Escrow.com published the second edition of its IPv4 Investment Index on August 27, 2026, reporting 107 transactions worth $9.37 million in the second quarter, average prices of $24.67 per address on large blocks, and a cumulative $557.7 million moved through the service since 2013.

What the quarter recorded

The second quarter closed with 107 IPv4 transactions passing through Escrow.com, worth $9.37 million in aggregate. That is the same deal count the company recorded in the second quarter of 2025, when the total came to $10.03 million. Average deal size fell to $87,538, a decline of roughly 6.6 percent year over year.

Volume across the preceding four quarters ran as follows, according to Escrow.com: $10.03 million on 107 deals in the second quarter of 2025, $15.12 million on 106 deals in the third, $8.95 million on 115 deals in the fourth, and $23.46 million on 123 deals in the first quarter of 2026.

That first-quarter figure carries a footnote worth reading. Escrow.com originally published $13.3 million across 89 deals in April. The revised number is $23.46 million across 123 deals, an upward restatement of 76 percent in value and 38 percent in deal count. The company attributes the gap to settlement timing, noting that IPv4 transactions can take months to complete once terms are agreed. Any reading of the second quarter therefore sits on provisional ground, and the same revision mechanism that lifted the first quarter may lift this one.

The report states that 374,000 addresses transferred during the quarter on its key findings page. Its outlook page states 366,000. Escrow.com does not reconcile the two figures, and neither carries a qualifier that would explain the difference.

Price is where the report puts its weight

Blended price across all Escrow.com IPv4 transactions in the second quarter reached $21.35 per address, against $20.60 in the first quarter. That is a move of 3.6 percent across three months, which the report itself characterises as barely moved.

The number the company leads with is different. Across three transactions involving blocks of /16 or larger, the average price came to $24.67 per address. Escrow.com frames that figure against a $20-by-year-end level it says brokers had predicted a quarter earlier. The index does not name those brokers, cite the forecast, or state the methodology behind it, so the comparison rests on an unpublished benchmark.

Three deals is a small base for a price signal. The report does not disclose the block sizes, the buyers, or the dates within the quarter, and a single unusually priced /16 would move a three-deal average by a wide margin. The blended figure covering all 107 transactions is the more robust of the two, and it moved 75 cents.

There is a further arithmetic gap. Dividing the quarter's $9.37 million by the 374,000 addresses reported as transferred produces roughly $25 per address. Dividing by 366,000 produces roughly $25.60. Neither matches the $21.35 blended figure, which suggests the blended price is calculated on a different basis than quarterly dollar volume over quarterly address volume. The index does not describe that basis.

Concentration sits at the top of the book

Deal count held steady. Deal composition did not.

The first quarter of 2026 contained three transactions above $1 million, worth $7.80 million combined. The second quarter contained two, worth $4.03 million, with the largest single settlement at $2.88 million. On a base of roughly one hundred deals per quarter, the removal of one seven-figure buyer accounts for most of the quarter-on-quarter decline in value.

Across the first half of 2026, five transactions cleared $1 million. Together they carried $11.83 million out of $32.83 million transacted across 230 deals. A little over two percent of deal volume therefore accounted for 36 percent of dollar value.

The distribution beneath that tier is heavily weighted toward small transfers. Escrow.com reports 155 deals under $50,000 in the first half, worth $2.6 million between them. A further 46 deals fell between $50,000 and $250,000, worth $5.9 million. Twenty-four deals ran from $250,000 to $1 million, worth $12.5 million. Five deals sat in the $1 million to $5 million band.

Two hundred and thirty transactions and a $32.83 million total reconcile cleanly across those four bands, which is more than can be said for the address transfer count.

The demand thesis, and what the dataset does not contain

Matt Barrie, chief executive of Escrow.com, tied the price movement to a specific driver.

"The IPv4 market has entered a structural phase shift. The emergence of agentic AI has created an entirely new category of demand. At the same time, legacy holders who were going to sell have largely sold, compressing inventory just as demand accelerates. With Escrow.com having facilitated over half a billion dollars in transfers across 80+ countries, we're launching this Index to give a market of this scale the transparency it deserves," Barrie said.

According to Escrow.com, every autonomous AI agent coming online requires its own IPv4 address, against a supply that has been fixed since the global pool of unallocated addresses was exhausted in 2011.

