Agentic AI Is Reshaping Network and Enterprise Infrastructure: What ERP and Operations Leaders Should Track
ERP

Agentic AI Is Reshaping Network and Enterprise Infrastructure: What ERP and Operations Leaders Should Track

AT&T's CEO says agentic AI will drive roughly nine times more enterprise network traffic by 2035, a claim with direct implications for ERP integration architectures and cloud infrastructure budgets.

2 min readJuly 26, 2026
Back to News
Photo by Christina Morillo on Pexels
TL;DR
  • -AT&T CEO John Stankey projects agentic AI will drive roughly 9x enterprise network traffic growth by 2035, with AI agents already generating 450% more traffic per task than humans.
  • -For ERP and operations leaders, that traffic curve puts middleware, integration layers, and cloud contracts under near-term pressure.
  • -Audit your current network and cloud capacity assumptions against agentic workload growth before renewing infrastructure or vendor agreements.

The traffic surge no ERP team has fully priced in

Agentic AI is not just an application story. It is a network and infrastructure story.

AT&T CEO John Stankey said this week that AI agents generate up to 450% more total traffic per task than a human doing the same work. He projected roughly 9x growth in enterprise traffic and 7x growth in consumer traffic by 2035 driven by agentic adoption.

That is the AT&T perspective, not an independent forecast. ERP and operations leaders should treat it as a directional signal, not a confirmed number.

But the direction matters. If Stankey's framing holds even partially, integration middleware, API gateways, and cloud egress contracts sized for today's workloads will face stress earlier than most roadmaps assume.

Stankey added that distribution of AI inference to the edge requires low latency and high-bandwidth connectivity. He argued fiber-enabled network convergence creates competitive advantage at that edge. That is a carrier's commercial argument. However, the underlying infrastructure constraint it describes is real and vendor-neutral.

Cloud waste compounds the pressure

A separate data point from an Insight study reported in May reinforces the concern. The study found organizations waste an average of 24% of annual cloud capacity, and that AI is driving a 12% year-on-year increase in hosting costs. The study also found 67% of organizations already view digital sovereignty as a critical strategic priority.

Those three numbers sit uncomfortably together. Cloud budgets are already inflating. Waste is already high. Agentic workloads will push volume higher still.

For finance leaders reviewing cloud and infrastructure spend, the combination creates a near-term audit trigger, not a 2035 problem.

Sovereignty adds a vendor-selection wrinkle

Also in this week's digest: Samsung is rumored to be in talks to invest roughly €1 billion in French AI firm Mistral, as European organizations seek alternatives to US technology dependence.

Asish Patel of Houlihan Lokey's Capital Solutions Group described growing recognition that Europe has potential to become a more independent technology market. That is an investor's view, not confirmed deal terms.

For ERP teams evaluating AI-embedded modules from US hyperscalers, the sovereignty question is becoming a procurement and compliance consideration, particularly in regulated European industries. The source does not confirm Mistral has closed any deal. That detail remains unknown.

What the Randstad signal adds for HR-linked ERP

The digest's headline also flags a Randstad HR-and-AI combination as this week's positive example. The source does not provide detail on what Randstad deployed or what outcomes it achieved. That limits what can be said here. HR module owners should look for the full Randstad story separately before drawing conclusions.

The decision that changes

Metrotechs analysis: if you are renewing network, cloud, or integration contracts in the next 12 months, the Stankey traffic projections are a useful stress-test input. Model what a 3x to 5x increase in API and integration traffic would cost under your current agreements. The full 9x figure is a 2035 projection from an interested party. The near-term multiple is the one worth pricing now.

The cloud waste finding from the Insight study is the more immediately actionable point. Over-provisioning, limited visibility, and inactive resources each contributed to that 24% waste figure. Any of those three conditions in your current environment represents recoverable budget before agentic workloads arrive in volume.

Sources and supporting resources
Next
Lenovo Feeds Device Telemetry Into ServiceNow's AI Control Tower to Fix IT Problems Early

Get Business Technology Updates

News, insights, and practical guidance across ERP, Cloud, Data, AI, digital transformation, and technology projects.

No spam. Unsubscribe anytime.