During HUAWEI CONNECT 2023, Yang Chaobin, Director of the Board and President of ICT Products & Solutions of Huawei, delivered a keynote titled “Embracing AI: Building Digital Infrastructure for an Intelligent World”. Yang believes that building digital infrastructure is vital in the AI era which features a vast range of models and applications. During the keynote, Yang said: “Huawei uses solutions for power and buses to help industries go intelligent. High computing, storage, and transmission power delivered by AI clusters can accelerate the development of AI foundation models. Furthermore, multi-scenario AI computing power and intelligent buses for industries can make ubiquitous intelligence a reality.”
Yang claimed that intelligence is becoming a key driver for both productivity improvement and economic growth. He also believes that AI evolution will reflect two trends: First, growing numbers of AI model parameters are leading to a significant increase in demand for computing, data storage, and network resources, meaning AI cluster infrastructure that offers higher computing, storage, and transmission power is required. Second, the adoption of AI across more industries means it is important to develop multi-scenario AI computing power and build intelligent connectivity buses for four key industry scenarios: WAN, campus, industrial manufacturing, and micro enterprise.
Launching high-computing-, storage-, and transmission-power solutions that facilitate leading AI foundation model training clusters
In terms of computing power, Huawei has introduced architectural and system innovations to support high-computing-power platforms for different scenarios, thus overcoming computing bottlenecks for AI foundation model training.
In terms of storage power, Huawei launched OceanStor A800, a leading AI knowledge repository storage system that features high-performance data storage based on an innovative architecture.
Regarding transmission power, Huawei launched the industry’s first high-connectivity AI computing DCN switch series, CloudEngine XH, and the industry’s highest-capacity ultra-broadband solution, Intelligent DC OptiX. These large-scale, high-capacity network solutions will deliver the transmission power necessary to unlock higher computing power.
Computing-network collaboration: Adopting foundation models across industries to realize ubiquitous intelligence
The adoption of foundation models across industries to help them go intelligent faster will require both computing power for AI training and inference, covering multiple application scenarios, and intelligent buses that connect numerous industries.
Huawei’s all-scenario AI computing products, including Atlas cluster series, training & inference servers and appliances, and AI accelerator modules, that facilitate AI deployment in diverse service scenarios can support centralized, at-the-edge, and on-device model training and inference.
Huawei also introduced an innovative intelligent connectivity bus solution that realizes intelligent connectivity for AI adoption across industry scenarios and has already been applied to typical industry scenarios like WAN, campuses, industrial manufacturing, and micro enterprises.
In WAN scenarios, lightweight and design-free intelligent SE-OTN products can facilitate the end-to-end ultra-broadband lossless connection and the quick deployment of intelligent WAN buses.
In campus scenarios, Huawei’s 10 Gbps intelligent campus solution can provide industrial campus customers with high-quality campus network experiences.
In industrial scenarios, the TSN switches that feature intelligent orchestration and low, deterministic latencies can support the agile deployment of both IT and OT services. This product can ensure both lossless data collection and ubiquitous intelligent analysis.
In micro-enterprise scenarios, Huawei’s one-stop, smart micro-enterprise suite improves small and micro enterprises’ network access experiences and enables them to adopt cloud-based AI applications that facilitate smart office and productions.
Working with industry partners to develop industry-specific AI platforms and foster a thriving AI ecosystem
Huawei focuses on innovation in foundational technologies, and adheres to a computing strategy of open hardware, open-source software, partner empowerment, and talent development. To date, we have worked with more than 30 hardware partners and over 1,300 software partners to deliver over 2,600 scenario-based AI solutions and develop more than 50 foundation models that have already been applied to a range of industry scenarios.
During the keynote, Huawei and a number of foundation model partners jointly launched a range of foundation model training and inference appliances for different industries and scenarios. These appliances are expected to accelerate the deployment and application of foundation models across industries.
