Ziply Fiber Quincy Ring Is Complete, but Power Still Sets the Pace

Ziply Fiber has completed an 874-strand fiber ring in Quincy, Washington, giving the region’s data centers two paths for moving fast-growing AI traffic. The Ziply Fiber Quincy ring addresses a basic reliability problem. A single damaged route should no longer isolate a connected facility or force all traffic through one constrained path.

The project arrives as operators request more bandwidth between data centers, cloud networks, and long-haul routes. Ziply says its wavelength-service sales have tripled over the past year. That shift turns local fiber from supporting infrastructure into a factor that can influence where operators deploy expensive computing systems.

Yet connectivity is only one limit on Quincy’s growth. Grant County Public Utility District already restricts additional industrial load while expanding the area’s transmission system. The result is a revealing infrastructure mismatch: Quincy now has more network diversity, but electricity still controls how quickly its AI market can expand.

The Ziply Fiber Quincy Ring Connects the Entire Cluster

The new ring gives Quincy’s major data centers redundant local routes and direct access to Ziply’s wider transport network.

Ziply announced the completed system on September 15, 2026. Its fiber ring announcement describes a fully redundant loop built for high-capacity connections throughout the Quincy data center market.

A fiber ring connects facilities through a loop rather than a single linear path. If construction, equipment failure, or another incident interrupts one section, traffic can move in the opposite direction. That architecture limits the ability of one physical break to disable a customer’s connection.

The design matters because a nominally redundant service can still contain hidden shared infrastructure. Two circuits provide limited protection if both use the same conduit, bridge crossing, or upstream route. Genuine path diversity requires physically separate segments and enough unused capacity to absorb redirected traffic.

Ziply’s Quincy deployment uses 874 fiber strands, according to details provided to Fierce Network. Fiber strands are the individual glass paths inside a cable. The high count gives the operator room for dedicated customer connections, wavelength services, future capacity, and alternative routing.

The ring reaches a concentrated group of facilities operated by companies including Microsoft, Sabey Data Centers, NTT, CyrusOne, and H5 Data Centers. Ziply commercial general manager Chris Gellos said the company now has those data centers connected through the local system.

That arrangement supports two distinct traffic patterns. An operator can carry data from Quincy toward another regional facility, or customers can exchange traffic within the Quincy cluster. Local connections can avoid an unnecessary trip to a distant internet backbone before returning to another nearby facility.

That distinction becomes important for AI infrastructure. Training systems regularly move model checkpoints, datasets, and operational telemetry across computing and storage environments. Inference services also depend on predictable connections among data centers, cloud regions, and end-user networks.

The ring includes a direct on-ramp to Ziply’s long-haul network. It therefore functions as more than a local resilience project. It joins Quincy’s data center campuses to a broader system spanning the Pacific Northwest, Mountain West, Midwest, and other markets.

The immediate change is straightforward. Operators in Quincy gain another carrier option with diverse local paths and long-haul access. The larger question is whether that network capacity can translate into additional computing deployments when the electrical system remains constrained.

AI Workloads Are Raising the Cost of a Fiber Failure

AI does not merely increase traffic volume; it raises the operational value of predictable bandwidth between expensive computing clusters.

Traditional enterprise applications often tolerate modest variations in network performance. Some AI workloads create a different operating profile. They combine large datasets, distributed storage, dense accelerator clusters, and frequent transfers between regional computing sites.

A training job can involve thousands of processors performing coordinated work. Much of that communication stays inside one facility, where specialized networking connects servers. However, the surrounding data pipeline still depends on metro and long-haul fiber for replication, backups, cloud access, and workload movement.

Those external connections become more important as operators distribute infrastructure across multiple buildings. A single campus might not provide enough power, cooling, or construction space for every deployment. Companies then need reliable links among nearby facilities and larger regional hubs.

Ziply says demand is already visible in its commercial results. Gellos told the industry interview that wavelength-service sales tripled during the previous year. Wavelength service assigns a customer dedicated optical capacity over a fiber network.

The company also applies a 40 percent utilization ceiling to individual fiber spans, according to the interview. That policy preserves headroom for rerouting traffic after a break. It is a capacity-management choice rather than a guarantee that every disruption will be invisible.

