The evolution of multimode fiber has been a cornerstone in the development of modern telecommunications and data center networking. Understanding this journey requires a look back at the early days of optical fiber technology. In the 1970s and 1980s, the first generation of multimode fiber, often referred to as OM1, was introduced. With a core diameter of 62.5 micrometers, OM1 was designed to work with LED light sources for short-distance applications. It served well in the nascent LAN environments of the time, but its bandwidth was limited, capping out at around 200 MHz·km. The 1990s saw the introduction of OM2 fiber, which featured a 50-micrometer core. This smaller core allowed for higher bandwidth—up to 500 MHz·km—and supported slightly longer distances, but the real breakthrough came with the advent of laser-optimized multimode fiber. The demand for higher speeds in enterprise networks and early data centers drove the creation of OM3 fiber, standardized by the ISO/IEC 11801 and TIA-568-C.3 standards. OM3 fiber, with its 50-micrometer core, was specifically designed to be used with VCSEL (Vertical-Cavity Surface-Emitting Laser) sources, offering an Effective Modal Bandwidth (EMB) of 2000 MHz·km. This represented a massive leap forward, enabling 10 Gigabit Ethernet (10GBASE-SR) to run up to 300 meters. The current state of OM3 fiber is one of widespread adoption. It has become the de facto standard for campus backbones, storage area networks (SAN), and horizontal cabling in many mid-sized enterprises. Its reliability and cost-effectiveness have made it a staple. However, as network demands grow, the limitations of OM3 fiber are becoming more apparent, particularly in hyperscale data centers and for emerging 400G applications. The physical infrastructure of a typical Hong Kong data center, such as those in the Tseung Kwan O Industrial Estate or the Cyberport, often relies on OM3 fiber for inter-rack connectivity. A single `fibre optic cable` run of OM3 can comfortably handle 10G and 25G traffic, but when operators look to scale to 50G or 100G per lane—as is required for 400G—the reach of OM3 becomes a critical limiting factor. For example, a standard 100GBASE-SR10 implementation over OM3 is limited to just 100 meters. In a large Hong Kong data center, where distances between network cabinets can easily exceed 150 meters due to cable management pathways and physical building layouts, this constraint forces designers to either use more expensive single-mode fiber solutions or to plan for a future upgrade to OM4 or OM5. The ubiquitous presence of OM3 in the installed base is a double-edged sword: it provides a known, reliable `` for legacy 1G and 10G connections, but it is now the primary barrier to cost-effective bandwidth scaling.OM3 fiber today is the workhorse of many existing enterprise and data center networks, but its specific applications and performance characteristics define its role in the current infrastructure landscape. The most common application for OM3 fiber remains the support of 10 Gigabit Ethernet over distances of up to 300 meters and 40GBASE-SR4 over 100 meters. In Hong Kong's financial district, for instance, many of the older trading floors and back-office systems in Central and Admiralty are interconnected using OM3 fiber. A typical `` in a server room might be a 10GBASE-SR SFP+ module plugged into a switch, with the corresponding `fibre optic cable` being an OM3 duplex LC connector patch cord. The performance characteristics of OM3 are well-understood: it has a modal bandwidth of 2000 MHz·km, an attenuation of 3.5 dB/km at 850 nm, and a numerical aperture of 0.2. These specs make it an excellent choice for building-to-building campus links that are under 300 meters. For enterprise storage connectivity, OM3 is used extensively in Fibre Channel SANs, supporting 16GFC over 100 meters and 32GFC over 70 meters. However, the limitations of OM3 are becoming increasingly problematic. The most significant limitation is its reach at higher data rates. As the industry moves from 10G per lane to 25G per lane for 100G and 400G optics, the modal dispersion in OM3 limits the effective distance. For example, the 100GBASE-SR4 standard, which uses four lanes of 25G each, can only reach 70 meters over OM3, whereas over OM4 it reaches 100 meters, and over OM5 it reaches 150 meters. This 70-meter reach is often insufficient in modern data center designs that utilize top-of-rack (ToR) switches and long cable runs to middle-of-row (MoR) or end-of-row (EoR) switches. In a typical Hong Kong colocation facility, the physical cabling distance from