Telecommunications providers are shifting from proprietary, hardware-based network functions to cloud-native software to reduce dependence on a single vendor. A 2026 survey of 50 Tier 1 operators found that 80% now run cloud-based network functions in at least two mobile core domains, up from 62% in 2024. It also found that 34% have deployed some form of Open RAN. This transition will shape the telecom infrastructure agenda in 2027. Across the tech summit circuit, including TechCrunch Disrupt in San Francisco, attention is moving toward running these networks reliably and economically over the long term. Founders and enterprise leaders have little patience for speculative AI narratives and want practical guidance on building these technologies into their business models.
The focus has shifted from predicting trends to building networks that can handle growing traffic and compute demand. That requires an understanding of both software and the physical limits of existing hardware, from cell sites to data centers. Operators must balance capital allocation, engineering talent, and regulation to stay competitive.
Key areas shaping the 2027 infrastructure agenda include:
- AI-native networking and the push toward greater network autonomy
- Open RAN’s promise of vendor flexibility and the realities of multi-vendor deployment
- Infrastructure investment shaped by power demand, edge computing, and private 5G
- Sustainability measures, early 6G research, and specialized chips for vRAN and AI-RAN workloads
Autonomous Networks and AI-Native Operations
AI-native network automation is changing how operators run their networks. Closed-loop automation takes over tasks that network operations center (NOC) staff have handled manually, detecting problems and acting on them to maintain service quality. Machine learning models, including reinforcement learning, can analyze network telemetry in near real time to tune radio parameters such as transmit power and antenna tilt and allocate resources across network slices with limited human intervention. Full autonomy remains rare, though. TM Forum benchmark figures cited in 2025 showed most providers between Level 2 and Level 3 on the industry’s autonomy scale, which tops out at Level 5. Only 4% reported reaching Level 4.
Automation matters most where traffic varies sharply, such as stadiums and dense urban areas. Self-healing functions can detect faults and reroute traffic or restart failed components automatically, often before customers are affected, reducing the need for round-the-clock manual intervention.
Open RAN and Vendor Lock-In
Open RAN specifications are opening up a part of the network that single-vendor systems have long dominated. By standardizing the interfaces between RAN components, such as the radio unit, distributed unit, and centralized unit, they let operators combine hardware and software from different vendors. In practice, change has come more slowly than promised. After a turbulent 2025 for the sector, Fierce Network reported that early deployments worldwide have mostly been single-vendor configurations that comply with O-RAN specifications, as established operators opt for simpler integration. Disaggregation can still make the market more competitive, letting procurement teams source commercial off-the-shelf servers and components at better price-performance as multi-vendor integration tools mature.
The main benefit is flexibility: operators can replace individual components without overhauling the whole system. Open interfaces also let specialized vendors supply individual elements, including applications for the RAN Intelligent Controller (xApps and rApps). This modular approach helps operators manage the capital spending required for 5G Advanced upgrades and, later, 6G.
Infrastructure Investment in 2027
Heading into 2027, the digital infrastructure sector is focused on capital planning and institutional investment. Events like Connect (X) bring together tower owners, fiber providers, and private equity firms to align financial strategy with technical requirements. Asset valuations increasingly account for the higher power density of AI workloads, a pressure that now extends from hyperscale campuses to operators’ own facilities.
Investors are adding data center scalability and access to renewable energy to the connectivity metrics they use to value assets. Power is a growing constraint. Data center electricity demand rose 17% in 2025 and is projected to double by 2030, according to the International Energy Agency. The IEA also warned that strained planning systems are holding up grid connections and other approvals. Operators face the same constraints. Virtualized mobile core and vRAN workloads run on general-purpose servers in regional data centers and central offices, and edge facilities that host AI inference compete for the same grid capacity as cloud campuses. Power availability is becoming a siting constraint for telecom infrastructure and a valuation factor for tower and edge assets.
