The Ericsson Mobility Report documents 3 billion 5G subscriptions. The number worth watching is buried in the traffic analysis: uplink is becoming the choke point, and AI is the reason why.
(Ericsson Mobility Report; 2026)
Q1 2025 to Q1 2026
(Ericsson Mobility Report; 2026)
grew faster than downlink in 2025
(Ericsson Mobility Report; 2026)
offerings now available globally
(Ericsson Mobility Report; 2026)
by 2031 under medium AI adoption
(Ericsson Mobility Report; 2026)
Mobile networks were dimensioned for download-dominant traffic. AI agents, smartglasses, and autonomous systems run in the opposite direction. The gap between today's uplink capacity and what AI workloads require is not a forecast problem. It is an infrastructure investment decision that enterprise technology leaders need to make now.
The headline from the Ericsson Mobility Report June 2026 is subscriber scale: 5G subscriptions passed 3 billion in the first quarter of 2026, and half of all mobile data traffic globally now runs over 5G. Those are consequential numbers, and the coverage will treat them as the story. They are not the story.
The finding that matters for enterprise technology leaders is buried in the traffic analysis section, where Ericsson examined uplink versus downlink growth rates across 55 service providers in 2025. In 43 of those 55 networks, uplink grew faster than downlink. In 17 of the 55, uplink grew at more than 1.5 times the downlink rate. Mobile networks have been designed, dimensioned, and commercially structured around the assumption that people download. That assumption is no longer accurate, and AI is the reason the inversion is accelerating.
I have written about this structural shift from the enterprise network side, most recently when covering the Cisco WAN research on agentic traffic and the Ericsson Cradlepoint W2255. The Mobility Report frames it from the carrier infrastructure side, which changes what the observation implies for CIOs and CTOs making network decisions.
The conventional view in most infrastructure planning is that uplink capacity is a future concern tied to augmented reality, autonomous vehicles, and smartglasses at scale. The Ericsson data pushes back on that framing. Uplink growth is already outpacing downlink growth for most service providers measured, driven primarily by behavior that exists today: higher-resolution smartphones uploading to cloud storage, short-form video creation and live streaming, AI assistant interactions that continuously send sensor and voice data upstream, and increasing enterprise IoT telemetry. The AI-on-phone behavior driving current uplink growth will compound, not plateau, as agentic applications become standard.
Ericsson's scenario modeling adds specificity. Under a medium AI adoption scenario, where 43 percent of subscribers use AI assistants on smartphones and 7 percent use the same applications on smartglasses, uplink traffic would be three times higher in 2031 compared to 2025. Under a high adoption scenario, that figure rises to five times. The networks currently being operated were not designed for any of these scenarios.
"Uplink capacity is not keeping pace with expected traffic growth and service requirements. Without substantial increases in uplink capabilities and more efficient resource utilization, service providers will struggle to deliver a consistent user experience for new emerging use cases."
That is from the Mobility Report itself, describing the structural gap. What makes it analytically significant is the architecture reason: mobile Radio Access Networks (RANs) are optimized around frequency division that allocates substantially more spectrum to downlink. Fixing uplink capacity at scale requires hardware investment, not just software configuration. 5G Standalone (SA) features and uplink-optimized antenna selection can close part of the gap in the near term, but the modeling suggests even advanced multi-antenna techniques leave a meaningful shortfall against where AI-driven demand is heading.
The second thread worth pulling from this report is the pace at which differentiated connectivity, meaning 5G standalone network slicing with guaranteed service levels, is crossing from pilot into commercial offerings. Six months ago, 65 out of 118 network-slicing offerings tracked were commercially available. The June 2026 update shows 84 out of 151, representing an annual growth rate of around 58 percent.
The Ericsson analysis is candid about where the gap is. Network readiness has crossed a threshold. The question service providers are now asking is not whether to deploy 5G SA but when and with what service portfolio. The execution problems are on the go-to-market side: customer segmentation, pricing logic, and in-moment engagement models that translate guaranteed connectivity into something a customer will pay for.
The SoftBank trial at the 2026 Formula 1 Japanese Grand Prix at Suzuka Circuit illustrates what the production architecture looks like. SoftBank deployed five network slices simultaneously on shared 5G SA infrastructure: a high-quality general connectivity slice, an extended reality (XR) slice, a point-of-sale payment slice, a millimeter-wave Wi-Fi slice for general subscribers, and a millimeter-wave wireless camera slice for broadcast. The result was that premium services and general attendee connectivity coexisted without degradation. General attendees on 5G SA saw 4.1 times faster downlink speeds compared to 2025 and 14.6 times faster uplink. The payment slice maintained stable low-latency connectivity throughout peak congestion periods.
