Huawei AI WAN: Building an AI-Centric IP Bearer Network to Drive New B2B Growth

Foundation models and AI agents are transitioning from experimental phases to widespread deployment, becoming the core engine driving industrial transformation. With global token productivity soaring, the token economy is now in full swing. According to IDC, AI agents will handle over 400 trillion tasks annually by 2030, ushering in a massive leap in industrial growth.

Amid this transformation, carriers are reshaping their identities, evolving from traditional network providers into key enablers of industry-wide intelligence. Networks now transcend mere connectivity. They serve as critical platforms for computing power scheduling, security, and intelligent operations, fostering cloud-network-edge-device-agent synergy and delivering integrated digital and intelligent services across industries. In response to this trend, AI WAN — the next-generation intelligent IP WAN built around AI — serves as the core enabler for carriers to break through computational limits, build a strong security foundation, and improve O&M efficiency.

Multi-dimensional awareness: Breaking compute limits, unlocking a new chapter in intelligent computing cloud services

Computing power now drives enterprise digital and intelligent transformation as a core engine. Cloud computing leasing is becoming the preferred choice. A bank in Anhui, China, struggled to deploy AI-driven anti-fraud and credit risk assessment services due to inadequate power supplies in its data center. A six-month reconstruction timeline slowed down computing power scaling. Moreover, the bank feared that the swift evolution of computing hardware could render its investments obsolete almost instantly — the "built to be outdated" risk. Gartner predicts that by 2028, 90% of Chinese enterprises will use leased cloud computing to agilely address the evolving needs of AI services.

To seize this strategic opportunity, global carriers are accelerating their transformation into "industry digital and intelligent partners". More than 30 mainstream carriers, including China Mobile, China Telecom, China Unicom, SKT, SingTel, DT, and Orange, have either launched or are planning to launch intelligent computing cloud services. By the end of 2025, the top three carriers in China had built over 180 EFLOPS of intelligent computing power. Through ecosystem collaboration, they have aggregated nearly 100 proprietary foundation models, over 100 third-party AI models, and hundreds of AI agents, offering enterprises full-stack services from computing power scheduling to AI agent invocation.

Currently, intelligent computing cloud services face two major hurdles in commercial use: First, traditional WANs struggle to support fine-grained scheduling and deterministic quality assurance, resulting in significant computational efficiency losses during long-distance transmission. Computational efficiency loss can reach up to 74% over 200 kilometers. Second, conventional centralized cloud-based training and inference methods fall short of strict compliance requirements in security-demanding sectors like finance and government, where sensitive data cannot leave the premises.

To address these pain points, Huawei AI WAN has made key technological breakthroughs in computing-network synergy, as shown in Figure 1.

Figure 1: Xinghe Intelligent WAN Solution

With multi-dimensional awareness and the proprietary Xingluo lossless algorithm, the solution tackles the load balancing and lossless transmission challenges of remote direct memory access (RDMA) — a mainstream communication technology for AI training and inference — achieving zero packet loss and computing efficiency loss within 5% over thousands of kilometers. By integrating the Xingluo vector engine, Huawei AI WAN elevates conventional data transmission to high-dimensional vector transmission, ensuring sensitive data remains within the enterprise for full compliance in high-security scenarios.

In one of the AI WAN projects, Huawei collaborated with China Telecom Shanghai to seamlessly connect customer-owned computing power with leased cloud resources across 110 kilometers, cutting model training and inference losses to under 3% while keeping critical data safely within the customer's premises. This innovation not only enhances carriers' network capabilities, but also builds a competitive edge of computing-network synergy, positioning carriers to lead in the intelligent world.

Resilient security: Building a strong security foundation and pioneering a new security-as-a-service paradigm

The rise of AI and quantum computing has brought unprecedented network security challenges. AI has dramatically reduced the barrier to launching attacks, amplifying their impact and acting as a threat multiplier. IBM's security report showed a 72% increase in AI-assisted attacks in 2025. Enterprises now face a critical shortage of security talent and defense capabilities, driving demand for third-party managed security services. In China, for example, more than 60% of large and medium-sized firms plan to adopt these services. Leveraging their ubiquitous network coverage, localized operations, and enterprise channels, carriers are well-positioned to redefine the security service market and open new pathways for growth.

