Skip to the stories
Microsoft 365 & SharePoint

Data Center Interconnect Planning for 2027: Lock In Capacity, Keep Paths Flexible

  • Natalie Stipan
  • 959 views

Lead times are stretching for optical gear. Secure DCI capacity early, prove true route diversity, and build passive plant ready for 800G and 1.2T upgrades.

Data Center Interconnect Planning for 2027: Lock In Capacity, Keep Paths Flexible

Most DCI conversations still start with the switch or the transceiver. My read is that the constraint has moved outward. The bottleneck now sits in fiber routes, duct space and supplier schedules, and teams that plan 2027 capacity around the server hall will find themselves waiting on things they never costed.

Demand is outrunning the build-out

Dell'Oro Group reports that revenue from wavelength-division multiplexing systems and ZR/ZR+ coherent optics for data center interconnect grew 45% year over year in the second quarter of 2026. North America accounted for more than 40% of global optical transport equipment revenue that quarter. Dell'Oro VP Jimmy Yu put it plainly: "There is strong demand for optical network equipment between neighboring data centers and across long-distance backbones."

That demand is pulling in two directions at once. Facilities need to exchange data at specific speeds, latencies and availability levels, and AI clusters spread across campuses and regions make those requirements tighter. Compute only delivers value if the links between sites perform, so interconnect has become a dependency for the whole build, not an afterthought.

Fiber on a map isn't fiber in service

CBRE's first-half 2026 research counts more than 20,000 route miles of new fiber projects announced during the year, linking emerging AI markets with established network hubs. That's a large number, but announced routes aren't operational connections. Rights-of-way have to be secured, construction completed, cable installed and links commissioned, and each step takes time that a press announcement doesn't show.

Power availability is already shaping where facilities go, which makes connectivity assessment part of site selection rather than a later phase. Before committing to a site, verify actual route availability, delivery schedules and achievable latency. Don't assume that a nearby fiber corridor will give you the capacity or resilience you need. Teams in this position typically learn that a corridor exists long before they learn who can light it, and when.

Two carriers, one failure point

Redundancy deserves the same skepticism. Two carrier contracts don't guarantee two independent physical paths. Shared ducts, bridges, building entrances or splice locations can expose both services to the same incident, and that's the failure mode that tends to surface during an outage rather than during design review.

Route diversity has to be checked end to end: from outside plant, through the entrance facility, and into internal distribution. If the diverse paths converge at a single entrance room or splice point, the contract paperwork is doing more work than the glass is. Ask carriers for route documentation that shows where their paths physically join, not just where they terminate.

Read manufacturing announcements carefully

Supply investment is clearly moving. A January 2026 hyperscaler agreement worth up to $6 billion supports expanded U.S. optical cable manufacturing. Separately, another manufacturer broke ground in September on an $80 million expansion in South Carolina, with construction scheduled for completion in the first half of 2028.

Those investments should strengthen future supply, but an announcement is not a delivery commitment. Construction completion is also a different milestone from qualified production at scale, and each expansion will ramp on its own schedule. Don't apply one blanket lead time across every supplier and product line.

The practical response is to reserve critical capacity early while keeping room to change specifications and release dates. Separate firm requirements from forecast demand, qualify compatible alternatives in advance, and hold targeted restoration spares. Indiscriminate stockpiling is the trap here. It ties up capital in components that may not match the architecture you end up deploying, and that architecture is still moving.

Build the physical layer for change

DCI covers campus, metro and long-haul links, and each carries different requirements. Single-mode fiber selection should follow distance, wavelength plans, attenuation budgets and the transmission system you expect to run. Cable construction has to match duct capacity, pulling loads, bend limits and environmental exposure. A cable that fits the design sheet can still fail at the first sharp bend in a real duct run.

At building entrances, high-count cables need manageable splicing and clean transitions into indoor plant. Ribbon-compatible splice management, protected enclosures, modular optical distribution frames and organized slack storage all make growth easier to absorb. Density helps, but technicians still need to reach individual connections without disturbing adjacent live circuits. A dense panel nobody can work in is a liability during the next change window.

Budget loss honestly. Account for fiber, connector and splice losses, then keep engineering margin for repairs and future changes. Factory-tested assemblies can make installation more repeatable, though field splicing stays necessary wherever routes and cable lengths demand it. Acceptance testing should verify end-to-end insertion loss, with optical time-domain reflectometer traces recording splice quality and fault locations. Those traces are the baseline you'll compare against when something goes wrong at 2 a.m.

Plan for 800G and 1.2T without ripping out plant

Coherent transmission is heading toward 800 Gbps, 1.2 Tbps and higher wavelength capacities. The passive plant should carry those upgrades without forcing wholesale replacement. In practice, that means reserving duct space, spare fiber capacity and distribution-frame expansion positions now, and using modular termination and splice arrangements with documented interfaces so future optics slot into known connections.

Access to a broad, compatible portfolio helps teams adapt cable construction, termination and protection to each location. Pair that range with realistic logistics, staged deliveries and installation skills. Avoid choosing hardware on port density alone, relying on unverified lead times, or customizing every connection when a standard arrangement would do.

Records matter as much as hardware. Keep accurate logs of fiber assignments, splice locations, test results and physical routes, and reassess them after every expansion. An apparently minor patching change can quietly undo a diversity design you paid for. Restoration procedures should name the responsible teams, the spares on hand and the access arrangements before an outage, not during one.

What to take into the next planning cycle

Successful DCI planning for 2027 and beyond comes down to three things: capacity that arrives on schedule, physical paths that hold up when something breaks, and infrastructure that can change as fast as the business does. If you only get one of those right, the other two will eventually cost you. Start with the route audit and the supplier conversations this quarter, because those are the items with the longest lead times.

Source: Natalie Stipan · datacenterpost.com

Neighbouring cells