Leaving most of a span unused can appear inefficient during normal operation. The reserve becomes valuable when traffic must move from a failed segment to surviving circuits. Without enough spare capacity, physical diversity alone can still produce congestion during an outage.

Ziply’s approach reflects the economics of modern data centers. An operator investing heavily in AI accelerators does not want those systems waiting because an external circuit failed. Connectivity represents a small share of the facility’s total investment, but it can affect the utilization of far more expensive equipment.

Quincy also sits within a wider regional network. In April, Ziply activated its 2,100-mile Northern Link Route between the Pacific Northwest and Chicago. The company says the system supports 400-gigabit wavelengths and recorded a 39.5-millisecond Seattle-to-Chicago round trip during testing.

The Northern Link route passes through markets including Seattle, Spokane, Missoula, Billings, Bismarck, Fargo, Minneapolis, and Madison. Its Quincy ring can feed traffic into that broader east-west corridor.

Together, the local ring and long-haul route form a more complete transport proposition. Quincy customers gain connectivity among nearby campuses, plus onward paths toward major cloud and interconnection markets. That combination is more useful than either asset would be alone.

However, Ziply has not disclosed signed customers, contracted capacity, traffic volume, or revenue tied specifically to the Quincy ring. Its companywide wavelength growth demonstrates demand, but it does not reveal utilization of this particular project.

That distinction matters. Announcing available infrastructure is not the same as proving adoption. Customer connections, activated wavelengths, and sustained traffic growth will show whether the ring becomes essential infrastructure or mainly additional optionality.

Quincy’s Data Center Advantage Was Built Before the AI Boom

The fiber project reinforces an established data center market rather than creating a new cluster from scratch.

Microsoft and Yahoo began developing large facilities around Quincy during the mid-2000s. Other operators followed, creating a market with hyperscale campuses, colocation facilities, carrier networks, and specialized construction expertise.

Three local advantages drove that expansion. The region offered relatively inexpensive land, access to hydroelectric power, and a cool climate that can assist facility operations. Washington tax incentives also reduced the cost of eligible data center equipment.

That combination attracted cloud and internet companies well before generative AI became the industry’s central investment theme. AI now adds another demand layer to infrastructure originally built for search, storage, consumer internet services, and general cloud computing.

Quincy’s concentration creates a network effect. Carriers can justify building additional routes because multiple large customers occupy a small geographic area. Operators then gain more choices for transit, interconnection, and disaster recovery.

The concentration also gives customers reasons to exchange traffic locally. Moving data directly between nearby facilities can reduce unnecessary long-distance transport. It can also support architectures that place computing, storage, and backup capacity across separate campuses.

Ziply has used the same broad strategy elsewhere. Its Hillsboro, Oregon, network serves another dense data center market. The company has also expanded connections around Seattle and other Pacific Northwest infrastructure hubs.

The Quincy ring strengthens competition among carriers already serving the area. Available providers have included regional and national fiber operators offering lit services, dark fiber, internet transit, or long-haul transport. Customers often use several carriers to reduce dependence on one network.

Lit fiber is managed capacity provided by a network operator. Dark fiber gives a customer unused strands that it can equip and operate. The two models serve different buyers, budgets, and levels of technical control.

Ziply’s 874 strands give it flexibility across those needs, although the company has not published the ring’s commercial mix. It can allocate capacity to managed wavelengths, dedicated connections, or future expansion as demand develops.

The more important competitive shift is route diversity. A new provider offers limited resilience if its physical path overlaps existing infrastructure. Ziply says the completed loop supplies diverse paths to facilities and connects directly to its wider backbone.

Quincy’s position also reflects its location between coastal cloud markets and inland transport routes. Seattle and Hillsboro provide major interconnection points, while eastern routes can reach Spokane, the northern plains, Minneapolis, and Chicago.

The completed ring makes that geography more usable for data center operators. It does not shorten every application path or replace larger internet exchanges. It does give customers another way to reach those markets without depending on one local route.

This is why the project is more consequential than an ordinary municipal fiber build. It targets a cluster containing some of the world’s largest computing operators. The value comes from the facilities connected, the traffic they generate, and the paths available after a failure.

Still, Quincy’s historical advantage rested heavily on electricity. AI infrastructure intensifies that dependency because modern facilities request larger blocks of continuous power. Fiber capacity can support growth only when the grid can energize the servers at the other end.

