How the AI Boom Is Shaping IPv4 Leasing and Sales

IPv4 leasing and sales are affected by the artificial intelligence (AI) boom mainly through the services built around computing capacity: hosted inference, dedicated customer endpoints, and connections to external systems. More graphics processing units (GPUs) do not automatically require more public Internet Protocol version 4 (IPv4) addresses. Additional address demand arises when those services need public connectivity that a provider’s existing pool cannot support.

For hosting providers and data centers, the next step is to count public endpoints and customer requirements before acquiring another block. For IPv4 holders, the opportunity is to serve that identifiable demand, with realistic expectations about utilization and pricing.

Table of Contents

• What does the evidence show about AI infrastructure growth?
• Where does an AI service actually need public IPv4?
• How does AI growth influence leasing and purchase decisions?
• A practical example: size the service, not the GPU cluster
• Which signals should trigger an IPv4 capacity review?
• What should IPv4 holders expect from AI-related demand?
• What should teams clarify about AI and IPv4 demand?
• Ready to source addresses for your AI services?

What does the evidence show about AI infrastructure growth?

The International Energy Agency (IEA) reports that global data-center electricity demand increased by 17% in 2025. Electricity consumption at AI-focused data centers increased by 50% in the same year. These figures, published in the IEA’s 2026 report on energy and AI, demonstrate the expansion of computing infrastructure.

They do not measure IPv4 purchases or leases. Our assessment is that this expansion creates additional opportunities for address demand where new services require public IPv4 connectivity. The sources cited here do not establish AI’s share of worldwide IPv4 transactions or a market-wide increase in prices caused by AI.

AI is also entering a market with an existing supply constraint. The RIPE Network Coordination Centre’s (RIPE NCC) June 2025 report on IPv4 and transfers explains that the last available IPv4 /8 blocks in the Internet Assigned Numbers Authority’s (IANA) original free pool were allocated to regional registries in 2011. Leasing and transfers redistribute existing address space; building more data centers does not replenish that pool.

Where does an AI service actually need public IPv4?

The useful distinction is between the internal compute network and the service’s internet-facing connections. NVIDIA’s DGX SuperPOD network-planning documentation explicitly distinguishes routable and non-routable address requirements. It is a concrete example of why internal network size cannot be treated as a public IPv4 requirement.

These are conditional demand mechanisms, not measured shares of the IPv4 market. An inference endpoint can also be private: Amazon SageMaker AI supports invocation through private network endpoints without sending that connection over the public internet. Providers should therefore count how customers access a service, not assume that every deployed model needs a public address. For the broader application patterns, see our guide to IPv4 leasing for hosted software platforms.

How does AI growth influence leasing and purchase decisions?

IPv4 leasing for AI infrastructure can suit a staged launch when the number of paying customers or the deployment duration remains uncertain. Purchase becomes worth evaluating when dedicated public addresses support a sustained production service and changing them would disrupt customer integrations. The deciding factor is the service commitment, not the size of the GPU order. Our IPv4 lease-versus-purchase comparison covers the wider commercial trade-offs.

Demand may also reach the IPv4 market indirectly. An AI business using managed hosting can consume addresses already held by its provider. The provider needs additional leased or purchased space only when its usable inventory no longer covers its requirements. More AI customers therefore need not produce an immediate, proportional increase in external IPv4 transactions.

Planning a launch or adding customer-facing capacity? Explore IPv4 leasing and purchase options on the InterLIR Global marketplace. IPv4 leasing starts at €99 per /24 per month, with rental terms from one month. Check the selected listing for availability and applicable conditions (activated within 24 hours).

A practical example: size the service, not the GPU cluster

Consider an illustrative GPU-hosting launch at one site. The operator has no spare public addresses available for this service. Its commercial offer requires two dedicated public addresses per customer; this is an example assumption, not a technical requirement for AI hosting.

A /24 contains 256 addresses in total. In this example, 208 planned addresses leave 48 before network-design reservations. That makes a /24 a candidate for evaluation, not a guarantee that every subnet layout will fit. Validate the usable capacity and routing design with the hosting network before ordering.

If a suitable /24 is available at €99 per month and that rate applies throughout a three-month rental, the address-rental subtotal is €297. This is an illustrative calculation using InterLIR Global’s starting rate, not a quote or a complete deployment budget. Confirm tax treatment and any additional charges; compute, transit, and other infrastructure costs are outside this calculation.

The useful sensitivity test is customer growth. With 120 customers receiving two addresses each, the same assumptions produce 288 addresses: 240 + 16 + 8 + 24. That exceeds one /24 even before network reservations. This is a service-driven reason to review capacity; adding GPUs behind existing endpoints would not necessarily have the same effect.

Which signals should trigger an IPv4 capacity review?

Use the service design, customer commitments, and current address inventory as inputs. Review these four signals before expanding an AI deployment:

1. Dedicated-address commitments: compare signed customer requirements with available allocations. Separate confirmed orders from sales forecasts.
2. New public locations: identify which launches need independent public endpoints and which can use existing frontends.
3. Outbound connection requirements: confirm whether customers actually require separate source addresses or can accept shared connectivity.
4. Observed consumption: reconcile the forecast with addresses assigned to active services and the reserve needed for the next launch stage.

Verify the result against the deployed allocation records and actual customer connectivity requirements. If the forecast changes, revise the address plan before extending a lease or committing to a purchase. This keeps procurement tied to operational demand rather than infrastructure announcements.

What should IPv4 holders expect from AI-related demand?

For address holders, an AI-related inquiry is a potential use case, not evidence that a block deserves a price premium. Ask what will use the addresses, how much capacity is committed, and how long the service is expected to run. Our guide to evaluating demand before monetizing IPv4 space explains the broader assessment.

Transfer counts also need careful interpretation. The RIPE NCC’s public IPv4 transfer tables record transferred blocks, organizations, dates, and transfer types, including business-structure changes. They do not report transaction prices or label the buyer’s AI workload. Those records alone cannot establish an AI-driven price increase.

For operators, the actionable opportunity is to secure the capacity their services require. For holders, it is to match available resources to that requirement. Neither decision needs a prediction that IPv4 prices will inevitably rise.

InterLIR Global offers rental terms from one month, with IPv4 addresses starting at €99 per /24 per month. Letters of Authorization (LOAs), Resource Public Key Infrastructure (RPKI) support, and reverse Domain Name System (rDNS) configuration are available (activated within 24 hours). Confirm the arrangements for the selected block against your deployment requirements.

Ready to source addresses for your AI services?

Visit the InterLIR Global marketplace to explore available IPv4 blocks for lease or purchase.