Huawei’s MWC 2026 announcements signal that agentic network infrastructure is no longer conceptual – the buildout has begun.Two emerging agent communication standards–one for enterprise, one for telecom–suggest the industry is converging on interoperability, not just automation.
Agentic networks are emerging more rapidly than most operators had anticipated. At Mobile World Congress 2026 in Barcelona, Huawei made two announcements on the table that, taken together, point to something larger than a product launch: the beginnings of an infrastructure layer designed specifically for a world where AI agents–not humans–are increasingly the entities making and managing network connections.
The first was the release of its Agentic Core solution, a three-engine network architecture intended to help operators handle the demands that AI agents will place on their infrastructure.
The second was the announcement of an open source project for A2A-T, the Agent-to-Agent for Telecom protocol–a standardised communication framework for multi-agent collaboration within telecom environments, co-developed with global industry partners and formalised through TM Forum in early February 2026.
On the surface, these are two separate announcements. But they are better understood as two pieces of the same problem: what does a network look like when the primary users of it are agents rather than people?\
The infrastructure problem no one has fully solved
Current network architectures were built for human-generated traffic patterns– predictable, relatively uniform, designed around consumer and enterprise use cases that have not changed dramatically in years. AI agents break those assumptions almost entirely.
Huawei’s Agentic Core framing is built around this tension. According to the company, as AI agents become embedded in next-generation devices–including embodied robots and autonomous vehicles–the number of connected entities is projected to grow tenfold.
An AI-powered industrial robot, per Huawei’s own figures, may require 100 Mbit/s bandwidth with 20 millisecond latency to function reliably. That is a specific, non-negotiable requirement that static, rule-based network architectures cannot accommodate at scale.
The response Huawei is proposing is a shift to what it describes as intent-driven networks–where AI agents inside the network itself interpret the requirements of different services and organisations, dynamically allocate resources, and handle policy generation and configuration automatically, in a closed loop. Rather than operators pre-configuring network behaviour for anticipated use cases, the network learns to respond to what is actually being demanded of it.
That is the vision. Delivering it is another matter, and it is where the protocol question becomes unavoidable.
Why agent interoperability is the harder problem
Parallel to Huawei’s telecom-specific work, Google launched its Agent2Agent (A2A) protocol in April 2025, initially with support from more than 50 technology partners, including Salesforce, SAP, ServiceNow, and PayPal. The protocol has since been donated to the Linux Foundation and now counts over 150 organisations in its ecosystem.
Its purpose: to create a common communication standard that lets AI agents built on different frameworks, by different vendors, discover each other’s capabilities and coordinate tasks without requiring custom integration code for every new pairing.
Google’s A2A and Huawei’s A2A-T are not the same thing, and they are not in direct competition. Google’s protocol is designed for the enterprise software layer–agents collaborating across business applications. A2A-T is designed for the network layer, addressing the reliability, security, and latency requirements that are specific to telecom environments.
As Huawei describes it, A2A-T is built to handle the kind of cross-domain, cross-vendor workflows that operators run in automated production–where collaboration efficiency, reliability under load, and security are non-negotiable.
But the broader dynamic is the same: without a shared protocol, every agent-to-agent interaction requires custom integration. That is not scalable in a world where the number of agents and the number of tasks they need to coordinate on is growing rapidly. The TM Forum’s formalisation of A2A-T as IG1453 in February 2026–and Huawei’s decision to open source the supporting software at MWC–is an attempt to prevent that fragmentation before it becomes entrenched.
From specification to deployment: why open source matters here
Telecom standards have a long history of being thorough on paper and slow in practice. The gap between a well-constructed specification and actual industry interoperability is where most standardisation efforts quietly stall. Huawei’s open source move for A2A-T is, in part, an acknowledgement of that history.
The project will include three components: an A2A-T Protocol SDK for standardised agent integration, a Registry Centre handling authentication, addressing, and skill management across multi-agent environments, and an Orchestration Centre with low-code and no-code visual workflow tools and pre-built solution packages. The explicit goal is to lower the barrier for operators and integrators who cannot or do not want to build from scratch–and to accelerate the transition from industry consensus to global deployment.
If adoption follows, the claim in Huawei’s own announcement is striking: system integration timelines that currently take months could compress to days. That is a significant efficiency claim, but it reflects a real inefficiency in the current environment. Integrating agent systems across vendors and domains, without a shared protocol, is precisely the kind of labour-intensive work that has historically made autonomous network operations difficult to scale.
What this means beyond Barcelona
For operators in the Asia Pacific region–where 5G rollouts are maturing but the commercial returns remain under pressure–agentic networks represent both an opportunity and a pressure point. The infrastructure investment required to support physical AI at scale, including the digital identity management, agent registration, and A2A session management that Huawei’s Agentic Core requires, is not trivial.
Neither is the organisational shift involved in moving from rule-based network management to intent-driven, AI-mediated operations. But the alternative–continuing to operate static network architectures while AI-generated traffic patterns grow increasingly unpredictable–is its own kind of risk.
The operators that benefit most from the agentic network transition will likely be those that engage with the protocol and infrastructure questions now, rather than waiting for the ecosystem to consolidate around them.
MWC 2026 will not be remembered as the moment agentic networks arrived. The harder work of implementation, standardisation, and real-world deployment is still ahead. But the groundwork being laid in Barcelona this week–a network architecture built for AI agents, and a shared protocol for how those agents communicate–is where that transition begins.
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.
Tech Wire Asia is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.