Key Takeaways From Mobile World Congress 2026

Mobile World Congress took place from 2–5 March 2026 in Barcelona and was attended by over 10,000 people. I had a fantastic MWC26. Being part of the PAC and Idate team, we screened countless announcements and engaged with clients over five days, yielding a wide range of insights on the evolving telco and tech markets.

We met with over 50 clients to discuss cutting-edge technology trends, present our analyses, and debate perspectives. Overall, MWC26 had an energetic and optimistic atmosphere. My personal top ten takeaways are that:

• Rising costs for energy, hardware, chips, and tokens will put pressure on telcos. In discussions with operators, we heard of massive chip price increases occurring over very short periods, as the hyperscalers continue to buy up large portions of the market. At the same time, energy prices were visibly surging during the event, given the geopolitical turbulence. Looking ahead, token consumption is set to rise sharply as AI agents become more widespread, with agentic workloads potentially exceeding human-driven usage by hundreds or thousands of times, further intensifying cost pressure. I am wondering whether energy costs are the Achilles heel of telcos’ sovereign data center ambitions.
• Everyone is talking about agentic AI, but few deployments are operating at scale. In the telco space, AI agents are moving beyond chat. I loved the demo of Deutsche Telekom’s Magenta AI Call Assistant – even though privacy and data protection-related questions remain. It illustrates a futuristic model for AI-augmented telephony, enabling real-time translation, automated call summaries, and contextual assistance directly within the network layer. Magenta AI Call Assistant requires no dedicated apps, hardware, or complex configuration, and activation remains user-controlled. Moreover, telcos are also trialing agents for network operations, with agents monitoring conditions, automating maintenance, and improving service quality. However, my impression is that most initiatives remain pilots, with scaling constrained by issues like unresolved guardrails around autonomous agent behavior.
• A new operating model must emerge for business in an agentic AI environment. AI is shifting from a productivity tool to the core operating logic of the enterprise. Thus, organizations must rethink how work, decisions, and value creation are orchestrated. At MWC26, Infosys introduced the Frontier Telco framework to support this transition. The framework outlines transformation across operations, finance, talent management, go-to-market strategy, and product design. It positions the enterprise as a system that combines intelligent operations with customer value orchestration and strategic human oversight. I am impressed by the simplicity of the Frontier Telco framework. While designed for telecom operators, I believe that it is applicable across industries.
• AI is reshaping network architecture as usage patterns evolve. Already, around 60% of digital content in China is AI-generated. The use of LLMs and the rise of AI agents are pushing traffic toward a more balanced downlink/uplink pattern rather than the traditional asymmetry. This shift forces a rethink of UL/DL capacity, real-time transmission, higher reliability, and secure bidirectional data integrity. At the infrastructure level, the All-Photonics Network (APN) concept introduced by the IOWN Global Forum envisions an open photonic layer enabling distributed data centers to share GPU and storage resources at ultra-low latency, while potentially reducing energy consumption. APN could be a building block for data center redesign in Europe, with large data centers in remote locations with renewable energies, like Spain or the North Sea, and distributed data centers close to AI consumption areas.
• Sovereignty is emerging as a strong differentiator for telcos. Positioned at the intersection of geography, trusted infrastructure, networks, and data centers, telcos are natural providers of sovereign data, cloud, and AI services. Their long-standing experience navigating complex regulatory frameworks also positions them well to enforce the guardrails required for AI models and agents operating in sovereign environments. As PAC has explored extensively, enterprise-focused telco units, such as T-Systems, are gaining traction in this space. At the same time, “networking for AI” is becoming essential as evolving traffic patterns and fiber capacity constraints risk limiting large-scale AI workloads.
• Specialized models create a new opportunity for telco-driven AI factories. A strategic tension is emerging between foundational model providers seeking to capture enterprise data within their ecosystems and data platform vendors aiming to remain the primary orchestration layer. In my experience, enterprises will demand both model-driven data access and native AI interfaces integrated with their data platforms. Telcos are well-positioned to develop smaller, domain-specific models and sovereign training platforms. Thus, initiatives like GSMA’s launch of Open Telco AI make a lot of sense. However, such initiatives depend on strong data governance and shared platforms, supported by consistent horizontal data models and clear guardrails for AI-driven network operations.
• Network APIs saw less hype than in previous years. While the discussion of network APIs themselves was less prominent than last year, greater attention was given to the use cases they enable. This suggests the market is moving from early hype toward more practical deployments. In my opinion, this shift also indicates a gradual maturation of the network API ecosystem. However, clear examples of direct API monetization remained scarce at MWC26. As PAC has highlighted in our API research, the larger opportunity is likely to come from ecosystem expansion and indirect monetization rather than standalone, token-based API revenue streams.
• AI is converging with virtual twin technology. We tested a solution from Dassault Systèmes designed for data center planning, though the underlying approach applies across many industries. The demonstration showed how far virtualization has advanced and how significantly it could reshape the way complex systems and products are designed, operated, and optimized. In an AI-driven environment, however, competitive advantage depends less on producing artifacts and more on deciding what to build, what to exclude, and when to stop. I believe that as AI automates execution, human value increasingly lies in strategic judgment, problem framing, and decision quality.
• 6G is gradually moving from a vague concept toward a clearer industry definition. 6G remains years away from mainstream adoption, with initial pilot deployments expected around 2029. However, MWC26 showcased more concrete visions from Ericsson, Nokia, Huawei, and operators such as T-Mobile. A common theme is that 6G will be AI-native and highly programmable, enabling new capabilities without requiring constant hardware replacement. As traffic volumes surge and uplink demand grows, cost per bit and energy per bit will become critical performance metrics. Over time, 6G will enable distributed intelligence spanning devices, edge, and cloud.
• The workforce impact of AI has slipped down the agenda. Despite extensive discussions about AI and agents, surprisingly little attention at MWC26 was given to the implications for talent and workforce transformation, with conversations involving Infosys and EY being notable exceptions. Technology reshapes job roles every few quarters. I believe that learning velocity is becoming more valuable, complementing academic credentials and certificates. AI fluency may increasingly differentiate graduates. At the same time, organizations must address “work slop”, polished but low-value AI-generated output, by deploying trustworthy, specialized agents that help employees produce more meaningful, high-quality outcomes.