Microsoft’s manufacturing partnerships with Corning and Heraeus to produce hollow‑core fibre (HCF) for the Azure fleet mark a turning point in how hyperscalers plan, build, and scale their optical backbones—moving a laboratory curiosity into industrial reality and setting the stage for ultra‑low‑latency, high‑capacity data centre interconnects that are explicitly optimised for AI workloads. Corning will handle large‑scale manufacturing at its North Carolina glass and cable facilities, while Heraeus (now operating under its Covantics/Comvance banner) will cover European distribution and selected US volumes from its European and US plants—moves that promise to accelerate Azure’s HCF rollout while exposing the cloud ecosystem to new manufacturing, integration, and standards challenges.
Hollow‑core fibre (HCF) guides light predominantly through an air core rather than a solid glass core. That architectural change reduces propagation delay (light travels closer to vacuum speed), lowers nonlinear optical effects, and—following recent research breakthroughs—can achieve attenuation figures comparable to or better than conventional silica single‑mode fibre. Microsoft has been one of the most aggressive hyperscaler backers of HCF: the company acquired Lumenisity in 2022, has published detailed Azure technical posts about HCF deployments, and has outlined plans for wide‑scale rollouts across its network. These moves culminate in Microsoft’s decision to partner with established glass manufacturers to scale production beyond the experimental plant footprint.
Recent peer‑reviewed and press‑covered results claim HCF designs with attenuation below 0.1 dB/km at 1,550 nm using advanced anti‑resonant hollow‑core structures (DNANF), a result that materially changes the economics and deployment model for HCF in backbone, metro, and data centre fabrics. Microsoft’s lab and field work—now being translated into mass production by Corning and into regional distribution/manufacturing by Heraeus—creates a credible pathway for Azure to scale HCF across metro and inter‑datacentre links where latency and signal integrity directly affect AI training and real‑time services.
That said, the transition remains a complex, multi‑year engineering programme: standards must catch up, mass production yields must improve, operational crews must be trained, and lifecycle data must be gathered. The near‑term reality is strategic, selective deployment where HCF’s advantages are most valuable; the mid‑term prize is rearchitecting certain tiers of the cloud network to be faster and more efficient at scale. Microsoft’s partnerships accelerate that timeline—but widespread adoption across carriers and enterprises will depend on reproducible manufacturing economics, robust interoperability standards, and multi‑year reliability evidence.
Source: Telecompaper Telecompaper
Hollow‑core fibre (HCF) guides light predominantly through an air core rather than a solid glass core. That architectural change reduces propagation delay (light travels closer to vacuum speed), lowers nonlinear optical effects, and—following recent research breakthroughs—can achieve attenuation figures comparable to or better than conventional silica single‑mode fibre. Microsoft has been one of the most aggressive hyperscaler backers of HCF: the company acquired Lumenisity in 2022, has published detailed Azure technical posts about HCF deployments, and has outlined plans for wide‑scale rollouts across its network. These moves culminate in Microsoft’s decision to partner with established glass manufacturers to scale production beyond the experimental plant footprint. Recent peer‑reviewed and press‑covered results claim HCF designs with attenuation below 0.1 dB/km at 1,550 nm using advanced anti‑resonant hollow‑core structures (DNANF), a result that materially changes the economics and deployment model for HCF in backbone, metro, and data centre fabrics. Microsoft’s lab and field work—now being translated into mass production by Corning and into regional distribution/manufacturing by Heraeus—creates a credible pathway for Azure to scale HCF across metro and inter‑datacentre links where latency and signal integrity directly affect AI training and real‑time services.
