- Verizon's AI Connect initiative is moving to take advantage of skyrocketing enterprise demand for network bandwidth and new network configurations driven by AI
- Customers seek prudent investments to be ready for increased capacity needs without getting too far ahead of demand
- Verizon also uses AI internally for customer experience and delivering AI in its products and services.
MOBILE WORLD CONGRESS, BARCELONA - AI is driving exploding enterprise demand for network bandwidth and new network configurations, and Verizon is on it, company executives said at an industry analyst roundtable this week.
Chipsets, data centers and power get all the attention in discussions of resources needed for AI, noted Scott Lawrence, SVP and chief product officer, Verizon Business. "What's missing from that conversation — a key ingredient in the recipe — is the role that the network will play in distributing AI workloads at scale," Lawrence said.
Building large language models (LLMs) requires centralized data, with demand for dark fiber, wave services and large bandwidth. Inferencing is becoming more commonplace, which is driving a transition to a more distributed computing environment where traffic moves device-to-device and edge-to-core-to-cloud, Lawrence said.
AI-driven network topology changes are the subject of our free report, published this week, "Redefining Connectivity for AI-Driven Enterprises: The unexpected evolution of SD-WAN." We talked with experts at VeloCloud by Broadcom, Walmart, Nature Fresh Farms, AT&T, and Orange Business.
In another free report, we explored how network topologies need to adapt to keep expensive GPUs fed with data: "AI and the Network: Optimizing Network Design and Operations to Meet AI Demands."
Verizon's AI Connect strategy and services, launched in January, are designed to serve that networking demand, Lawrence said.
Building on strengths
AI Connect is built on Verizon's key assets across the U.S., with space, power and cooling to help expand edge computing, based on edge services launched five years ago, Lawrence said.
Verizon also leverages its fiber assets across 71 major metropolitan U.S. cities, supporting dark fiber and lit wave fiber services. Verizon is rolling out 400 gigabit wave services this year, with 800 gigabit to follow shortly, and it tested 1.6 terabyte service in November.
Part of Verizon's strategy is serving the GPU-as-a-Service (GPUaaS) market, Lawrence said. "Many of [these providers] are infrastructure-disadvantaged because they haven't had time to build up their fiber networks like the hyperscalers." Verizon is partnering with Vultr to provide data center space, power, cooling and fiber networking facilities "to extend and connect their existing footprint of data centers," he said.
Platforms like Vultr have recently been dubbed "neocloud providers." Last week, Juniper launched a solution purpose built for this market.
Verizon is not alone in seeing and anticipating networking demand driven by AI. Long-haul and middle-mile players such as Lumen, Zayo and Windstream are investing in networking capacity to meet demand. Zayo plans to build 5,000 miles of long-haul fiber. Flexential CEO Chris Downie warned that network could be the cloud's next coveted resource.
What enterprises want
Verizon's enterprise customers want to invest in AI prudently, avoiding "getting caught up in the hype" and implementing "a solution that's looking for a problem," Daniel Lawson, Verizon Business SVP global solutions, said. They seek solutions to improve operations, change the customer experience and differentiate from competition.
For example, one big box retailer has a GPU in every store, using it to monitor shrinkage and other applications, including automated cashless checkout, Lawson said. He didn't name the retailer, but his description is very similar to technology discussed by Walmart in Fierce's report on redefining network connectivity. Another enterprise is looking to build applications to help train 15-30 new regional suppliers annually.
Early enterprise AI adopters tend to be "on the larger side," Lawson said, and in "asset-heavy industries" such as manufacturing, logistics, supply chain and healthcare, with enormous volumes of valuable data at the edge. These enterprises value new enterprise network architectures to bring "the model to the data instead of trying to bring the data to the model," he said.
Enterprises demand convergence. They don't care if connectivity is wireline or wireless; they just want it to work, and offer the same kinds of SLAs on wireless as on wireline, Lawrence said.
Also, enterprises want to lay the foundation for future applications without building ahead of demand.
"They're not wanting to build the church for Easter, so to speak," Lawrence said.
They want to have capacity available to move quickly and efficiently to solve problems without getting too far ahead of need.
Different lenses
Verizon views AI through different. The first is "applied AI," which enhances customer and employee experience and drives operational efficiencies. Verizon is applying generative AI and AIops to its own business and has been using AI for "well over a decade," resulting in "significant operational and overall experience improvements" by deploying AI and generative AI.
Verizon has deployed Fast Pass, a contact center solution using Google Cloud services across 30,000 agents, resulting in significant call time reduction and improved customer satisfaction, Lawrence said.
Another lens is "embedded AI," where AIops and generative AI are integrated into existing and new products and services.
Verizon plans to launch Verizon Digital Assistant for the small and medium business market this month. This assistant integrates conversational AI into SMS texting so customers can ask for information on business hours, venues and other such information.
Additionally, AI is embedded in devices Verizon sells from Apple and Samsung.