Meet the Speakers

Klemens Birner – Product Manager, Colt Technology Services

Parvez Khan – GM, Network Architecture Engineering (Transmission), TATA Communications

Chaz Kramer – SVP Product, Fiber and Transport, Zayo

Kent Jordan – Advisor, Product & Technology Marketing, Ciena

Kyle Hollasch – Lead Analyst Transport Hardware – Cignal AI

Presentation Summary

Compiled with AI assistance from the session transcript.

Key takeaways

• Hyperscalers still build much of their own network – the majority in some regions – but network operators play a central role where geography, regulation, or time-to-deploy favor them.
• MOFN demand is real, global, and already turning service provider revenue: North American SP spend rose 32% year over year last quarter.
• EMEA and India are the major markets outside North America – EMEA the larger of the two, India the fastest-growing.
• Ciena framed the technology direction: ROADM designs are simplifying toward point-to-point, C+L is becoming near-standard, and new multi-rail line systems are enabling scale-across AI and a new “lit fiber” service.

What is a MOFN?

• Definition – a dedicated, single-tenant optical network (Layer 0 + Layer 1) built and run by an operator exclusively for one customer and handed back as a managed service. The customer today is almost always a hyperscaler, increasingly a neocloud. The concept is old; only the name is new.
• The managed-services spectrum, by how much control and operational pain the customer absorbs:
  • Dark fiber self-build – customer acquires fiber and runs everything; the first choice for hyperscalers in North America, though MOFN remains significant there too.
  • Open line system (OLS) – operator delivers only the Layer 0 line system; tenant brings its own optics.
  • MOFN – the full managed service: operator’s fiber and hardware, Layer 0 and Layer 1, one tenant.
  • Managed wavelengths – lit capacity sold per wavelength on a shared network; buy them all and you have effectively recreated a MOFN.

Why hyperscalers buy instead of build

• Speed – footprint already in the ground turns a network up in months, not the years a build takes.
• Regulation and borders – crossing borders means licensing regimes operators have spent decades learning; India even requires licenses to lay fiber, making the operator mandatory.
• Fiber scarcity – on dense metros, long-haul routes, and regional chokepoints, the operators who built the footprint are often the only path to capacity. Fiber is gold.
• Sovereignty and security – full control of topology, vendor, and platform, plus encryption – and the simple fact that building networks is not the hyperscaler’s core business.

Market context

• AI turned on the hardware-spending switch in 2023; Cloud and Colo hardware spend has tripled to quadrupled since, overwhelmingly North American.
• When capacity ships as a MOFN, the spend lands as service provider revenue – the line to watch as the model scales.
• Last quarter: North American SP spend +32% YoY, EMEA ~20%, worldwide ~10%.
• EMEA – the largest MOFN market outside North America; hyperscaler connectivity flows largely through carriers across a fragmented, border-heavy map.
• India – the fastest-growing; Ciena’s Indian SP revenue more than doubled YoY on MOFN demand, orders up 40%+, with an Indian operator among Ciena’s top five customers by bandwidth shipped.
• North America – hyperscaler self-build is the first choice, but MOFN remains significant: three of Ciena’s top five customers by capacity in a recent quarter were traditional operators.

Ciena: the technology partner’s view

Ciena’s Kent Jordan mapped how the equipment is evolving to meet MOFN and AI-era demand.

• ROADM simplification – early MOFNs used colorless, directionless, contentionless (CDC) ROADMs for rerouting flexibility. Traffic is now point-to-point and fibers at full capacity on day one, designs are collapsing to colorless MUX-DMUX and Transponder Direct Attach. ROADMs still win where systems grow for years or need multi-degree distribution; fixed filters suffice for day-one full-fill.
• C+L band – table stakes in long-haul, now increasingly integrated rather than bolted on, and pushing into metro as the cheapest way to double capacity on scarce fiber.
• Coherent optics roadmap – high-performance transponders reach 1.6 Tbps per wavelength and maximize spectral efficiency; 400G/800G coherent pluggables are entering metro MOFN RFPs, giving operators a range of optics to match capacity and reach.
• Quantum-safe encryption – PQC (Post-Quantum Cryptography) algorithms built directly into performance-optimized transponders (no extra hardware), or communicate with external QKD (Quantum Key Distribution) over an open API. Both give operators a path to harden against future quantum threats.
• Federated management – a one-way, view-only feed giving hyperscalers consolidated visibility across multiple providers and domains while each operator keeps full control and can select what to expose.
• Multi-rail and scale-across – traditional amplifier huts do not scale to the hundreds of fiber pairs that scale-across AI demands. Multi-rail platforms raise fiber density and cut power at the hut, and the same technology lets an operator offer a new shared “lit fiber” DCI service.