Why Microsoft is investing in hollow‑core fibre
The technical imperative: latency and nonlinearity
Traditional silica single‑mode fibre (SMF) slows light to roughly two‑thirds of its vacuum speed because photons transit through glass (refractive index ~1.47). By contrast, HCF routes most optical power through air, producing a measurable propagation‑speed advantage and therefore lower group delay per kilometre. In practical terms, many HCF designs deliver approximately 30–50% lower latency per kilometre when compared to SMF—an attractive metric where every microsecond counts across distributed AI training, real‑time inference, and latency‑sensitive financial or gaming workloads. Reduced nonlinear impairments (self‑phase modulation, four‑wave mixing, etc.) also make HCF more tolerant of higher launch powers and denser wavelength‑division multiplexing.The economic and operational argument
Lower intrinsic attenuation (recent claims place DNANF‑class HCF near or below 0.1 dB/km) means fewer inline amplifiers are required for longer links, reducing capex and operational complexity for some spans. When coupled with latency benefits, HCF permits data centre designers to reimagine topology constraints—allowing more geographically distributed clusters to operate within the same latency budget, or to place backup and compute resources further apart without an undue performance penalty. Microsoft’s public technical notes highlight both the performance and operational trials—showing HCF carrying live traffic in Azure regions while integrating with existing coherent optics and DC equipment.Strategic control of supply and IP
Owning or tightly controlling HCF manufacturing and distribution is a strategic move. Microsoft’s Lumenisity acquisition provided IP, teams, and a prototype manufacturing plant; partnering with Corning and Heraeus brings industrial scale, supply‑chain resilience, and proven cable packaging and connector know‑how. Hyperscalers face intense competition for fibre capacity and cables (and for skilled manufacturing slots), so securing manufacturing partnerships ahead of volume demand is a prudent hedge to avoid bottlenecks during rapid data‑centre expansion cycles driven by AI.The partners and their roles
Corning: scaling HCF in North Carolina
Corning has been contracted to operationalise Microsoft’s HCF manufacturing processes in the U.S., producing both the HCF glass preforms and finished fibre/cable assemblies at its North Carolina facilities. Corning’s role extends beyond drawing fibre: converting the laboratory recipe into high‑yield, repeatable runs, integrating HCF into cable constructs suitable for duct and intra‑datacentre use, and delivering the connectivity hardware (splices, connectors, cable jackets) that keep Azure’s operational processes consistent. Corning’s decades of fibre and cable manufacturing experience and its global production footprint are precisely the strengths Microsoft needs to go from hundreds to thousands of kilometres of deployed HCF.Heraeus: European distribution and regional manufacturing
Heraeus (operating as Heraeus Covantics / formerly Heraeus Comvance) brings a legacy in synthetic quartz and telecom fibre manufacturing, with draw towers in Denmark and material plants across Europe and the U.S. The company’s stated role is to support European distribution and supply certain U.S. volumes from its mixed regional plants. Heraeus’s presence fills a critical regionalisation gap: hyperscalers cannot depend solely on a single production centre if they intend to deploy HCF across EMEA and transatlantic routes quickly. However, public reporting of Heraeus’s exact contractual terms and capacity commitments remains limited and should be treated as a strategic supply arrangement rather than a fully disclosed outsourcing contract.What the technical breakthrough means for deployment
DNANF and attenuation breakthroughs
Recent research describing double‑nested anti‑resonant nodeless fibre (DNANF) claims attenuation at or below 0.091 dB/km at 1,550 nm—figures that historically belonged to the lowest‑loss silica fibres and were thought to be unreachable by HCF. If reproducible and manufacturable at scale, those losses eliminate the key historical drawback of HCF (high leakage and high loss) and open the door for HCF use in much longer spans, including metro and some regional backhaul links that feed or connect data centres. Major news and technical outlets have reported these results as a watershed moment; Microsoft’s lab and pilot activity shows the company is already translating those gains into operational links.Practical integration with coherent optics and transponders
Field and lab trials demonstrated that HCF can be coupled with existing coherent optics without wholesale redesigns of transponders—though vendor firmware and link‑engineering practices need tailoring. Azure teams report live traffic and DC interconnect runs where HCF links interoperate with standard coherent DWDM gear. That compatibility reduces the barrier to entry for operators: rather than waiting for a whole new ecosystem of HCF‑native transponders, hyperscalers can achieve near‑term benefits by pairing HCF with the coherent optics they already deploy.Manufacturing and supply‑chain realities
From lab draws to continuous high‑yield production
