Compared with the previous Lightyear dataset, median U.S. DIA pricing moved from approximately $975 to $901 for 1 Gbps and from $2,555 to $1,987 for 10 Gbps. That is roughly an 8% reduction at 1 Gbps and more than 20% at 10 Gbps.

Meanwhile, the Fiber Broadband Association’s deployment-cost research found that physical construction moved in the opposite direction. Underground fiber deployment costs increased about 3% year over year, while aerial deployment increased about 14%. Labor represented the majority of deployment cost in both cases.

The strongest signal heading into 2027 is not a forecast from an analyst. It is what some of the world’s largest network buyers are actually signing.

The enterprise bandwidth ladder has moved dramatically. Only a few years ago, 100 Mbps was a serious primary business circuit. Today, Lightyear reports that 9.5% of the DIA connections sourced through its dataset are already 100 Gbps or above.

AT&T expanded 400G wavelength service across more than 40 U.S. metros during 2026, with eligible 400G handoffs across roughly 440,000 properties containing more than 2.3 million business tenants.

The economic case should continue improving. TeleGeography forecasts average 400G wavelength prices on key routes declining around 16% annually between 2025 and 2032.

Fiber cable and optical transceivers are related markets, but they are not the same supply chain. A company can have plenty of glass in the ground and still struggle to obtain the exact high-speed electronics needed to light it.

AI-network growth has driven extraordinary demand for 800G and emerging 1.6T transceivers. Industry research has identified indium-phosphide lasers and related high-speed optical components as important production constraints.

Co-packaged optics moves optical engines much closer to the switching silicon, reducing the electrical distance between the switch chip and the optics. At enormous AI-network speeds, that can reduce power and improve bandwidth density.

2026 has brought genuine movement from demonstrations toward deployments, and 2027 should produce more real-world CPO systems. But the transition will not make pluggable optics obsolete overnight.

The latest direction is actually more nuanced than “CPO versus pluggable.” Cisco and NVIDIA announced work in October 2026 on an open next-generation 400G-per-lane pluggable-optics specification, demonstrating that the pluggable ecosystem itself continues evolving aggressively.

XGS-PON has become the mainstream next-generation fiber-access platform. Its symmetrical 10G capability is enough for most residential users and a surprisingly broad portion of the business market.

The next steps are becoming more interesting. 25G PON is already commercially deployed for high-capacity business access, mobile transport and specialized applications.

Nokia announced in October 2026 that its new multi-PON platform supporting GPON, XGS-PON, 25G and 50G on the same port is scheduled for general availability in 2027. That gives operators a way to increase capacity without deciding years ahead exactly where each PON generation will be needed.

The federal Broadband Equity, Access, and Deployment program has spent years in planning, mapping, rulemaking and award processes. By mid-2026, nearly all state and territory final proposals had been approved.

Federal officials estimate BEAD will serve roughly 3.8 million locations with around $18.4 billion awarded for deployment under revised plans. Actual completion stretches well beyond 2027, but more projects are moving from paperwork toward field activity.

Pennsylvania provides a useful example. Its published BEAD timeline places planning, design, permitting and pole work in late 2026 and network construction beginning in 2027.

Fiber Broadband Association research estimated that the industry needs tens of thousands of additional construction and technician workers to absorb planned federal and state broadband investment.

Its more recent deployment-cost research shows why this matters financially: labor accounts for roughly 72% of underground fiber deployment cost and 64% of aerial deployment cost.

A cheap spool of cable therefore does not automatically produce a cheap fiber build. In many off-net projects, civil construction and labor remain far more important than the glass itself.

Traditional long-haul planning focused heavily on route distance, attenuation and available strands. AI campuses add another question: how many fibers can physically fit into the duct?

Corning says typical data-center-interconnect fiber counts have already doubled in recent years. Emerging AI campuses can require thousands of fibers per cable run and multiple conduits with enormous aggregate counts.

That is why manufacturers are emphasizing smaller-diameter fibers, higher-density ribbon cables, new connector systems and, eventually, technologies such as multicore fiber.