a ToR switch to an EoR switch can easily exceed 70 meters when you account for vertical riser pathways and horizontal cable trays. A direct consequence is that network architects must carefully manage their floor plans or prematurely upgrade to OM4, negating the cost savings of using existing OM3. Another limitation is future-proofing. While OM3 can support 40G and 100G, its performance is marginal for these standards. The newer SWDM (Shortwave Wavelength Division Multiplexing) technology, which allows multiple wavelengths over a single fiber, is less effective over OM3 than over OM4 or OM5. For instance, SWDM4 (which uses 4 wavelengths) can support 100G over 100 meters on OM4, but only about 75 meters on OM3. This forces network operators to leave less room for growth. Furthermore, the market in Hong Kong, where space is at a premium in data centers, requires high-density cabling. The limitations of OM3 mean that to achieve the same port count over a given distance, you might need more fiber strands or more expensive optics, which directly impacts the total cost of ownership. Therefore, while OM3 remains a competent and cost-effective solution for current 10G and 25G networks, its limitations in reach and bandwidth density make it a strategic bottleneck for the next wave of network scaling.The emergence of OM4 and OM5 fibers has addressed the limitations of OM3, providing a clear performance ladder for network architects. At the core of the comparison is bandwidth and distance. OM3 fiber offers an EMB of 2000 MHz·km, while OM4 provides 4700 MHz·km, and OM5 provides an EMB of 4700 MHz·km as well, but with the added capability of supporting four wavelengths in the 850-950 nm range. This difference translates directly to reach. For 100GBASE-SR4, OM3 supports 70 meters, OM4 supports 100 meters, and OM5 supports 100 meters for the same standard, but OM5 shines for SWDM. The advantages of newer multimode fibers, particularly OM5, are substantial for modern data centers. OM5, also known as WBMMF (Wide Band Multimode Fiber), is the first multimode fiber designed to explicitly support multiple shortwave wavelengths. This is a game-changer because it allows a single `fibre optic cable` pair to carry multiple data channels using SWDM technology, reducing the number of fiber strands required for high-speed links. For example, instead of needing 8 fibers (4 pairs) for 100GBASE-SR4, a single pair of OM5 fibers can carry 100G using SWDM4, which uses four different wavelengths (e.g., 850, 880, 910, and 940 nm). This drastically simplifies cable management and reduces costs in high-density environments. In a typical `` scenario in a hyperscale data center in Hong Kong, such as those operated by HKColo or PCCW, the shift from OM3 to OM5 is driven by the need for 400G connectivity. The 400GBASE-SR8 standard can run over OM5 up to 100 meters, while OM3 struggles to reach 70 meters for an equivalent 400GBASE-SR16 (which is now considered obsolete). The advantages of OM4 and OM5 are not just about raw distance; they are about flexibility. An OM5 fiber can support 10G, 25G, 50G, 100G, 200G, and 400G over various distances, meaning an operator can use the same `fibre optic cable` for multiple generations of technology simply by changing the pluggable optics at the ``. This is in stark contrast to OM3, which may require a new fiber installation when upgrading from 10G to 100G over a certain distance. In Hong Kong, where real estate costs for data centers are astronomical, the ability to maximize bandwidth per square foot using OM5's wavelength division multiplexing is a critical financial advantage. A table below illustrates the distance differences for common Ethernet speeds:| Standard | OM3 (m) | OM4 (m) | OM5 (m) ||---|---|---|---|| 100GBASE-SR4 | 70 | 100 | 100 || 100G SWDM4 | 75 | 100 | 150 || 400GBASE-SR8 | 70 | 100 | 100 || 400G SWDM4 (4 lanes 50G) | 75 | 100 | 150 |The key takeaway is that OM5 effectively doubles the reach for SWDM applications compared to OM4 in some cases, and significantly extends the distance over OM3. This makes OM5 the most future-proof choice for any greenfield data center installation, especially in high-value markets like Hong Kong.Technological advancements are continuously reshaping the potential of multimode fiber, moving it beyond its traditional role as a simple short-reach medium. One of the most significant innovations is Shortwave Wavelength Division Multiplexing (SWDM). Traditionally, multimode fiber relied on a single wavelength at 850 nm. SWDM, defined by the TIA-492AAAD standard, uses multiple wavelengths in the 850-950 nm