HCF is materially different from conventional SMF to make and package. Draw‑tower processes, preform geometries, coating application, and environmental control all require bespoke tooling. Translating a lab recipe into a factory‑grade process involves yield engineering, contamination control, and long process‑qualification cycles. Corning’s announced role is exactly this: converting Microsoft’s process into mass‑manufacturing yields that meet telco reliability standards. Expect a multi‑quarter (if not multi‑year) ramp phase where trial runs, burn‑in testing, and end‑to‑end cable qualification dominate operations.Regional capacity and redundancy
Heraeus’s European plant footprint and Corning’s U.S. facilities provide distributed capacity—but the total global HCF output is still limited compared with decades of SMF manufacturing. Hyperscalers must plan rollouts carefully to match production schedules, anticipate spares and reclamation strategies, and maintain fallback paths on conventional SMF during ramp‑up. Multi‑supplier strategies (Corning, Heraeus, plus other fibre manufacturers active in HCF pilot programs) reduce single‑point risks but raise standards‑alignment and quality‑control complexity.Cost, pricing, and the unit economics question
HCF materials and processes are currently more expensive than commodity SMF owing to specialised preforms, more complex draw geometry, and slower throughput. Even with lower amplifier needs and potential layout efficiencies, the near‑term unit cost per kilometre of HCF cable will likely exceed equivalent SMF runs. The economic case therefore depends on two variables: (1) how much Azure values latency and lower nonlinearity in specific links, and (2) how quickly production scale and process improvements reduce per‑unit costs. Corning’s manufacturing scale and industrial expertise are intended to accelerate that cost curve, but the transition will not be cost‑neutral overnight.Operational and engineering challenges
Splicing, connectors, and field work
HCF has different modal and reflection properties that demand updated splicing and connector procedures. Field splicing crews must adopt equipment and SOPs calibrated for HCF to meet insertion‑loss and back‑reflectance targets. Azure’s internal operations teams have already developed new testing and installation practices during pilot deployments, but broad field operations at hyperscaler scale will require retraining, inventory changes, and new test gear procurement—costs and lead times that are often underestimated.Standards, interoperability, and vendor ecosystems
Global standards bodies and industry consortia currently lack a complete set of production‑grade standards for HCF cable geometry, connector end‑face specs, or long‑term reliability test suites. Until standards mature, multi‑vendor interoperability (between different HCF producers, transponder vendors, and cable systems integrators) will remain a risk. Microsoft’s approach—vertical integration via Lumenisity IP plus external manufacturing partners—reduces that risk inside Azure’s own deployments, but external carriers and cloud partners will need standards that allow multi‑party networks to interoperate reliably.Long‑term reliability and ageing
Glass membranes in anti‑resonant HCF are extremely thin; long‑term mechanical and photonic ageing behaviours at field temperatures and under repeated mechanical stress require multi‑year telemetry. Pilot deployments and accelerated ageing tests are encouraging, but the industry has no multi‑decade field data for DNANF‑class fibres. Azure will need robust field telemetry, lifecycle replacement plans, and conservative SLAs for initial commercial HCF links.Strategic implications for cloud and telecom markets
Data centre architecture and geographic placement
If HCF’s latency advantage is realised at scale and the economics become attractive for critical links, data centre placement strategies could shift. Operators could distribute compute clusters more widely while remaining within acceptable latency budgets for large distributed AI models, enabling better disaster resilience, renewable power matching, and cost optimisation across power and land markets.Competitive landscape and supplier power
Securing manufacturing partnerships with large, experienced suppliers (Corning and Heraeus) gives Microsoft a head start. Competitors will either need to develop their own HCF supply chains, partner with alternative manufacturers, or purchase capacity from third parties—creating bidding pressure and potential supplier concentration. Over time, larger suppliers that commit to HCF tooling and processes will exert pricing and scheduling leverage unless new players enter the market quickly.Carrier and enterprise adoption
Carriers historically adopt new fibre technologies more slowly due to cost and standards concerns. However, where latency and signal fidelity are commercially valuable (high‑frequency trading, national research networks, AI exchange fabrics), carriers will trial and commercialise HCF. Microsoft’s public deployments and scaling announcements act as a market signal that HCF is moving from R&D to production, increasing commercial credibility for telcos and system integrators.Strengths of Microsoft’s approach
- Vertical control plus manufacturing scale: Owning IP through Lumenisity while partnering with Corning and Heraeus balances innovation with industrialisation.
