Key takeaways from IDC’s Telco Forum 2026 in Barcelona point to an industry moving from pilots and promises to structural change. The transformation reaches into capital allocation, operating models, data architecture, and the definition of what a network is. The headline is blunt: training captured attention, inference will determine winners. Telcos control the last mile, the edge footprint, and the trust fabric that governments and large enterprises require. That combination turns connectivity into intelligence and positions national operators at the center of sovereign digital infrastructure.
Four Megatrends Shaping Telecom Through 2030
Structural Reinvention Is Accelerating
Four strategies now run in parallel: the shift to TechCo models that treat the network as a programmable platform; delayering that separates ServCo, NetCo, and InfraCo to improve asset productivity; consolidation to fix sub-scale economics; and a new form of convergence that mixes fixed, mobile, and satellite bundles to defend ARPU and cut churn. The common goal is capital efficiency with commercial agility. Operators treating these as portfolio levers rather than one-way bets are pulling ahead.
Network Investment Is Tapering and Redirecting
The 5G non-standalone buildout has crested in many markets, and fiber-to-the-premises peaks are fading. IDC forecasts a 1.5% decline in global telecom CAPEX in 2026, bringing the total to approximately $320 billion, with CAPEX intensity projected to fall from 22% in 2024 toward 18% by the end of the decade. Spend is shifting, not disappearing. Satellite partnerships reduce the need for extreme coverage CAPEX. A CAPEX-to-OPEX mix is rising as operators adopt software, public cloud, and third-party AI infrastructure. More free cash is flowing into shareholder returns and targeted digital infrastructure bets.
AI Adoption Is Moving to Inferencing and Sovereign AI
The center of gravity is shifting from training to running models at scale in real time, close to data. IDC’s survey data shows that AI compute spending is approaching a pivot point in 2027, when inferencing will overtake training as the dominant driver of AI infrastructure investment.2 That favors distributed operators. Low-latency edge sites, regulated data handling, and national trust positions make telcos natural providers of sovereign AI services for government, healthcare, finance, and critical industries. The strategic task is to expand data center and edge footprints, strike credible co-location alliances, and build clear service catalogs with defined SLAs.
Low Earth Orbit Partnerships Are Strategic, Not Adjacent
Starlink holds an early lead as a preferred partner. Use cases are regional: direct-to-device augmentation and rural broadband in North America; resilience and disaster recovery in Europe; backhaul and transport in the Asia Pacific. The boundary is dissolving into a hybrid connectivity model where terrestrial, LEO, and high-altitude platforms interoperate. Operators that translate this into simple, reliable coverage commitments will set a bar that pure terrestrial challengers cannot meet.
Balance, Pivot, Revolt: The 2026 Imperative
Balance defines this year’s internal posture
According to IDC’s C-Suite Tech Survey (September 2025, 45 telecom respondents), 52% of telco C-suite leaders have AI implementation as a top-three priority. 50% simultaneously have technology modernization in the same tier. Transformation budgets are finite. Operators must prove each dollar in IT either reduces unit cost or enables monetization.
IDC forecasts a modest 5.2% growth in spend on operations and monetization systems in 2026, reaching approximately $54 billion, as telcos invest in the IT systems needed to monetize the billions in CAPEX already deployed in new wireless and fixed networks. AI returns collapse without data quality, inventory integrity, and clean integration patterns.
The autonomous network ambition illustrates the gap starkly. TM Forum data from 2025 shows that only 4% of operators self-report achieving Level 4 autonomous network status, yet 85% aspire to reach that level by 2030. There is an extraordinary gap between declared intent and current capability. The blockers are not new: fragmented inventories, interoperability gaps, and no single source of truth for network and service data. These are the same barriers that slow AI adoption more broadly because they are the same problem.
Pivot is now about ruthless sequencing
Winning operators are standardizing data models and cleaning network inventories as preconditions for AI-driven assurance and fault management. In EMEA, service assurance and fault management rank among the top automation priorities for 2026. RAN, IP access, and core domains are moving to Level 4 capability, one domain at a time. That sequenced approach is more realistic than a blank-check quest for full autonomy.
Talent is the other pivot dimension:
97% of telcos recognize material gaps in AI development and adoption skills, and while
65% are investing in AI-enabled learning tools, and 58% expanding internal upskilling programs, only
42% currently offer skills training. There is a meaningful gap between recognition and action.