The operational lesson from Suzuka is one that any enterprise deploying private networks should internalize: monitoring granularity matters as much as network capability. SoftBank found that standard 15-minute key performance indicator (KPI) collection intervals were too coarse to detect and respond to service-level agreement (SLA) breaches in real time. Per-minute observability with closed-loop automation was what made slice management operationally meaningful. Network slicing without updated monitoring infrastructure leaves a ceiling on how well quality can actually be guaranteed.
Fixed Wireless Access (FWA) has become the clearest near-term 5G monetization case, and the June 2026 data is striking. The share of FWA service providers offering the service over 5G has grown 14 percentage points in one year, reaching 71 percent. That is the largest single-year increase in four years. In North America, the three largest carriers added nearly 1 million FWA net connections in the first quarter of 2026 alone, and their combined FWA base now exceeds 17 million connections. In India, Jio and Airtel together are approaching the same number.
The Ericsson analysis draws a meaningful distinction between FWA markets that are succeeding and those that are not. Successful markets share two characteristics: 5G mid-band population coverage sufficient to deliver a fiber-competitive experience, and a monetization model based on speed tiers rather than data buckets. Speed-based tariff plans are now offered by 57 percent of FWA service providers globally, up from 51 percent a year ago. The regions where FWA monetization is struggling are precisely those where 5G mid-band coverage remains thin or where pricing models have not evolved beyond volume caps.
The relevance for enterprise planners is not the consumer FWA story itself, but what it signals about carrier investment priorities. Carriers that are successfully monetizing FWA have the revenue incentive to continue investing in mid-band coverage and 5G SA infrastructure. Carriers that are not are likely to lag on the network capability improvements that enterprise AI workloads will require.
The 5G subscriber milestone is real, but the constraint that will determine whether enterprise AI on mobile networks succeeds is uplink capacity. The infrastructure investment needed to close that gap is not uniform across carriers or geographies. Where a carrier sits on the FWA monetization and 5G SA deployment curve is a reasonable proxy for where they are on the uplink investment curve.
A section of the Mobility Report, written with research partner Arthur D. Little, surfaces a finding that reframes the enterprise AI adoption conversation. Among more than 100 large-enterprise chief officers surveyed across North America, Europe, and Asia, 88 percent expect their AI solutions to depend on real-time data. Only 18 percent have widely adopted cellular technologies that provide secure, reliable, real-time mobile connectivity. Only 8 percent have fully scaled AI across multiple business areas.
The gap between AI ambition and AI execution in enterprises is not primarily a model quality problem or a data science talent problem. It is a connectivity and infrastructure readiness problem that enterprise technology leaders are underweighting. The Mobility Report frames this as an opportunity for service providers to reposition from infrastructure vendors to strategic partners. From an enterprise buyer's perspective, the implication is different: connectivity must be evaluated as part of the AI deployment architecture, not as a pre-existing commodity assumption underneath it.
This is the same argument I made in the context of private 5G for physical AI deployments, and it applies equally to enterprise AI applications that depend on mobile endpoints, field devices, and distributed sensors. The mobile network is not neutral infrastructure for these workloads. Its uplink capacity, latency characteristics, and slice architecture determine whether the application can deliver its intended outcome.
The first implementable specifications for 6G, formally designated as 3rd Generation Partnership Project (3GPP) Release 21 based on International Telecommunication Union (ITU) International Mobile Telecommunications (IMT)-2030, are targeted for finalization by late 2028 or early 2029. The first commercial 6G services are expected around 2030. Early adopters are projected to be the United States, China, Japan, South Korea, and the Gulf Cooperation Council countries, all of which launched 5G relatively early.
What makes this relevant for enterprise planning now is the trajectory of the uplink problem. The Mobility Report describes 6G features, including uplink/downlink decoupling and contention-based uplink access, as the structural response to the asymmetry that AI traffic is creating. Those capabilities are not available in current 5G deployments. The gap between where uplink demand is heading and where current network configurations can deliver is the gap that 6G is being designed to close. Enterprise technology leaders whose AI roadmaps extend into the 2028 to 2032 window need to factor that transition into their infrastructure planning.
The 6G discussion is also already centering on commercialization and use cases from the start, not technical capability. The Mobility Report notes that agentic AI will be native at 6G introduction, smartglasses will enter the 6G era with their full capabilities, and many chipset vendors are positioning for next-generation participation from day one. That commercial alignment suggests the 5G-to-6G transition could produce a device supercycle, an unusually strong multi-year upgrade wave, rather than the incremental adoption curve that characterized 5G's first three years.
The Ericsson Mobility Report documents the uplink gap structurally and projects it will widen materially through 2031. For any enterprise betting AI agent workloads on mobile connectivity, whether through private 5G, carrier networks, o
