In the ever-evolving landscape of cloud and communications infrastructure, one trend stands out among the rest: AI. As we venture into 2024, AI is set to have a profound impact on how networking infrastructure is built to support AI-enabled applications.
Enabled by the advent of AI, there are unique characteristics that set it apart from traditional cloud infrastructure. The training of large language models (LLMs) and other AI applications necessitates exceptionally low latency and high bandwidth. This demand for speed and connectivity is driving the need for highly distributed and accelerated platforms.
Generative AI (GenAI) is a particular subset of AI that creates text, images, sounds, and other outputs based on natural language queries. This emerging field is propelling computing trends towards more complex and powerful infrastructures. These infrastructures need to address the full spectrum of functionality, ranging from chips to specialized networking cards to distributed high-performance computing systems.
Leaders in the networking industry, such as Cisco, have recognized the significance of AI in networking and have actively embraced it in their marketing materials and investor conference calls. However, two companies that have attracted substantial investor interest are Nvidia and Arista Networks.
Nvidia, renowned for its comprehensive stack of networking elements, including the BlueField networking platform, has become a prominent player in the AI networking space. Arista Networks, on the other hand, has garnered significant attention from investors due to its status as a key networking supplier to AI giants like Microsoft.
While Nvidia and Arista Networks lead the industry, numerous private companies are making their mark in this market as well. Let's take a closer look at some of these up-and-coming players:
1. Arrcus: Arrcus offers Arrcus Connected Edge for AI (ACE-AI), a solution that utilizes Ethernet to support AI and machine learning (ML) workloads. ACE-AI targets communications service providers, enterprises, and hyperscalers who seek flexible networking solutions for their AI infrastructure.
2. DriveNets: DriveNets provides a Network Cloud-AI solution that employs a Distributed Disaggregated Chassis (DDC) approach to interconnect GPUs in AI clusters via Ethernet. This scalable platform serves as an alternative to InfiniBand, demonstrating improved job completion times in AI training clusters.
3. Enfabrica: Enfabrica, a startup founded in 2020, has developed an accelerated compute fabric switch (ACF-S) that enhances connections between network elements and AI systems. By streamlining AI processing and reducing latency, Enfabrica aims to lower the total cost of ownership for AI systems.
As the AI revolution unfolds, the role of InfiniBand, a high-bandwidth technology frequently employed in AI systems, is being actively debated. While Nvidia has established itself as a leader in InfiniBand, it has also diversified its offerings by developing Ethernet-based solutions. Ethernet holds economic advantages, but it requires software adaptations, coupled with SmartNICs and data processing units (DPUs). The emergence of Ethernet-based networking solutions presents an opportunity for upcoming companies to make their mark in the market.
Automation is another aspect of AI that is transforming the infrastructure landscape. Observability, the process of gathering and analyzing IT systems data, is being revolutionized by AI-driven automation. Companies like Kentik and Selector utilize AI and machine learning to monitor network traffic, correlating various data sources for capacity planning, cloud cost management, and troubleshooting.
WebAssembly (Wasm), an abstraction layer that streamlines application deployment in the cloud, is also proving beneficial for AI infrastructure. Companies like Fermyon are leveraging Wasm to compile different types of code efficiently, resulting in faster startup times for web applications.
The impact of AI on networking extends to multicloud environments as well. As more data is shuttled between clouds for collection, organization, and analysis, secure connectivity becomes paramount. Networking companies focused on edge data and application processing are expected to thrive in this scenario. Aviatrix, for instance, offers a multicloud networking platform that integrates connectivity with public cloud platforms, enhancing security and observability features.
Automation also plays a crucial role in integrating multidomain, hybrid, and multicloud environments. Companies like Itential facilitate infrastructure orchestration using APIs and pre-built automations, enabling organizations to establish flexible connectivity to data sources.
Furthermore, AI has paved the way for providers offering AI in IT and cloud environments. ClearBlade, for example, specializes in IoT software that facilitates stream processing and creates digital twins of IoT environments for real-time monitoring and operations. These advancements empower organizations with actionable insights and improved performance.
As we delve further into 2024, it is evident that AI's influence on networking and infrastructure will continue to thrive. While the hype surrounding AI may eventually taper off, the long-term capital spending plans for AI infrastructure deployments indicate that its impact on networking and infrastructure will be substantial. The future of networking and infrastructure looks to be tightly interwoven with the advancements and innovations revolving around AI.