The index contains no transaction-level attribution supporting that claim. Buyers are aggregated by country, not by sector or use case, and no field in the published dataset distinguishes an address block bought for agent infrastructure from one bought for a hosting business, a mobile network, or a content delivery estate. The demand thesis is asserted alongside the numbers rather than derived from them.

The premise also sits awkwardly against how large automated systems actually address themselves. Published crawler address files show operators running from cloud provider allocations rather than their own networks. OpenAI's file listed 21 IPv4 prefixes as of August 2026, most of them inside Microsoft Azure ranges, as documented in coverage of GPTBot. That is a pattern of shared, provider-held space rather than per-agent allocation. Where address pools have visibly expanded, the expansion has been lumpy and institutional: Applebot added twenty-one IPv4 prefixes in August 2026, taking its published total from 2,400 addresses to 7,056, an increase of 194 percent, all inside a single /16 in consecutive runs consistent with one allocation.

None of that disproves rising demand from automated systems. It does suggest the buyers are hyperscalers and service providers acquiring in bulk, which is what the whale concentration in Escrow.com's own book shows, rather than a long tail of individual agentic AI deployments each acquiring an address.

Independent measurement of the traffic side has been running for some time. Automated requests reached 57.5 percent of web page requests in June 2026, the first recorded crossover past the halfway mark. HUMAN Security reported automation growing eight times faster than human web traffic in April 2026, and blocking rates climbing even as agent volume dipped the following month. The traffic is real. The link between that traffic and per-address purchasing remains an inference.

A federal programme, restated upward

Escrow.com identifies a second demand source in the United States Broadband Equity, Access and Deployment programme, valued at $42.45 billion.

The company corrects its own prior reporting here. The inaugural index published $22 billion in the first quarter, which Escrow.com now states was wrong. According to the report, most eligible states and territories hold final proposals approved by the National Telecommunications and Information Administration and are signing subgrant agreements, with construction running through 2026 and into 2028. That is a change from the position in April, which the report characterises as money that had not yet moved.

Two self-corrections in one document, on the first-quarter transaction total and on the programme value, are a reasonable signal about how much weight a single quarter of this index can carry.

Google in the address market

The report notes that Google appeared as a top recipient of IPv4 address space in July, attributing that to multiple news reports rather than to primary allocation records. Escrow.com states explicitly that it cannot stand behind the market-wide volume estimates carried in those reports, while treating the named event as a fact in its own right.

The company flags the position as noteworthy given that Google runs one of the more advanced internal IPv6 networks in the industry. An operator with mature IPv6 deployment acquiring legacy IPv4 space points to the backwards compatibility problem the report describes elsewhere: IPv6 provides a far larger pool, but enterprises still need legacy address blocks to remain reachable across the whole internet.

Google's own posture toward IP data in advertising has moved in the same direction over the same period. The company added IP address ingestion to Customer Match through the Data Manager API on May 28, 2026, with the European Economic Area, the United Kingdom and Switzerland carved out. On June 1, 2026, AdSense gained a publisher control that restores the full fourth octet of an IPv4 address in programmatic bid requests, off by default. From August 3, 2026, Google began using IP addresses for measurement and personalisation across those same three European territories. After years of truncation, the direction of travel on IP signal access has reversed.

Where the money comes from and where it goes

Escrow.com reports buyers in 89 countries and sellers in 88 across the full history of its IPv4 book, covering 3,944 deals since 2013. Barrie's statement refers to more than 80 countries, and the pitch accompanying the report describes 3,944 deals in 89 countries without distinguishing buy side from sell side.

The United States leads both sides. Between 2020 and the second quarter of 2026, American buyers accounted for $194.0 million and American sellers for $199.7 million. On the buy side, Israel follows at $38.0 million, then China at $37.0 million, Germany at $30.3 million and France at $20.6 million. The United Kingdom, the Philippines, Spain, Switzerland and Romania complete the top ten, ranging from $11.8 million down to $8.0 million.

The sell side looks different. The Netherlands ranks second at $54.1 million, ahead of Japan at $31.1 million, India at $19.9 million and the United Arab Emirates at $13.8 million. Australia, Romania, the United Kingdom, Spain and Germany fill out the list. Germany appears in both top tens, buying $30.3 million and selling $8.1 million, a net importer of address space by a wide margin. The Netherlands, home to a dense concentration of hosting and transit infrastructure, is the clearest net exporter.

