During the gold rush, fortune seekers chased riches while suppliers profited by selling the tools needed to find them. The AI Supercycle presents a similar opportunity. As operators address new operational challenges, they must also identify the enabling capabilities—the modern picks and shovels—that can monetize high-value connectivity and establish their role in the AI ecosystem.

How is the rise of AI traffic the new gold rush?

We are moving from an era dominated by voice, data and video to one in which AI traffic increasingly becomes the predominant workload. The profound impact and exponential adoption demonstrate the transformative power of AI. As a result, many technology companies are making large investments to develop AI applications and the critical infrastructure required to support them. This modern tech rush mirrors the high-stakes gold fever from the late 19th century, which lasted about seven years, but compressed into a much tighter timeframe.

Why are new picks and shovels needed to build an intelligent optical network?

In this AI-driven landscape, the optical network is no longer a passive conduit. It is an active innovation enabler that can create new economic value. Training and real-time inference workloads have distinct data-transfer requirements: they are highly bursty, demand ultra-low latency, low packet loss and require deterministic performance for reliable model operation (see Table1).

Training

Inference

High bandwidth

Low latency

Minimal packet loss

High reliability

Table 1 - Training and inferencing attributes

AI efficiency, reliability and performance depend directly on network quality, as explained by the following factors:

The cost of idle GPUs: High-performance GPUs are indispensable AI assets, but they’re expensive. If data cannot move efficiently between processing stages or distributed datacenters, these resources sit idle and erode infrastructure ROI. Ultra-low-latency, highly reliable, high-throughput and intent-aware optical networks are, therefore, essential to keep compute resources in use, secure and operating at peak performance. The imperative of distributed datacenters: Large centralized datacenters require extensive real estate, capital and energy. Combined with the need for lower application latency and localized processing, these pressures are driving investment toward smaller, distributed AI compute resources. Reusing established facilities and connecting them with high-performance optical infrastructure becomes critical.

The cost of idle GPUs: High-performance GPUs are indispensable AI assets, but they’re expensive. If data cannot move efficiently between processing stages or distributed datacenters, these resources sit idle and erode infrastructure ROI. Ultra-low-latency, highly reliable, high-throughput and intent-aware optical networks are, therefore, essential to keep compute resources in use, secure and operating at peak performance.
The imperative of distributed datacenters: Large centralized datacenters require extensive real estate, capital and energy. Combined with the need for lower application latency and localized processing, these pressures are driving investment toward smaller, distributed AI compute resources. Reusing established facilities and connecting them with high-performance optical infrastructure becomes critical.

What type of picks and shovels can telecom providers offer from their infrastructure?

This shift in AI network demand creates significant opportunities for telecom providers to monetize existing and evolving infrastructure. Nokia WaveSuite AI helps service providers expand their portfolios and become indispensable solution partners to their end-subscribers in the following ways:

AI-optimized connectivity: Providers can move beyond generic bandwidth by offering premium, SLA-based services tailored to AI workloads, including dedicated high-bandwidth, low-latency and resilient optical paths. These services address subscribers that cannot allow network bottlenecks to disrupt AI operations. Edge AI hosting: Using distributed points of presence, providers can host inference resources and smaller AI models closer to users or enterprises who are beginning to adopt them. This reduces latency for applications such as industrial automation and smart city services while limiting reliance on costly centralized datacenter expansion. Network-as-a-service (NaaS) for AI: Nokia WaveSuite automation capabilities abstract network complexity and enable on-demand, programmable spectrum or wavelength services across tenant-based virtual network domains. Enterprise IT teams, AI developers and OSS/BSS systems can use APIs to request and scale connectivity dynamically, accelerating premium AI service deployment and creating new revenue streams.

AI-optimized connectivity: Providers can move beyond generic bandwidth by offering premium, SLA-based services tailored to AI workloads, including dedicated high-bandwidth, low-latency and resilient optical paths. These services address
Edge AI hosting: Using distributed points of presence, providers can host inference resources and smaller AI models closer to users or enterprises who are beginning to adopt them. This reduces latency for applications such as industrial automation and smart city services while limiting reliance on costly centralized datacenter expansion.
Network-as-a-service (NaaS) for AI: Nokia WaveSuite automation capabilities abstract network complexity and enable on-demand, programmable spectrum or wavelength services across tenant-based virtual network domains. Enterprise IT teams, AI developers and OSS/BSS systems can use APIs to request and scale connectivity dynamically, accelerating premium AI service deployment and creating new revenue streams.

Why do these new picks and shovels provide differentiation?

