Twenty-five microseconds sounds meaningless to a person clicking a website.

In an electronic trading system, it can be enormous.

Market-data feeds travel from trading venues to algorithms. Algorithms react. Orders travel back toward exchanges. Competing firms may be attempting to react to the same price movement at nearly the same moment.

In that environment, shaving tens of microseconds from the physical route can materially alter who receives information or reaches a venue first.

McKay Brothers and Relativity Networks announced their partnership on November 2, with the initial focus on interconnecting important financial-market locations in New Jersey.

McKay has purchased Relativity’s ChronoCore anti-resonant hollow-core fiber for commercial deployment and plans to offer both dedicated fiber-pair IRUs and managed connectivity services from 10 Gbps through 400 Gbps.

The planned network also includes branches toward data-center development projects near financial corridors, connecting the financial-services use case with growing AI training and inference infrastructure.

A faster transmission medium cannot rescue a terrible route.

Financial-network engineers have spent years optimizing physical path length because every unnecessary mile adds propagation delay.

Hollow-core fiber changes the tradeoff because it can travel a longer distance while maintaining the same propagation time.

For example, if a conventional route is 10 miles long, a hollow-core route could be roughly 14.7 miles and still have approximately the same propagation time.

That can be strategically valuable because the geographically shortest path is not always the easiest or cheapest path to construct.

Hollow-core fiber attempts to bring propagation speed much closer to wireless while retaining the fundamental operational model and bandwidth advantages of a physical optical cable.

A new academic review published November 5 examines hollow-core fiber from a network-design perspective rather than simply comparing one kilometer of HCF with one kilometer of ordinary fiber.

Its central conclusion is important: selective deployment can capture a significant share of HCF’s network-wide benefit without rebuilding every link.

In its edge-data-center consolidation studies, upgrading only about 6% to 24% of total network fiber length to HCF could produce large benefits, including scenarios with as many as 50% fewer required edge data centers.

Other modeled use cases examined selective HCF deployment for higher-power optical transmission, bidirectional capacity scaling and coexistence with quantum channels.

Not every segment of a trading network has equal economic value.

A link between two administrative offices does not need the same propagation speed as a route connecting a market-data source to a trading engine.

Even within a low-latency network, only a few critical corridors may dominate total propagation delay.

For most businesses, hollow-core fiber is not going to make Microsoft 365, Salesforce or ordinary Internet browsing noticeably faster.

A 25-microsecond propagation improvement is tiny compared with application processing, Internet routing, database queries, cloud workloads and other delays commonly measured in milliseconds.

The AI use case has a different economic logic from trading.

Data centers increasingly need to be built where large quantities of electrical power are available. That location may not be immediately beside an existing compute campus.

If hollow-core fiber reduces propagation delay by roughly one-third, two facilities can be placed farther apart while remaining inside the same network-latency budget.

The area accessible around a facility can therefore increase much more than the linear distance increase suggests.

The propagation savings are real, but end-to-end latency still contains many other components.

Hollow-core fiber now has published deployments, commercial manufacturing partners and paying customers.

What it does not yet have is a broadly transparent commodity market comparable with ordinary G.652 or G.657 single-mode fiber.

That means a useful business case should not assume HCF costs 2×, 5× or 10× standard fiber without an actual quote.

If application latency is measured in tens of milliseconds, saving 25 or 50 microseconds may be operationally irrelevant.

In those environments, a company will usually obtain far more benefit by improving route diversity, reducing application processing time, choosing a closer cloud region, increasing bandwidth or eliminating unnecessary network hops.

The new November 2026 academic review provides perhaps the most practical road map.

Instead of assuming every conventional fiber strand must eventually become hollow-core, the researchers model hybrid networks where only selected high-value segments are upgraded.

Depending on the use case, substantial benefits appeared with only a fraction of the network converted.