Better Fiber Does Not Remove Quincy’s Power Ceiling

Quincy’s central tension is now clear: network capacity is expanding faster than the power system can accommodate new industrial demand.

Grant PUD serves eight data center operators in Grant County. Those customers used approximately 280 average megawatts during 2025, according to the utility’s data center guidance.

An average megawatt measures sustained consumption over time rather than a momentary peak. The figure shows that data centers already represent a large, continuous load on the county’s electrical system.

Demand accelerated as operators sought capacity for AI infrastructure. Grant PUD said its forecasting and transmission analysis identified operational risks in the Quincy industrial zones. It introduced interim customer load limits beginning in April 2025.

The limits do not require existing industrial customers to reduce consumption below current levels. Instead, they restrict additional load until the utility completes more transmission capacity. Grant PUD describes them as a temporary reliability measure.

That policy changes how operators should interpret the new fiber ring. Connectivity can make a location more attractive, but it cannot provide the electrical service needed for new servers. The two systems must expand together.

Grant PUD’s Quincy Transmission Expansion Plan is intended to address that gap. The program includes six transmission-line segments, a new switchyard, and expansions of the Mountain View and Wanapum facilities.

The utility says the plan will raise the Quincy area’s load limit from 372 megawatts to 750 megawatts. Its transmission expansion plan lists different construction and service dates across the individual projects.

Several initial components target service during 2027, while the broader program extends beyond that point. Grant PUD has also discussed completing the collection of planned upgrades by 2029.

Those timelines introduce execution risk. Transmission projects require permitting, rights-of-way, specialized equipment, contractors, and coordinated construction. Delays in one component can affect when customers receive additional capacity.

The expansion has also created local conflict. New transmission corridors cross privately owned land, and Grant PUD has pursued property rights needed for construction. That process demonstrates how AI infrastructure can move from corporate capital plans into community land-use disputes.

Power procurement creates another uncertainty. Transmission lines move electricity, but they do not generate it. Grant PUD must secure enough energy to meet new loads while maintaining reliability and complying with Washington’s clean-energy requirements.

The utility says large power users should fund the additional resources and infrastructure required for their growth. That principle seeks to prevent the cost of serving data centers from shifting unfairly to households, farms, and small businesses.

Implementation will matter more than the principle alone. Infrastructure costs can be assigned through connection agreements, rates, deposits, and customer-funded projects. The exact structure affects both local ratepayers and the economics of future data center construction.

This power constraint prevents an overly simple reading of Quincy’s AI boom. The region possesses desirable land, fiber, operating experience, and access to hydroelectric resources. It does not possess unlimited immediately available electricity.

The Ziply Fiber Quincy ring solves a real network problem within that larger system. It improves routing options and makes failures easier to manage. It does not resolve pending power requests or accelerate transmission construction by itself.

For operators, the practical question is therefore not whether Quincy has good fiber. It is whether power delivery, carrier capacity, site construction, and equipment availability align on the same schedule.

Redundancy Claims Need Operational Evidence

The ring’s architecture is credible, but its value will depend on physical separation, customer adoption, and performance during real disruptions.

Ziply describes the loop as fully redundant and says each data center receives two paths. That design is appropriate for critical infrastructure. However, public announcements provide limited detail about conduit maps, shared crossings, switching behavior, or customer-specific service configurations.

Physical diversity can fail in subtle ways. Separate cables might converge at a bridge, railroad crossing, utility corridor, or central office. A single excavation or equipment incident can then affect both routes.

Operators typically assess those risks through route audits and contractual service designs. They may request maps, verify entry points, use separate carriers, or combine managed wavelengths with dark fiber. Those details remain customer-specific.

The ring’s 40 percent utilization policy adds another layer of resilience. Spare capacity can receive traffic shifted from a damaged span. Yet actual results depend on the size of the failure, traffic distribution, and whether multiple incidents occur together.

Automated rerouting also involves tradeoffs. A network can restore service quickly while still producing brief packet loss, latency changes, or congestion. Applications with strict performance needs may require additional safeguards above the carrier layer.

Ziply has not published outage tests or recovery measurements for the Quincy ring. It also has not identified which facilities have activated service. Those omissions are normal for a newly announced commercial network, but they limit independent evaluation.