Huawei AI WAN provides secure routers that help carriers build a multi-dimensional resilient security foundation that covers devices, links, and networks. This foundation provides end-to-end protection against external threats and ensures 24/7 service availability. It consists of three dimensions: Device-level intrinsic security, link-level quantum security, and network-level threat prevention. Among them, the Huawei Intelligent Flash Defense Solution upgrades security operations from sample-based detection to full-scale identification and from manual response to real-time closed-loop responses. The solution detects attacks within 5 seconds and initiates defense in just 10 seconds. China Mobile used this solution to transform its security system from a cost center into a value engine, unlocking new market space for B2B value-added services.

Collaborating with carriers, Huawei developed an innovative architecture for the security resource pool, offering centralized AI modular firewalls to enable AI-powered online detection of unknown threats. This improves the unknown threat detection rate from the industry average of 80% to 95%, creating high-value security private lines.

Quantum computing breakthroughs threaten to disrupt traditional cryptography. Tasks requiring tens of thousands of years on conventional computers can be accomplished in just 10 minutes on quantum machines. Harvest Now, Decrypt Later (HNDL) attacks are now a real-world risk. Global adoption of quantum security is accelerating, with over 90% of high-income countries having unveiled national quantum security strategies.

To this end, Huawei AI WAN integrates communication with quantum security, empowering carriers to rapidly launch quantum security services.

First, by upgrading to post-quantum cryptographic (PQC) algorithms at the data layer, carriers gain quantum protection with a simple software upgrade. In addition, Huawei AI WAN supports X-Sec multi-point dynamic deployment, improving the efficiency of provisioning quantum private lines by 60% and helping carriers capture the first-mover dividends of quantum security value-added services. Second, Huawei AI WAN introduces a board with native quantum key distribution (QKD), achieving integration of the QKD network with the existing communications network. Its unique noise suppression algorithm enables communication signals, key negotiation signals, and quantum signals to be transmitted over a single fiber, requiring no new devices or additional fiber installation and reducing overall investment by more than 60% (as shown in Figure 2).

Figure 2: Native QKD with communication-quantum security integration

Based on Huawei AI WAN, China Telecom Ningxia has launched the industry's first quantum security service package, achieving a 50% premium over traditional private lines and successfully opening up a new track for the commercial use of quantum private lines.

Network autonomy: Boosting O&M efficiency and setting a new benchmark for digital and intelligent O&M

The service level agreement (SLA) is at the core of carriers' B2B competitiveness, and high-quality SLAs depend on efficient and stable network O&M. As networks expand and services become more complex, traditional O&M methods are falling short. Frequent network changes increase the risk of human error. In the past two years alone, carriers globally have experienced 64 major incidents caused by routine O&M issues, resulting in over US$30 billion in economic losses. At the same time, cost pressures are mounting and the O&M workforce is shrinking, widening the gap between labor supply and O&M demand.

The industry now agrees that AI agents are essential for achieving automated, intelligent, and even hands-free O&M. A Google Cloud report reveals that 56% of carriers have scaled up their use of AI agents, with network O&M ranking as the second most common application scenario. TM Forum highlights that once AI agents are fully adopted, 70% of repetitive O&M tasks can be fully automated by machines, 20% of complex tasks can be AI-assisted with manual final review, and only 10% of extreme cases require deep manual intervention.

Huawei AI WAN, centering on network agents, uses AI routers with precise awareness to deliver round-the-clock intelligent protection across all domains, as shown in Figure 3.

Figure 3: Intelligent network O&M solution

First, configuration simulation and pre-verification prevent human errors from entering the network. Second, using the CMOS-Risk algorithm, the system continuously learns from 800,000 high-quality fault cases and live network scenarios. This allows the system to precisely reconstruct the network architecture, evaluate the impact of route changes in seconds, and cut the full diagnosis time for a network with 8,000 NEs from two months to just two hours, preemptively identifying and mitigating risks. Third, when a fault occurs, the system rapidly pinpoints the root cause and dispatches tickets accurately via AI agent collaboration. For hardware issues, it guides engineers to resolve them in a single onsite visit; for software issues, it triggers automatic self-healing.

Powered by iMaster NCE-IP's digital network map and NetMaster's network AI agents, Huawei AI WAN builds an end-to-end closed-loop system that covers from intent recognition to intelligent analysis, decision simulation, and automated execution. The solution helped MTN South Africa develop the world's first TM Forum-certified Autonomous Network (AN) Level 4 intelligent IP network, setting a new benchmark for network O&M.

AI has opened new growth opportunities for carriers. Huawei is ready to collaborate with global carriers to build AI-centric next-generation IP bearer networks — breaking computing limits, securing infrastructure, and boosting O&M efficiency — to unlock new B2B growth through digital and intelligent transformation.