AI Is Already Here. BEAD Is Still Warming Up.

AI demand is no longer some future theory living in a pitch deck. Agentic AI, automation platforms, inference workloads, data pipelines, and cloud infrastructure all need compute & networks. Networks still rely on IPv4 for compatibility, access, routing, customer environments, security policies, and operational simplicity.

This is the boring physical layer behind all the exciting AI headlines. Everyone wants to talk about models, but fewer people want to talk about the infrastructure those models require. Infrastructure is where the bill eventually lands. This is not speculation, and not a theoretical demand curve. Real addresses are moving into real infrastructure.

BEAD is the other major demand story, but it has not fully arrived yet. For years, BEAD has lived in conference panels, regulatory filings, revised grant maps, and government announcements. Eventually, though, paperwork becomes construction. Fiber gets built & customers get activated

When that happens, broadband operators will not be entering a calm, well-stocked market. They will be walking into the same IPv4 cage match already occupied by hosting providers, AI platforms & cloud companies. The market is already tightening before one of its largest potential demand drivers has fully arrived.

The Market Forgot IPv4 Was Scarce

At the beginning of 2026, /17s and larger across ARIN, RIPE, and APNIC commonly traded around $9 per IP. Today prices have already increased to $12 and $15 per IP.

That is a 33% to 55% move in only a few months. The correction from $50-$60 per IP down to roughly $9 did something dangerous: it made buyers confuse cheaper with abundant.  IPv4 did not become less finite. Sellers blinked, buyers waited, sentiment cooled, and pricing reset. Now the tables have turned.

Sellers are less desperate & large subnets are harder to find. Buyers with real infrastructure needs are back in the market and society relies on them to keep the internet to keep working.

One Soft ARIN Metric Does Not Kill the Trend

Yes, ARIN transfer requests declined to 135 transfer requests in April, below both the 2026 monthly average and the 150 average monthly request volume seen during 2025. On its own, that statistic could be interpreted as weakening demand, but it is not the same thing as demand disappearing.

Transfer requests and transfer volume are not identical. A market can have fewer requests while still moving larger blocks, and that is exactly the kind of behavior we are seeing as larger infrastructure buyers pursue meaningful address space instead of nibbling around the edges. One soft metric does not mean the market took a nap. It means you should look at the whole chart before declaring the patient dead.