- Proven field trials: Azure has public reports of HCF carrying live traffic and delivering reduced latency while integrating with existing coherent optical gear.
- Strategic geographic coverage: Corning’s U.S. capacity and Heraeus’s European footprint create regional redundancy and faster deployment windows for Azure regions.
- Rapid technology maturation: Recent DNANF results lower a historic barrier (high HCF loss), creating a plausible path from niche to mainstream use.
Risks, unknowns, and cautionary notes
- Standards and interoperability remain incomplete. Until global optics and cabling standards explicitly accommodate DNANF/HCF geometries and test methods, multi‑vendor networks could face compatibility and reliability issues.
- Manufacturing scale‑up is nontrivial. Translating lab yields to factory yields requires months of tooling, process control, and yield improvement—during which costs stay elevated and production is constrained. Corning’s role mitigates but does not eliminate this risk.
- Long‑term field reliability data are limited. DNANF membranes are thin and the field behaviour over decades is not yet proven; conservative lifecycle planning is prudent.
- Supply concentration risk. If Corning and Heraeus prove to be the dominant suppliers initially, Microsoft and other buyers could face supplier scheduling pressure or price volatility until broader industrial capacity emerges.
- Cost parity is not immediate. Near‑term HCF runs will likely cost more per kilometre than SMF; the value proposition depends on selective application to latency‑sensitive links rather than replacing SMF wholesale.
What this means for Azure customers, carriers, and infrastructure planners
- Azure regions with HCF links will gain tangible latency and signal‑fidelity benefits for distributed AI workflows and other latency‑sensitive services. These advantages will be most pronounced where link distance, amplifier count, or nonlinearity previously constrained performance.
- Enterprise and carrier customers should expect an initial period where HCF is available selectively (metro DC interconnects, strategic backbone routes) rather than universally; procurement and architecture teams should model hybrid SMF/HCF fabrics for the next 24–36 months.
- Standards bodies and vendor vendors must accelerate specifications for HCF connectors, splicing, and test methodologies to enable broader multi‑vendor adoption without creating long‑term operational debt.
Conclusion: a pivotal infrastructure inflection with careful engineering ahead
Microsoft’s decision to move from experimental HCF deployments toward industrial production through partnerships with Corning and Heraeus is a major inflection in cloud networking infrastructure. The combination of improved attenuation in new HCF designs, Microsoft’s early deployments and IP, and Corning’s and Heraeus’s manufacturing muscle creates a credible commercialization path for HCF in Azure’s data‑centre fabric. The payoff—lower latency, higher capacity per fibre, and reduced nonlinear penalties—addresses real technical constraints for distributed AI and low‑latency services.That said, the transition remains a complex, multi‑year engineering programme: standards must catch up, mass production yields must improve, operational crews must be trained, and lifecycle data must be gathered. The near‑term reality is strategic, selective deployment where HCF’s advantages are most valuable; the mid‑term prize is rearchitecting certain tiers of the cloud network to be faster and more efficient at scale. Microsoft’s partnerships accelerate that timeline—but widespread adoption across carriers and enterprises will depend on reproducible manufacturing economics, robust interoperability standards, and multi‑year reliability evidence.
Source: Telecompaper Telecompaper