RAN early kills the case. Refresh cycles or new buildouts amortize differently.
2. Scale. Below ~5,000 cells, the integration overhead per cell is too high. Above ~20,000 cells, it amortizes well.
3. Mature internal SI capability. Operators who own integration in-house extract more value than those outsourcing.
4. Mid-band spectrum. Massive MIMO parity is best in mid-band. mmWave and ultra-low-band are still slightly behind in O-RAN ecosystem.
5. Regulatory pressure. Markets where governments push diversification (UK, US, Germany, Japan) tilt the field — vendor diversity has political value beyond pure TCO.

When traditional RAN wins on TCO

1. Small operators. <2,000 cells, integration cost dominates. Single-vendor traditional makes economic sense.
2. Tight upgrade cycles. Traditional vendor stack with mature ops is faster to upgrade than a multi-vendor stack with immature CI/CD.
3. Specialized deployments. High-band mmWave dense urban, ultra-rural ultra-low-band. Vendor-specific products outperform the O-RAN ecosystem in 2026.
4. Regulatory simplicity. When the operator's only requirement is "work," not "diversify," traditional is the lower-risk path.

The integration cost equation

A rough working model for greenfield O-RAN at scale (>10,000 cells):

• Hardware capex: 70-80% of traditional
• Software capex: 90-110% of traditional (depends on negotiation)
• Year 1-2 integration NRE: 5-10% of total program cost as one-time premium
• Year 3+ opex: 75-90% of traditional steady-state

Net 5-year TCO: 80-90% of traditional. Net 10-year TCO: 70-80% of traditional. The longer the horizon, the better O-RAN looks.

For brownfield partial swaps, the math is harder: integration cost is concentrated in the first 2 years and the upside is diluted across the unconverted footprint. Many operators staging brownfield migrations report year-3 break-even with cumulative savings only emerging in year 5+.

What the numbers do not capture

• Optionality. Multi-vendor sourcing reduces lock-in risk. Hard to price, real value.
• Innovation velocity. xApp/rApp ecosystem moves faster than traditional vendor roadmaps. Compounds over time.
• Talent. Engineers who run O-RAN networks are scarcer and more expensive in 2026, but the talent pool is growing.
• Geopolitical hedging. Reducing dependence on any single vendor or country is a strategic asset for some operators.

> O-RAN is not cheaper than traditional RAN on day one. It is cheaper over a decade, conditional on scale, internal capability, and the operator's willingness to absorb integration cost early.

The operators winning on O-RAN TCO are the ones who treated it as a multi-year program with sustained internal investment, not a procurement decision.