Telecom providers need differentiation beyond speed and bandwidth. Although both remain foundational, the true differentiators will be the experience and insights delivered by the network. Nokia’s optical automation solutions, WaveSuite and Transcend, provide a set of enabling capabilities that can simplify the operational lifecycle of AI-driven services either on shared infrastructure networks or on dedicated network buildouts such as managed optical fiber networks (MOFNs):

Simple order creation: WaveSuite monetization capabilities offer self-service portals and APIs that make ordering complex AI connectivity more like requesting a cloud instance. Service catalog specifications and intent-based requests reduce the need for deep networking expertise and streamline fulfillment. Fast service delivery: WaveSuite monetization capabilities can reduce delivery times from weeks to minutes. Rapid activation allows AI models to be deployed, trained, tested and scaled without waiting for manual network provisioning. Flexible service types: AI applications have different performance, resilience and security requirements. WaveSuite monetization service catalogs enable providers to offer choices matched to each subscriber’s requirements, including: Restorable services for basic recovery and less critical AI tasks, such as data retrieval prompting, versus agent-to-agent automation critical for physical AI with strict SLAs. Protected services with automatic failover and sub-50 ms recovery for mission-critical inference. Diverse routing that separates primary and backup paths for resilience against localized failures between router endpoints. Encrypted services that mitigate end-to-end data security risks for sensitive or proprietary information. Domain resource partitioning: Nokia’s optical automation capabilities support the logical partitioning of network resources for tenanted end-subscribers. Role-based access provides each authenticated subscriber with an isolated view of their network domain, including topology, optical routes, performance KPIs and fault information required to validate SLA compliance. Intelligent resource management: WaveSuite monetization capabilities optimize network asset allocation and utilization through: Order scheduling to pre-book capacity for events such as large-scale LLM training or planned data migrations from training datacenters to inferencing datacenters. Resource reservation to secure specific optical paths or bandwidth temporarily, ensuring critical AI operations perform without contention. Together, these capabilities improve predictability, prevent resource conflicts and reduce operational risk. Pay-per-use allows network operators to offer tailored data rate capacity for each AI application, rather than one-size-fit all.

Simple order creation: WaveSuite monetization capabilities offer self-service portals and APIs that make ordering complex AI connectivity more like requesting a cloud instance. Service catalog specifications and intent-based requests reduce the need for deep networking expertise and streamline fulfillment.
Fast service delivery: WaveSuite monetization capabilities can reduce delivery times from weeks to minutes. Rapid activation allows AI models to be deployed, trained, tested and scaled without waiting for manual network provisioning.
Flexible service types: AI applications have different performance, resilience and security requirements. WaveSuite monetization service catalogs enable providers to offer choices matched to each subscriber’s requirements, including: Restorable services for basic recovery and less critical AI tasks, such as data retrieval prompting, versus agent-to-agent automation critical for physical AI with strict SLAs. Protected services with automatic failover and sub-50 ms recovery for mission-critical inference. Diverse routing that separates primary and backup paths for resilience against localized failures between router endpoints. Encrypted services that mitigate end-to-end data security risks for sensitive or proprietary information.
Restorable services for basic recovery and less critical AI tasks, such as data retrieval prompting, versus agent-to-agent automation critical for physical AI with strict SLAs.
Protected services with automatic failover and sub-50 ms recovery for mission-critical inference.
Diverse routing that separates primary and backup paths for resilience against localized failures between router endpoints.
Encrypted services that mitigate end-to-end data security risks for sensitive or proprietary information.
Domain resource partitioning: Nokia’s optical automation capabilities support the logical partitioning of network resources for tenanted end-subscribers. Role-based access provides each authenticated
Intelligent resource management: WaveSuite monetization capabilities optimize network asset allocation and utilization through: Order scheduling to pre-book capacity for events such as large-scale LLM training or planned data migrations from training datacenters to inferencing datacenters. Resource reservation to secure specific optical paths or bandwidth temporarily, ensuring critical AI operations perform without contention. Together, these capabilities improve predictability, prevent resource conflicts and reduce operational risk. Pay-per-use allows network operators to offer tailored data rate capacity for each AI application, rather than one-size-fit all.
Order scheduling to pre-book capacity for events such as large-scale LLM training or planned data migrations from training datacenters to inferencing datacenters.
Resource reservation to secure specific optical paths or bandwidth temporarily, ensuring critical AI operations perform without contention. Together, these capabilities improve predictability, prevent resource conflicts and reduce operational risk.
Pay-per-use allows network operators to offer tailored data rate capacity for each AI application, rather than one-size-fit all.

There is no distributed AI without the optical network’s connectivity fabric. By using Nokia’s best-in-class automation capabilities, telecom providers can move beyond supplying capacity to become indispensable, high-value partners in the AI era.

About Pino Dicorato

Pino leads Solution Marketing for Nokia’s Network Infrastructure optical network automation environment portfolio, which is paving the path to transition, scale, simplify and digitize the future for the optical transport industry. With over 20 years of experience in the telecom industry, his roles have ranged from business development and consulting within various industry markets to R&D roles in photonic hardware development, ASIC/FPGA design, and software automation for high-scale manufacturing. Away from work, Pino enjoys high-intensity interval training, playing soccer, biking and traveling worldwide with his beautiful family.

Connect with Pino on LinkedIn

About Matias Schneeberger

Matias Schneeberger works as Product Line Manager for the Optical Network Division in Nokia. He has 15 years of experience working in the telecommunication industry in different roles from Network Integration, Professional Services and Product Management.His main goal is to build Software products and solutions that help Communication Service Providers to address their operational challenges. His current areas of focus are network optimization, open network management, and advanced monetization capabilities. Matias holds a B.Sc. in Electrical Engineering from Universidad de la Republica, Uruguay, and an Executive MBA from Politecnico di Milano, Italy.

Connect with Matias on Linkedin