Moving Compute to the Edge
Edge computing moves processing from centralized data centers to sites closer to users, such as central offices, aggregation points, and enterprise premises. Applications that need very low latency, such as industrial robotics and connected vehicles, are a major driver. Operators can host these workloads in their own edge facilities and deploy distributed user plane functions from the 5G core there, so traffic is handled locally instead of being routed back to a central site. This reduces the load on long-haul transport networks and lowers backhaul costs.
These same sites can host AI inference for latency-sensitive services such as video analytics, making them natural candidates for AI-RAN.
Private Networks and Enterprise Control
Private LTE and 5G networks give enterprises more control over their internal communications and data. Manufacturers and logistics companies increasingly use licensed or shared spectrum to get reliable connectivity isolated from public network congestion. Shared spectrum models such as CBRS have made this possible for organizations that are not mobile operators, but that access is now less certain. The 2025 One Big Beautiful Bill Act gave CBRS no formal protection, directing the FCC to auction at least 800 megahertz of spectrum in a range that includes the band. Private enterprise deployments still make up a relatively small share of CBRS use, which mobile operators and fixed wireless providers dominate.
Reducing the Environmental Footprint
Energy efficiency is now a core operational priority for operators. The radio access network and the data centers that host core and edge workloads account for much of a network’s power draw, and European policy is adding structure. Under the EU Code of Conduct for the sustainability of telecommunications networks, published in January 2026, participating organizations commit to implementing tracking mechanisms in 2026 and sharing lessons learned by mid-2027, making mid-2027 the first results checkpoint. Techniques such as cell, carrier, and symbol sleep modes let radio sites cut power draw when traffic is low.
The Road to 6G
While the industry continues to roll out 5G Advanced, work is underway to define 6G. Much of the early standards work centers on mid-band spectrum. Sub-terahertz frequencies, which could support data rates above 100 gigabits per second, won’t be part of the first 6G release. They aren’t on the agenda for the 2027 World Radiocommunication Conference, according to researchers from Ericsson, Nokia, NVIDIA, and several universities. These frequencies also present major engineering challenges, particularly high path loss and signal blockage by buildings and other obstacles.
Researchers are developing reconfigurable intelligent surfaces, panels that can redirect signals around obstacles, to address some of these limits. Integrated sensing and communication, which would let networks detect objects and movement while carrying data, is another area under study for 6G.
Compute Power and Chip Innovation
AI development is closely tied to semiconductor advances, and telecom networks increasingly depend on the same silicon roadmap. As operators virtualize the RAN and move mobile core functions onto cloud infrastructure, network performance depends on the servers and accelerators underneath. vRAN deployments rely on general-purpose processors paired with hardware acceleration for compute-intensive Layer 1 processing, and the choice of silicon shapes both capacity and energy use at each site.
AI-RAN takes this further. It runs radio signal processing and AI workloads on shared accelerated computing, so the same hardware at a cell site or edge facility can handle RAN functions during peak traffic and AI inference when capacity is spare. Industry groups such as the AI-RAN Alliance are exploring whether this model can turn network sites into revenue-generating compute assets, though the commercial case is still being proven.
Denser compute brings a thermal challenge that is especially acute at the network edge. A 2025 Uptime Institute survey of more than 1,000 data center professionals found that 22% use direct liquid cooling, while most operators still rely on air cooling. Respondents generally saw liquid cooling as necessary above 20 kW per rack. Central offices and edge sites were built for lower-density telecom equipment, so adding accelerated vRAN or AI inference may require cooling and power upgrades that operators need to build into site budgets. Silicon choices now affect network capex, site power budgets, and vendor options, not only data center performance.
Conclusion
In 2027, operators that focus on measurable outcomes rather than projections will be better placed. That means raising automation beyond Level 3 to lower operating costs, using open interfaces to broaden vendor choice, and planning power and cooling for vRAN, core, and edge sites before adding AI workloads. Spectrum decisions, including the future of CBRS, and early 6G standards work will set the limits on what networks can deliver next.