Why address economics reach the advertising stack

An IP address is not only a routing resource. It is one of the most heavily used identity and quality signals in digital advertising, and its supply conditions shape how well that signal works.

Address scarcity pushes networks toward sharing. The more parties sit behind a single public address, the less that address says about any one of them. Stanford University researchers found that 5 percent of IP addresses send 55 percent of all web requests, that fewer than 0.2 percent of domains hold a unique IPv4 address, and that 44 percent of client addresses carry two or more user agents. Frequency caps, geolocation and invalid traffic filters all read a signal that is already heavily multiplexed.

Accuracy measurements have pointed the same way for longer. A study of nearly a billion address records found IP-to-postal linkages accurate 13 percent of the time, with the best vendor reaching 18 percent and the worst 4 percent. FreeWheel published its own case in February 2026 for moving connected television targeting away from raw addresses toward deterministic identity, citing dynamic allocation, VPN routing and household churn.

Then there is deliberate contamination. Residential proxy software found inside more than a quarter of sampled Samsung Tizen applications converts consumer devices into exit nodes, so that scraping and automation traffic arrives carrying a household address rather than a data centre one. Verification vendors assess quality partly by checking addresses against databases of data centre ranges and known proxy networks, a method whose resolution degrades when unrelated parties share an exit point.

Rising prices for legacy address blocks do not cause any of that on their own. They do reinforce the economics behind it. A network operator facing $24.67 per address for a large block has a stronger incentive to place more subscribers behind fewer public addresses, and every such decision further dilutes the signal that buyers, verification vendors and measurement systems are reading. The advertising industry's IP-based infrastructure was built when addresses were cheap and mostly one-to-one. Neither condition holds.

What the index measures, and what it does not

Escrow.com is one escrow provider. Its book is not the IPv4 market, and the report makes no claim that it is. Transactions arranged directly between parties, brokered elsewhere, or settled through registry transfer without an escrow intermediary do not appear.

The 2025 comparison in the report illustrates how sensitive these figures are to single transactions. Deal volume held at 425 for the year, ten percent below 2024 and inside the five-year range. Total value fell from $85.7 million to $42.9 million. Escrow.com attributes essentially all of that decline to one transaction: 2024 contained a single $43.1 million deal, and 2025 did not.

Three data points on large-block pricing, one revised quarter, two irreconcilable address counts, and a demand narrative carried by a chief executive statement rather than by transaction fields. The index does supply something the market has largely lacked, which is a published, quarterly, dollar-denominated series from a party that sees settlement. The appropriate weight to put on any single quarter of it is small.

Timeline

Summary

Who: Escrow.com, which operates as the escrow intermediary for IPv4 address transfers, published the report through its research division. Matt Barrie, its chief executive, supplied the accompanying statement. The buyers and sellers counted in the dataset span 89 and 88 countries respectively, led on both sides by the United States.

What: The second edition of the IPv4 Investment Index, covering the second quarter of 2026. It reports 107 transactions worth $9.37 million, average deal size of $87,538 down 6.6 percent year over year, blended pricing of $21.35 per address, large-block pricing of $24.67 per address across three deals, and $557.7 million transacted across 3,944 deals since 2013. It also revises the first quarter of 2026 upward from $13.3 million to $23.46 million and restates the United States BEAD programme from $22 billion to $42.45 billion.

When: The report was published on August 27, 2026, covering the quarter ended June 30, 2026, with historical series running back to 2013 and geographic breakdowns from 2020.

Where: Transactions settled globally through Escrow.com. American buyers accounted for $194.0 million and American sellers for $199.7 million between 2020 and the second quarter of 2026, with the Netherlands the largest non-American seller at $54.1 million and Israel the largest non-American buyer at $38.0 million.

Why: IP addresses function as a primary identity, geolocation and fraud detection signal across programmatic advertising, connected television and measurement systems. Supply has been fixed since 2011, and rising prices for legacy blocks strengthen the incentive for network operators to place more users behind fewer public addresses, which degrades the resolution of every downstream system reading that signal. The report also asserts that autonomous AI systems constitute a new category of address demand, a claim its published dataset does not test.