50 & 100G PON
Isn't 1G enough? What about 10G? Those who shop in warehouse clubs can confirm that the more you have of something, the more you will use. The Internet is no exception. But it isn't just the speed. Lower-latency, more reliable connections, and overall lower prices are also key goals to upcoming Passive Optical Networks (PON) improvements.
Broadband Internet access might be taken for granted by those who have access to it, but not everyone has access. Those who do have access to broadband, still might not have access to 1Gbps, but perhaps speeds between 10-100Mbps. And within these categories, many might indeed "have access", but find the pricing to be unaffordable or not a good enough value.
The chart at the left comes from the Pew Research Center, one of our favorite sources of research. It's easy to spend a few hours looking at the research provided by Pew, but we recommend taking even just a quick further look at their Internet, Broadband Fact Sheet.
This particular chart shows the overall use of Broadband in the United States from the year 2000 to June of 2025. In the ongoing surveys conducted by Pew, broadband was not defined as an explicit speed or data rate, but was generally referred to in the survey as use of the Internet via smartphones and desktop computers, implying higher connection speeds. As the chart shows, up to around 80% of the United States uses broadband of some kind, but this means that 20% of the country does not. Interestingly, the trend has generally flattened after 2015, and has even started to drop a few percentage points since 2023.
For reasons like this, it may be surprising to hear that 50G and even 100G Passive Optical Networks (PON) are not only being tested, but in small areas are already available, at least at interim speeds that can eventually be enabled higher (e.g. 25G PON and 50G PON). As with most technology, lessons are learned from ongoing improvements, which tend to make technology less expensive over time. Developing 50G networks, well before they appear to be needed, will almost certainly contribute to lower prices for all PON networks, making it easier to achieve 1G to the prem for a greater share of the population.
Reliability
It isn't entirely just about speed. The International Telecommunications Union (ITU) tries to keep overall reliability high, and where possible, compatible with existing GPON and XGS-PON networks, which helps reduce costs for service providers and customers alike. But these things aren't so easy. As speed increases from 10G to 50G, for example, bit-error-rates tend to worsen, and recoverable signals naturally drop. The ITU 9804.3 spec (currently the latest high-speed PON specification for 50G) requires Optical Distribution Networks (ODN) to tolerate path losses between 29 and 32dB, and lengths of up to 40km, which is no small feat, as 50G speeds impose an approximate 5dB penalty over 10G speeds for a given distance.
PON protection schemes, which are not currently common due to their expense, may also become more important as users come to depend on much higher speeds. The same 9804.3 specification describes PON protection Type B (protects from the central office Optical Line Terminal, OLT), Type C (protects from the OLT, and includes plant like outside splitters), and a new Type W (which only provides OLT level protection). Other factors like laser drift start to matter, when multiple wavelengths are in use at sub-CWDM spacing, typically requiring lasers to remain within ±2nm of their center frequency, adding cost to both the OLT and ONU.
Coexistence
There are also new categories of Coexistence Elements, numbered 1-3 (with more likely coming as they are needed). This is necessary due to the different migration paths from XGS-PON and 10G-EPON, to 25G or 50G TDM PON. New wavelengths like 1342nm (50G) or 1358nm (25G) downstream, and 1286nm upstream, may or may not have to coexist with 1270nm, and others. True, triple co-existence is available (using CE3), but this may not be needed in all cases. Vendors offering 50G equipment also provide a choice of 1300nm up, or 1270nm up, depending upon whether co-existence is desired with just GPON (CE1), or just XGS-PON (CE2). These types of details are worked-out by each vendor and customer at deal time, but it shows how wavelength-agnostic equipment is becoming a key feature for modern PON equipment. This also means that customer Optical Network Units (ONU) must support broadband, colorless channels, which until recently was a costly and sometimes difficult challenge for vendors to manufacture consistently.
Technical Challenges
50G and 100G might sound like "easy doubles" from 25G, but the challenges start to mount. Problems with distortion, loss budgets, and affordability present real challenges that need work before large commercial adoption, although most of the technical issues are now resolved. It's more an issue of cost. For 100G in particular, the customer ONU will likely need to support Coherent Optics, which in essence requires digital signal processing to both recover and transmit a very high speed signal over distance. For example, laser light can be modulated at different power levels, (not just at different on-off speeds), increasing data throughput. However, this means the recoverable light signal must always remain above the optical noise floor at two or more different light levels, for billions of flashes per second. Over 40km. With no errors. Cheaply.