Colt: Multi-vendor Private Wave

• Who – roughly 30 years delivering dedicated optical networks, now branded Private Wave (formerly High-Speed Services). Strongest in Europe, extended across the Atlantic by the Lumen EMEA acquisition, with footprint in Singapore, Hong Kong, and Japan.
• Multi-vendor by design – builds Private Wave in a multi-vendor environment and can deliver path-protected, node-protected, and active-active resilience, even across two or three different vendor platforms on the same service.
• Dedicated vs. shared – positions Private Wave as fully dedicated and fully reserved against shared wavelength and spectrum services, handing the customer control of topology, vendor, and platform.
• Beyond hyperscalers – demand now spans neoclouds, financial institutions, large enterprises, and carriers that lack in-region footprint and want speed without a do-it-yourself build.
• On pluggables – few requests so far; gray handover remains the norm, and higher-power plugs can eat into router and switch scaling.

Tata Communications: the India scale play

• Who – India’s enterprise connectivity leader, with 80,000+ km of terrestrial fiber, 500,000+ km of subsea, 550+ Indian POPs, and 100+ data centers – and every hyperscaler in India on its network.
• MOFN footprint – one petabyte delivered across national long-distance and metro MOFNs, 20+ networks spanning Mumbai, Delhi, Hyderabad, and Chennai, and roughly 18,000 km of NLD span. (A Tata predecessor, VSNL, delivered India’s first internet link at 9.6 kbps.)
• The India model – the operator builds and manages while the hyperscaler often drives technology choice and funds equipment, fiber, and operations; the operator owns the SLA. Licensing makes the operator a mandatory partner.
• A demanding environment – fiber-cut rate ~15x higher than developed markets and fiber loss degrading ~3x faster (relentless construction, fiber not yet an essential service, aging plant); limited OPGW utility fiber forces deep underground builds (below 2 m) across multiple agencies under uneven rights-of-way charges. Long asymmetric paths – Delhi to Chennai ~3,200 km primary / ~4,000 km protection – push customers toward high availability, physically diverse paths, low latency, tight MTTR, and monitoring portals.
• The roadmap – multi-rail optics (toward 160 fiber pairs per rack), coherent pluggables for space and power, open line systems, AC-free amplifier sites, and longer-term hollow-core and multi-core fiber – against a backdrop of Indian data center capacity growing from ~1.7 GW to ~9 GW by 2030.
• On pluggables – watching the space, weighing reach, the service demarcation point, and regulation; expects metro deployments first.

Zayo: fiber and waves at the MOFN seam

• Who – one of the largest fiber and long-haul wave players in North America, densified by the Crown Castle fiber acquisition to nearly 50 million fiber miles. Owning both the fiber and the waves business puts MOFN squarely at the intersection.
• The demand inflection – hyperscale, neocloud, and GPU-as-a-service enterprises are driving it; customers who once bought a terabyte on shared waves now ask for a full system, and past roughly 6.4 Tbps a MOFN is the only way to scale with control of your own asset.
• Operating is the hard part – building a MOFN is one thing; running it is another. Customers who buy fiber and gear to do it themselves hit a wall six to eighteen months in on headcount, centralized monitoring, day-two provisioning, and software support they never priced accurately.
• A path to ownership – Zayo offers governance that lets a customer eventually take ownership of the asset, after learning the operational lessons first – and positions MOFN as more cost-effective than DIY over the first five years.
• On pluggables – flexible where it already knows how to operate the architecture; a cost-versus-scalability call.

Themes from the Q&A

• Transponders vs. pluggables – MOFNs are mostly long-haul, still transponder territory (1.6 Tbps per wavelength); coherent 400G/800G pluggables are entering at the metro edge and in RFPs, but operator pull is light and reach is the gate.
• AI training vs. inference – training is concentrated, point-to-point, and high-fill (C or C+L by fiber richness); inference is distributed and favors flexible multi-degree ROADMs, with AI pushing toward Tier 2 and Tier 3 sites over time.
• The operator’s edge – against self-building hyperscalers, operators win on speed, lifecycle expertise (NOC, sparing, SLA and MTTR), fiber-route and regulatory access, and the reality that operating a network long-term is harder than building it.

What to watch

• Where the fixed-filter vs. ROADM line settles as day-one full-fill becomes the norm.
• Whether coherent pluggables push into long-haul MOFN or stay metro.
• Hyperscaler vendor mandates vs. operator-led RFPs as hyperscalers standardize MOFNs on their own gear.