5G Hits 3 Billion Connections — But Telcos Are Confident on AI and Slow to Deploy It

By Hector Herrera | May 16, 2026 | Vertical: Telecom

Global 5G connections surpassed 3 billion in 2025, marking 5G as mainstream infrastructure rather than a frontier buildout. But new data from Omdia reveals the sector's defining paradox: 94% of telecom leaders are confident they'll capture AI-driven growth, while 66% haven't started deploying AI-driven network operations. The confidence-implementation gap is the largest structural tension in the industry right now.

The 3 Billion Milestone in Context

Three billion 5G connections is a number worth pausing on. The journey from zero to three billion connections happened in roughly six years — faster than 4G LTE reached equivalent scale. Growth has been led by China (which accounts for approximately 60% of global 5G subscriptions), followed by South Korea, the United States, and parts of Western Europe. Asia-Pacific and Latin America drove subscriber growth in 2025, and telecom connectivity revenues rose overall as a result.

The milestone matters because 5G was always a two-part story: first build the infrastructure, then monetize it. The infrastructure phase is largely complete in major markets. The monetization phase is what operators are struggling with.

What AI-Native Networks Would Actually Do

The telecom industry's AI bet centers on a specific vision: networks that manage themselves. In current operations, detecting a cell tower failure, diagnosing its cause, rerouting traffic, and dispatching a repair crew involves multiple human steps and significant latency. An AI-native network would handle the detection, diagnosis, and traffic rerouting automatically — faster and with fewer errors.

The value proposition for operators is OpEx (operating expense) reduction. Labor and network maintenance costs are massive at scale; AI-driven operations promise to compress them significantly. For enterprise customers, AI-native networks unlock capabilities that have been promised for years but rarely delivered:

• Network slicing — dedicated virtual network capacity for specific applications (a factory floor gets guaranteed low-latency connectivity, independent of consumer traffic)
• Ultra-reliable low latency (URLLC) — the technical standard that enables remote surgery, autonomous vehicle coordination, and industrial robot control
• Predictive maintenance — fixing network issues before they become outages

The Gap Between Confidence and Action

The Omdia data is stark. Of the telecom leaders surveyed:

• 94% express confidence in capturing AI-driven growth
• 66% have not commenced implementation of AI-driven network operations
• 61% have not deployed advanced 5G capabilities (the features that make 5G substantially different from 4G)

This is not a technology readiness problem — the tools exist. It is a systems integration problem. Legacy network architectures, built on proprietary hardware and closed vendor ecosystems, are deeply difficult to retrofit with AI-native software layers. The operators who built out 5G on traditional infrastructure are now facing the cost and complexity of another major technical transition before the first one has fully paid off.

The gap is widest in North America and Europe, where legacy architecture is most entrenched and integration friction is highest. Greenfield markets — operators in regions building new 5G networks from scratch — have an advantage: they can design AI-native from the start.

Revenue Is Already Moving

Despite the deployment lag, telecom revenue grew in 2025. This suggests operators are finding some monetization on current 5G infrastructure even without AI-native capabilities — primarily through consumer subscriptions and early enterprise deals. But the growth ceiling is lower without AI operations.

Carriers that achieve AI-native operations will have structural cost advantages that allow either margin expansion or competitive price reductions — both of which matter in a market where consumer price sensitivity is high and enterprise contracts are hard-fought.

What to Watch

The 6G development cycle is already underway. Major operators are publishing 6G architecture papers and working with standards bodies to define the next generation's specifications. The risk for operators that don't close the AI deployment gap now is entering the 6G cycle while still paying the technical debt of 5G integration. That compounding problem would be expensive.

Watch also for vendor consolidation in the network AI space. Ericsson, Nokia, and Samsung have all launched AI-native network products; hyperscalers including AWS, Google Cloud, and Microsoft Azure are also competing for network AI infrastructure contracts. The market is fragmenting before it converges, and operators picking vendors now are making decade-long commitments.

Hector Herrera covers AI in telecommunications, infrastructure, and the systems connecting them. He is the founder of Hex AI Systems.