Shorter distances help resolve these problems, but that would mean improvements to outside plant, with either improved fiber, or adding more optical amplifiers, or buying improved equipment for the OLT and ONU ends. Fiber is often being replaced or added for new networks, as modern fiber has better overall attenuation, zero water peak, and for the latest G.657-D category, is so bend-insensitive that it can wrap around a pencil and not break. But all of these things cost money.
What's Out There
The race is on for very-high speed PON. Most major equipment vendors can already provide off-the-shelf 25G solutions, and in many cases 50G, or a path to get there quickly without significant equipment or fiber upgrades. While not fully commercial as of this writing, 100G has been conducted in multiple trials, and it works well. Mostly all that remains is getting OLT/ONU costs down.
One popular example is Nokia's Lightspan MF-series optical line terminal (OLT) (MF-14 pictured below), which supports everyday speeds up to 25G, but makes use of many of the above technologies, allowing for easy turn-on of 50G and in the not-so-distant future, even 100G. Nokia has said the Lightspan MF is intended to eventually reach 100G, presumably without needing to perform significant upgrades. This is important to service providers who have had to make significant upgrades over the years from the original BPON to GPON to XGS-PON, with each requiring significant new hardware upgrades along the way.
100G trials have already been conducted with Nokia's Lightspan MF-series platform with operators like Australia's NBN Co, showing the technology for 100G is here right now, if not quite ready for widespread commercial release.
It isn't entirely just about speed. The International Telecommunications Union (ITU) tries to keep overall reliability high, and where possible, compatible with existing GPON and XGS-PON networks, which helps reduce costs for service providers and customers alike. But these things aren't so easy. As speed increases from 10G to 50G, for example, bit-error-rates tend to worsen, and recoverable signals naturally drop. The ITU 9804.3 spec (currently the latest high-speed PON specification for 50G) requires Optical Distribution Networks (ODN) to tolerate path losses between 29 and 32dB, and lengths of up to 40km, which is no small feat, as 50G speeds impose an approximate 5dB penalty over 10G speeds for a given distance.
PON protection schemes, which are not currently common due to their expense, may also become more important as users come to depend on much higher speeds. The same 9804.3 specification describes PON protection Type B (protects from the central office Optical Line Terminal, OLT), Type C (protects from the OLT, and includes plant like outside splitters), and a new Type W (which only provides OLT level protection). Other factors like laser drift start to matter, when multiple wavelengths are in use at sub-CWDM spacing, typically requiring lasers to remain within ±2nm of their center frequency, adding cost to both the OLT and ONU.
Coexistence
There are also new categories of Coexistence Elements, numbered 1-3 (with more likely coming as they are needed). This is necessary due to the different migration paths from XGS-PON and 10G-EPON, to 25G or 50G TDM PON. New wavelengths like 1342nm (50G) or 1358nm (25G) downstream, and 1286nm upstream, may or may not have to coexist with 1270nm, and others. True, triple co-existence is available (using CE3), but this may not be needed in all cases. Vendors offering 50G equipment also provide a choice of 1300nm up, or 1270nm up, depending upon whether co-existence is desired with just GPON (CE1), or just XGS-PON (CE2). These types of details are worked-out by each vendor and customer at deal time, but it shows how wavelength-agnostic equipment is becoming a key feature for modern PON equipment. This also means that customer Optical Network Units (ONU) must support broadband, colorless channels, which until recently was a costly and sometimes difficult challenge for vendors to manufacture consistently.
Technical Challenges
50G and 100G might sound like "easy doubles" from 25G, but the challenges start to mount. Problems with distortion, loss budgets, and affordability present real challenges that need work before large commercial adoption, although most of the technical issues are now resolved. It's more an issue of cost. For 100G in particular, the customer ONU will likely need to support Coherent Optics, which in essence requires digital signal processing to both recover and transmit a very high speed signal over distance. For example, laser light can be modulated at different power levels, (not just at different on-off speeds), increasing data throughput. However, this means the recoverable light signal must always remain above the optical noise floor at two or more different light levels, for billions of flashes per second. Over 40km. With no errors. Cheaply.