band, typically four wavelengths spaced 30 nm apart. This allows a single `fibre optic cable` pair to carry multiple independent data streams. The development of SWDM transceivers is directly tied to the performance of OM5 fiber. For instance, a 40G SWDM transceiver can use four 10G lanes over four wavelengths on a single OM5 fiber pair. This technology is critical for data centers that are space-constrained. In a Hong Kong data center, where every rack unit and every tray of cables translates directly into cost savings, the ability to reduce the number of `fibre optic cable` strands by a factor of four is revolutionary. The next frontier is multimode fiber for 400G and beyond. The IEEE 802.3bs standard defines 400GBASE-SR8, which uses eight lanes of 50G PAM4 modulation over eight fibers (four pairs). This works over OM3, OM4, and OM5, but with different reach limitations. However, the real breakthrough for 400G on multimode is the 400GBASE-SR4.2 standard for bidirectional transmission, which uses two wavelengths (850 nm and 910 nm) over a single fiber pair to achieve 100G per lane for a total of 200G, combined with four fiber pairs for 400G. This is where OM5's wideband capability is essential, as it supports these dual-wavelength optics more effectively. Looking further ahead, 800G and 1.6T Ethernet standards are being discussed, and multimode fiber is likely to play a role using a combination of increased lane count and SWDM. The industry is currently exploring the limits of VCSEL (Vertical-Cavity Surface-Emitting Laser) technology. Improvements in VCSEL design include better temperature stability, higher modulation bandwidth, and lower relative intensity noise (RIN). New VCSELs capable of operating at 50 Gbaud and 100 Gbaud for PAM4 are being developed, which directly benefits multimode fiber. For example, a 100 Gbps VCSEL operating at 850 nm can drive a single lane of 100G over OM5. These laser improvements allow for higher data rates per wavelength, which in turn reduces the number of fibers required for a given port speed. Furthermore, the development of single-mode VCSELs for long-reach applications is also on the horizon, but for the short-reach data center market (under 500 meters), multimode VCSELs remain the most cost-effective optical source, especially when paired with OM5 fiber. This synergy between SWDM, high-speed VCSELs, and wideband fiber is creating a new lease on life for multimode solutions in the era of 400G, 800G, and beyond. An `` in a future data center might not even be labeled as "mpo or single-mode", but rather as a "VM" slot for "Versatile Multimode", capable of accepting any SWDM optic from 10G to 400G.Despite the advancements in newer fibers, OM3 fiber retains a significant role in legacy systems, primarily due to the immense capital investment already sunk into existing infrastructure. Compatibility is the core of OM3's continued relevance. IT managers in Hong Kong, for example, who have spent millions of dollars cabling their office towers in Wan Chai or their manufacturing facilities in Kwai Tsing with OM3 fiber, cannot simply rip and replace it. They must ensure that any new equipment is backwards-compatible with their existing OM3 plant. The good news is that most modern transceivers, especially 10G and 25G SFP28 modules, are designed to work with OM3 fiber. A 25GBASE-SR SFP28 module will operate perfectly over an OM3 `fibre optic cable` up to 70 meters. Similarly, 100G QSFP28 SR4 modules will work over OM3, although at a reduced distance of 70 meters instead of 100 meters. In a legacy system scenario, this compatibility allows for a gradual upgrade path. For instance, a company can upgrade its core switches to 100G, using OM3 fiber for the short links between core switches, while keeping 10G to the server over the same OM3 plant. This hybrid approach is common. However, the limitations become apparent when trying to reach longer distances or higher speeds. This is where migration strategies come into play. One practical strategy is "stranded cable" management. Many legacy OM3 installations were deployed with multiple spare fiber strands. An operator can use two strands (a duplex pair) from an existing 12-fiber MPO trunk to create a dedicated 100G link using SWDM4 optics, provided the distance is under 75 meters. This avoids pulling new cable. Another strategy is to use media converters or transceivers that offer adaptive equalization. Some advanced transceivers can partially compensate for the higher modal dispersion of OM3, extending its reach for 100G applications by a few meters. This is not a silver bullet, but