The 874-strand count should also be interpreted carefully. It describes physical inventory, not immediately sold bandwidth. Capacity depends on installed optical equipment, wavelength configurations, customer demand, and how strands are allocated across the route.

Likewise, Ziply’s reported tripling of wavelength sales applies to the company’s broader business. It supports the claim that high-capacity transport demand is rising. It does not establish that Quincy alone produced that growth.

Customer behavior offers the clearest test. Large operators commonly purchase diverse connections from multiple providers. If Quincy facilities add Ziply as a secondary carrier, the ring can improve resilience even without becoming their primary path.

A more significant outcome would involve customers using the loop for regular high-volume traffic between facilities. That pattern would demonstrate that local interconnection has become part of their operating architecture, not simply backup capacity.

Price and contract terms will influence those decisions, although Ziply has not disclosed them. Buyers will compare route diversity, restoration commitments, capacity options, network reach, and operational support against competing carriers.

The project also faces a sequencing problem. New fiber can arrive before new power, leaving capacity underused while customers wait for electrical allocations. Conversely, the early build can ensure connectivity is ready when transmission constraints ease.

Building ahead of demand is common in network infrastructure. Fiber has a long useful life, and adding strands during initial construction is usually easier than reopening the route later. The risk is commercial timing rather than immediate technical obsolescence.

That makes the current announcement an infrastructure milestone, not proof of a completed AI expansion cycle. The loop is available. Its strategic importance will become measurable only as customers activate services and the grid supports additional computing capacity.

Three Signals Will Show Whether the Ring Changes Quincy’s Market

Customer activation, power expansion, and additional regional rings will determine whether Ziply’s project produces lasting competitive value.

The first signal is disclosed customer adoption. Ziply does not need to name every customer, but activated wavelengths or on-net facility announcements would demonstrate commercial use. Sustained growth in data center transport would strengthen the case that AI demand is driving the build.

Usage evidence should include more than a restatement of companywide sales growth. Quincy-specific connections, new interfacility services, or expanded long-haul capacity would offer a clearer measure. A lack of visible adoption would not prove failure, but it would weaken the project’s immediate significance.

The second signal is progress on Grant PUD’s transmission program. Construction milestones and in-service dates will determine when the area can support more industrial load. Movement from the current 372-megawatt limit toward 750 megawatts would reinforce Quincy’s expansion case.

Delays would have the opposite effect. Fiber would remain available, but prospective data centers could face longer waits for power. Existing operators might still purchase more bandwidth, although the market’s physical growth would stay constrained.

The third signal is Ziply’s rollout of similar rings along Northern Link. Gellos has indicated that the company is evaluating data center clusters around that route. Additional metro systems would turn the long-haul corridor into a chain of connected computing markets.

That strategy would place Quincy inside a broader portfolio rather than treating it as a standalone build. Customers could buy local access and east-west transport from one operator across several cities. Competitors would then face pressure on both reach and route diversity.

Execution will decide whether that network becomes differentiated. Ziply must connect the right facilities, preserve genuinely separate paths, and carry traffic reliably. It must also coordinate expansion with customer deployments that depend on power, construction, and computing hardware.

For developers and enterprise buyers, the lesson extends beyond Quincy. AI infrastructure planning cannot evaluate power, fiber, and data center space independently. A strong location needs all three resources available on compatible timelines.

Teams considering regional AI capacity should watch actual service activations, not just construction announcements. They should also examine physical routes, reserved failover capacity, power-delivery dates, and the financial responsibility for supporting infrastructure.

The Ziply Fiber Quincy ring gives the market a stronger network foundation. It adds 874 strands, alternative paths, local interconnection, and access to a 2,100-mile transport corridor. Those capabilities directly address the growing movement of data around AI systems.

The remaining constraint sits outside the cable. Quincy’s grid must catch up with the computing demand that attracted Ziply’s investment. Until that happens, fiber will make the region more resilient without making its supply of deployable power unlimited.

The next few months should reveal whether customers begin activating the new routes and whether Grant PUD’s projects remain on schedule. Watch those signals together. They will show whether Quincy is becoming a balanced AI infrastructure market or a well-connected cluster waiting for electricity.