Shorter distances help resolve these problems, but that would mean improvements to outside plant, with either improved fiber, or adding more optical amplifiers, or buying improved equipment for the OLT and ONU ends. Fiber is often being replaced or added for new networks, as modern fiber has better overall attenuation, zero water peak, and for the latest G.657-D category, is so bend-insensitive that it can wrap around a pencil and not break. But all of these things cost money.
What's Out There
The race is on for very-high speed PON. Most major equipment vendors can already provide off-the-shelf 25G solutions, and in many cases 50G, or a path to get there quickly without significant equipment or fiber upgrades. While not fully commercial as of this writing, 100G has been conducted in multiple trials, and it works well. Mostly all that remains is getting OLT/ONU costs down.
One popular example is Nokia's Lightspan MF-series optical line terminal (OLT) (MF-14 pictured below), which supports everyday speeds up to 25G, but makes use of many of the above technologies, allowing for easy turn-on of 50G and in the not-so-distant future, even 100G. Nokia has said the Lightspan MF is intended to eventually reach 100G, presumably without needing to perform significant upgrades. This is important to service providers who have had to make significant upgrades over the years from the original BPON to GPON to XGS-PON, with each requiring significant new hardware upgrades along the way.
100G trials have already been conducted with Nokia's Lightspan MF-series platform with operators like Australia's NBN Co, showing the technology for 100G is here right now, if not quite ready for widespread commercial release.
Nokia Lightspan MF-14 OLT (top left), Nokia ONU with 25G capability today (top right), and Adtran SDX 6400 OLT (bottom right).
Adtran's SDX 6400-series OLT operates up to 50G PON today, all in a small 1 rack-unit chassis. The system also supports XGS-PON and GPON, allowing operators to purchase SDX equipment today, but deploy PON networks at different speeds, making upgrades easy as new plant is added and campaigns are rolled-out to customers whenever the provider is ready.
The Calix Broadband Platform (announced late October 2025, not pictured) brings AI from Vertex AI and Google Gemini to make the product easy to use for both broadband service providers and end users alike. Calix is advertising 50G speeds, which is hardly surprising as the vendor was one of the leaders in deploying XGS-PON and NG-PON2 to its popular platforms like the E7 and E9.
What to Deploy Now
That's a good question. It depends a lot on what technology has been deployed already, the take-rate, customer expectations, what competitors may be doing, and a few other factors. Frankly, there's also a subjective factor. Some believe that 10G to 25G is too short a step, and that 50G should be the immediate goal. Others feel a more gradual approach is sufficient, and perhaps even warranted, since so few people currently have access to 1G and 10G. There's also the question of how much bandwidth is truly needed by a typical residential user or small-office/home-office, the typical end users of PON, and whether they will stay after their introductory price ends. Most customers today would certainly like to have high connection speeds, but as the PEW graph showed above, growth at the national level is no longer assured.
Ciena is a relative newcomer to access-facing PON networks, although the company is well known for their premium optical transport and add-drop multiplexer solutions. Ciena studied the cost to move from existing 10G networks to faster 25G and 50G networks (see chart below). There was a relatively low increase in cost from 10G to 25G, around 40% more than just the cost of deploying 10G. Considering the extra revenue potential of 25G (with its higher PON data rate, and potentially more splits to reach more customers), 25G seems compelling. It would not be unreasonable to suppose a 25G PON network could generate at least double the revenue of a 10G network, and possibly quite a bit more.
Getting to 50G is a significant jump, requiring digital signal processing (DSP), Service-Oriented Architectures (SoA), Electro-absorption Modulated Lasers (EML), and Intensity Modulation with Direct Detection (IMDD) transmitters (!). These are all examples of the special technologies needed to make 50G work. However, suppose one had a sufficient market size for 50G PON — the jump to 100G would then be relatively cheap. For most service providers, the largest cost factor in PON deployments is typically the fiber plant itself, with the OLT and ONU equipment coming in at a distant second. If a provider was looking to add service to a high-density environment like apartment buildings, a 50G solution (with a view to an eventual 100G offering) might make a great deal of sense rather than 25G, because the most appropriate fiber could be deployed today, with the future speeds in mind. This may not be as attractive for suburban or rural areas, where 25G and even 10G might still suffice for the next few years.
Ciena's ONU Cost for Higher BW PON comparison, from their "Breaking Barriers: The path to access modernization from 25GS-PON to Coherent PON" blog.
Copyright, AG Advice and Support, 2026.