it can bridge a critical gap. The most common migration strategy, however, is the "forklift upgrade"—but done in stages. In a Hong Kong data center, the typical plan is to first upgrade the backbone links between core and distribution switches to OM4 or OM5, leaving the OM3 to serve the access layer. Then, as access switches are refreshed, they are connected with new OM5 trunk cables. The legacy OM3 cable is often left in place, either as a spare or decommissioned slowly. The `` metaphor is apt here: OM3 serves as a legacy on a power strip—it might not support the latest high-power device, but it is perfectly fine for plugging in a low-power lamp. In practical networking terms, that lamp is a legacy 1G or 10G server or storage device. To efficiently manage this transition, network operators often create a detailed cable management database. A simple list of common migration actions is shown below:- Action: Use SWDM optics for 100G on existing OM3 (Reach: ~75m).- Action: BiDi (Bi-Directional) optics to use two fibers for a single link (effective for 40G and 100G).- Action: Parallel optics (SR4) for 40G/100G on OM3 (Reach: 100m for 40G).- Action: Replace specific high-traffic trunk cables with OM5 while keeping OM3 for low-traffic links.The key is that OM3 is not a dead technology; it is an enabler for a smooth, capital-efficient transition.Looking forward, the future of multimode fiber, and OM3's place within it, will be shaped by market trends, adoption rates, and the evolving architectural needs of data centers and enterprise networks. The market trends for multimode fiber in 2024-2028 are clear: a steady decline in OM3 procurement for new installations, a plateauing of OM4, and a strong growth curve for OM5, particularly in hyperscale and colocation data centers. In the Hong Kong market, which is a major financial and connectivity hub in Asia, the adoption rate of OM5 is accelerating. According to industry reports, the number of new data center builds in Hong Kong (e.g., the new NTT/HKIX data center in Tseung Kwan O) are specifying OM5 as their minimum standard for inter-rack connectivity. The drive is purely economic: the cost of OM5 fiber is only marginally higher than OM4 (often less than 10% premium), but it offers significantly higher bandwidth potential per fiber pair. The evolving role of multimode fiber in data centers is shifting from a general-purpose medium to a specialized short-reach, high-density solution. For distances under 150 meters, OM5 is becoming the default. For distances above 150 meters, single-mode fiber with DWDM (Dense Wavelength Division Multiplexing) is taking over. This bifurcation is healthy for the industry. Multimode fiber is no longer trying to compete with single-mode for long-reach; it is winning on cost-efficiency and density for the crucial inter-switch and top-of-rack to server links within a single data hall. The `fibre optic cable` of the future for data centers is likely a hybrid solution: OM5 for inside the row, and single-mode for between rows and buildings. The `extension socket` of the future will be a high-density MPO-16 or CS connector with OM5 fiber, capable of handling 400G or 800G with SWDM8 optics. For enterprise networks, the story is a bit different. Many enterprises in Hong Kong are still running 1G to the desktop and 10G in the backbone, for which OM3 is perfectly adequate. However, as video conferencing, AI/ML workloads, and cloud computing demands grow, even enterprises will need to upgrade. The prediction for enterprise adoption is that they will skip OM4 entirely and move directly from OM3 to OM5. This is because the price difference is small, and the future-proofing is significant. A major Hong Kong university, for example, might upgrade its campus backbone from OM3 to OM5 over a 3-5 year plan, reusing the OM3 in low-traffic administrative buildings. The ultimate fate of OM3 is to become a legacy technology, much like copper Cat5e cable is today. It will continue to operate reliably for years, supporting older equipment and serving networks where speeds are not growing. But the future of new investment is clearly in OM5, and the innovation is focused on making multimode fiber an even more powerful tool for the emerging era of hyperscale computing, AI, and 400G/800G Ethernet. The transition is already underway, and the infrastructure decisions made in Hong Kong today will define the bandwidth capabilities of its digital economy for the next decade. The hybrid reality of OM3 in legacy, OM4 in steady state, and OM5 in the vanguard is the most likely scenario for the next five years.