Survey: 59% of Enterprises Cite Bandwidth Issues for AI Workloads
Rising latency and bandwidth constraints are hindering AI adoption as organizations struggle to move data at machine speed.
Key points
- Bandwidth issues reported by organizations rose to 59% from 43% last year.
- Latency challenges nearly doubled, increasing from 32% to 53% annually.
- AI agents generate up to 450% more network traffic than human tasks.
Recent research indicates that nearly six in ten organizations now report bandwidth issues related to artificial intelligence operations. This figure represents a significant increase from the 43% reported just one year ago, according to Network World. The data highlights a growing infrastructure bottleneck that is complicating enterprise AI strategies.
Latency challenges have also surged, rising from 32% to 53% over the same twelve-month period. Enterprises are experiencing performance gaps that impact user experience and lead to unnecessary spending on the wrong parts of the technology stack. Leaders often blame AI models for workflow failures, but the root cause frequently lies in underestimated connectivity requirements.
The Data Movement Bottleneck
Enterprises have invested billions in AI models, compute power, licenses, and GPUs over the past two years. Despite this massive capital expenditure, many organizations cannot reliably access the systems they have built. Network World reports that leaders consistently underestimate the connectivity needs beneath their AI strategy. This oversight creates a disconnect between model capability and actual operational performance.
The dialogue around AI has shifted from questioning the technology's validity to focusing on making it work at scale. However, a model is only as effective as the infrastructure connecting it to users, applications, and other models. The current infrastructure struggles to support the distributed, bursty, and concurrent traffic patterns generated by modern AI agents.
Impact on Network Traffic Patterns
Traditional enterprise networks were designed for predictable human transactions, such as a user searching for information and receiving a result. AI workloads break this established pattern entirely. According to the report, a typical AI agent can generate up to 450% more total network traffic per task than a human performing equivalent work. This volume and type of traffic differ fundamentally from human-driven network usage.
The traffic generated by AI is distributed, bursty, and occurs at machine speed rather than human speed. A successful AI pilot may prove the model works, but it does not guarantee the enterprise network can sustain it. Engineers at Cisco Live earlier this year observed these shifts in real time, noting traffic spikes that diverged sharply from the human traffic patterns their infrastructure was designed to handle.
What should you do?
Organizations should audit their current network capacity against projected AI workload demands. It is critical to distinguish between compute limitations and data movement constraints. Monitoring tools should be configured to detect bursty, machine-speed traffic patterns that differ from standard human interaction logs.
What to do and how to stay safe: AI infrastructure
- Review network bandwidth utilization metrics to identify current bottlenecks in AI data transfer.
- Evaluate latency reports to determine if delays stem from model processing or network connectivity.
- Monitor traffic patterns for bursty, concurrent spikes characteristic of agentic AI workloads.
- Assess infrastructure readiness for machine-speed data movement before scaling AI deployments.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Why are AI workflows collapsing despite massive investment in models?
Leaders often underestimate the connectivity requirements, leading to bandwidth and latency issues that hinder access to built systems.
How much more network traffic do AI agents generate compared to humans?
A typical AI agent can generate up to 450% more total network traffic per task than a human performing equivalent work.
How have bandwidth issues changed over the last year?
Organizations reporting bandwidth issues increased from 43% to 59% over the past year, according to recent research.