A candid example from the UK illustrates the problem
A subscriber paying £15 per month for 10 GB, consistently using 6 GB with predictable Disney Plus and roaming needs, was offered a device upgrade and 30 GB for £18, or unlimited for £24, at renewal. No personalization. No complementary service upsell. The same operator offered 40 GB for £7.50 on a comparison site. That is a 50% ARPU cut and a fourfold value giveaway. The issue is not one offer. It is a commercial model built on customer inertia. AI agents are already automating plan comparisons and renewals for consumers and, soon, for enterprises. The purchasing friction that protected mediocre offers is eroding. Operators that cannot present clear, personalized value in real time will see churn accelerate in ways that overwhelm traditional retention playbooks.
5G From Product to Platform, and the 6G Horizon
The latter half of the 5G cycle presents an opportunity if operators redefine 5G. With IDC forecasting a 2.0% annual growth rate in global mobile connections through 2029, total connections will exceed 9 billion soon. In saturated markets, the focus shifts to retention and value per connection, highlighting the importance of commercialization.
The larger opportunity is 5G as a platform. The first phase focused on speed and latency, while the next phase views 5G as a foundation for extended reality, drones, V2X, private 5G, and RedCap devices. These applications scale in tens of thousands, necessitating different economics and commercial models. The main barrier to enterprise adoption is integration complexity, cited by 46% of enterprises.
However, there’s significant interest:
74% are keen on network slicing,
49% plan to boost fixed wireless access investment, and
58% are interested in satellite connectivity, despite misconceptions about satellite capabilities. Managing expectations is crucial.
Looking ahead, 6G commercialization is anticipated around 2029, with technical specifications still being developed by 3GPP. Key advancements include AI-driven control loops for deeper network autonomy, integrated sensing at radio sites, quantum-resistant security, and potential sub-terahertz spectrum. This evolution aims to transfer intelligence from data centers to the physical world, enabling real-time coordination among robots, vehicles, and infrastructure. The main challenge for operators is to present these capabilities in a way that developers and enterprises can easily implement without needing custom integration.
The Enterprise Connectivity Opportunity
Enterprise connectivity budgets are increasing, with many planning significant growth. Bandwidth demand has surged, especially among large organizations. Perception is a challenge. Cloud providers are favored over network service providers, which are often seen as difficult to work with.
Network as a Service (NaaS) could simplify connectivity by offering on-demand services, but concerns about lock-in and pricing remain. Clear economics and transparent SLAs are vital for adoption.
What Operators Should Do Now
1. Treat Inferencing As the Design Point
Re-architect edge sites, transport, and caching policies for high-volume, low-latency inferencing workloads. Prioritize power density, thermal design, and GPU scheduling at the edge. Package these capabilities into sovereign AI services with explicit data residency guarantees.
2. Fix the Inventory
Establish a single source of truth for network and service inventories. Converge data models, remove duplicates, and automate reconciliation. Tie funding for AI use cases to measurable progress on data quality. No clean inventory, no budget.
3. Sequence Autonomy by Domain
Build Level 4 capabilities where value is immediate. Start with assurance and fault management in RAN, IP, and core. Publish before-and-after KPIs: trouble ticket volume, mean time to resolve, and power consumption per site.
4. Build Commercial Intelligence, Not Just Network Intelligence
Replace blanket offers with model-driven personalization across renewals and upsell. Expose coverage, resilience, and application-aware QoS in a language procurement teams understand. Once AI agents shop on behalf of customers, opaque offers will lose by default.
5. Reframe the Enterprise Conversation
Lead with outcomes and integration. Present NaaS, slicing, and satellite augmentation as tools in a solution, not products looking for a problem. Pair telco architects with industry specialists from systems integrators in presales. Win rate follows credibility.
The Bottom Line
Barcelona confirmed that telcos hold the ingredients for a durable advantage precisely when AI inferencing, sovereignty requirements, and hybrid coverage demands converge. The constraints are equally real. Legacy complexity still suppresses AI returns. Automation ambitions outpace organizational readiness. Commercial models leak value. Enterprise buyers still question telcos’ breadth and ease of engagement.
The operators that close the gap will not treat these as parallel workstreams. Architecture decisions, data governance, talent development, product design, and go-to-market positioning are one integrated system. If inferencing is the design point, edge investment, inventory hygiene, domain-level autonomy, and commercial personalization are not separate initiatives. They are the same plan expressed across different functions.
The more consequential strategic question is whether telcos claim the sovereign AI layer before cloud hyperscalers establish it by default. Hyperscalers lack the physical edge, the regulatory standing, and the last-mile relationship. Telcos have all three and remain poorly positioned in enterprise perception. That gap is the real risk. Operators that close it through clearer positioning, credible NaaS execution, and published performance data will occupy a structurally defensible position. Those that treat sovereign AI as a marketing narrative without the service catalog to match will cede the ground to partners who move faster.
