Data Center MPO Cable: How to Specify 40G and 100G Fiber Trunks
A factory-terminated trunk gets one chance to be right. Once a high-density cabling order has been built, there is no field fix for a wrong polarity method or an interface that does not match the transceiver on the other end.
That is the specification problem behind every data center MPO order, and it only gets harder as project sizes grow. Here is how to settle the interface, the polarity, and the breakout before the trunks are built, not after.
A factory-terminated trunk gets one chance to be right. Once a high-density cabling order has been built, there is no field fix for a wrong polarity method or an interface that does not match the transceiver on the other end.
That is the specification problem behind every data center MPO order, and it only gets harder as project sizes grow. Here is how to settle the interface, the polarity, and the breakout before the trunks are built, not after.
A factory-terminated trunk gets one chance to be right. Once a high-density cabling order has been built, there is no field fix for a wrong polarity method or an interface that does not match the transceiver on the other end.
That is the specification problem behind every data center MPO order, and it only gets harder as project sizes grow. Here is how to settle the interface, the polarity, and the breakout before the trunks are built, not after.
A hundred billion dollars of construction, and one cabling decision
US data center construction spending reached $98.5 billion across 153 projects in the year to date, according to ConstructConnect's October report, published at the end of September. The number that matters more is the average: $1.18 billion per project in 2026, against $507 million in 2025. Projects have roughly doubled in size in a year. A further 65 projects worth $26.1 billion sit in preconstruction with starts targeted before the year ends.
Buildings that size are not cabled the way a comms room is cabled. Field termination at that scale is a labor problem nobody can staff, which is why the structured cabling inside a modern facility arrives factory terminated, as MPO and MTP trunks running between distribution zones, broken out to duplex connections only at the last hop.
That shifts the risk. A field-terminated link is slow to install and easy to fix. A factory-terminated trunk installs in minutes and is almost impossible to fix, because the specification was locked when it was manufactured. Everything therefore depends on getting the order right, and the three things that go wrong are the same three every time: the interface, the polarity, and the breakout.
US data center construction spending reached $98.5 billion across 153 projects in the year to date, according to ConstructConnect's October report, published at the end of September. The number that matters more is the average: $1.18 billion per project in 2026, against $507 million in 2025. Projects have roughly doubled in size in a year. A further 65 projects worth $26.1 billion sit in preconstruction with starts targeted before the year ends.
Buildings that size are not cabled the way a comms room is cabled. Field termination at that scale is a labor problem nobody can staff, which is why the structured cabling inside a modern facility arrives factory terminated, as MPO and MTP trunks running between distribution zones, broken out to duplex connections only at the last hop.
That shifts the risk. A field-terminated link is slow to install and easy to fix. A factory-terminated trunk installs in minutes and is almost impossible to fix, because the specification was locked when it was manufactured. Everything therefore depends on getting the order right, and the three things that go wrong are the same three every time: the interface, the polarity, and the breakout.
US data center construction spending reached $98.5 billion across 153 projects in the year to date, according to ConstructConnect's October report, published at the end of September. The number that matters more is the average: $1.18 billion per project in 2026, against $507 million in 2025. Projects have roughly doubled in size in a year. A further 65 projects worth $26.1 billion sit in preconstruction with starts targeted before the year ends.
Buildings that size are not cabled the way a comms room is cabled. Field termination at that scale is a labor problem nobody can staff, which is why the structured cabling inside a modern facility arrives factory terminated, as MPO and MTP trunks running between distribution zones, broken out to duplex connections only at the last hop.
That shifts the risk. A field-terminated link is slow to install and easy to fix. A factory-terminated trunk installs in minutes and is almost impossible to fix, because the specification was locked when it was manufactured. Everything therefore depends on getting the order right, and the three things that go wrong are the same three every time: the interface, the polarity, and the breakout.
MPO, MTP, and Why the Switch Vendor Does Not Change the Trunk
The first point of confusion is the name. MPO is the multi-fiber connector standard, defined in IEC 61754-7 and TIA-604-5. MTP is one manufacturer's version of an MPO connector, built to tighter tolerances and intermateable with standard MPO. Every MTP is an MPO. Not every MPO is an MTP. If a specification says MTP, it is naming a brand; if it says MPO, it is naming the standard.
The second point of confusion is bigger and costs more. Buyers routinely search for a trunk cable by the brand of switch it will plug into, as though a cable were a vendor accessory. It is not. The trunk is passive glass with connectors on each end, and it is specified by the optical interface the transceiver presents, not by whose logo is on the chassis.
What actually determines the cable is the transceiver. A 40G SR4 module uses 8 of the 12 fibers in an MPO, four transmitting and four receiving, with the middle four unused. A 100G SR4 module does the same at higher lane rates. A 100G SR10 implementation uses 20 fibers across a 24 fiber MPO. Two switches from different manufacturers running the same SR4 optics take the same trunk. One switch running SR4 on one side and a breakout to duplex LC on the other does not.
So the right question is never which brand of cable. It is which transceiver, at which end, across which distance, over which fiber type. Answer those and the trunk specifies itself.
The first point of confusion is the name. MPO is the multi-fiber connector standard, defined in IEC 61754-7 and TIA-604-5. MTP is one manufacturer's version of an MPO connector, built to tighter tolerances and intermateable with standard MPO. Every MTP is an MPO. Not every MPO is an MTP. If a specification says MTP, it is naming a brand; if it says MPO, it is naming the standard.
The second point of confusion is bigger and costs more. Buyers routinely search for a trunk cable by the brand of switch it will plug into, as though a cable were a vendor accessory. It is not. The trunk is passive glass with connectors on each end, and it is specified by the optical interface the transceiver presents, not by whose logo is on the chassis.
What actually determines the cable is the transceiver. A 40G SR4 module uses 8 of the 12 fibers in an MPO, four transmitting and four receiving, with the middle four unused. A 100G SR4 module does the same at higher lane rates. A 100G SR10 implementation uses 20 fibers across a 24 fiber MPO. Two switches from different manufacturers running the same SR4 optics take the same trunk. One switch running SR4 on one side and a breakout to duplex LC on the other does not.
So the right question is never which brand of cable. It is which transceiver, at which end, across which distance, over which fiber type. Answer those and the trunk specifies itself.
The first point of confusion is the name. MPO is the multi-fiber connector standard, defined in IEC 61754-7 and TIA-604-5. MTP is one manufacturer's version of an MPO connector, built to tighter tolerances and intermateable with standard MPO. Every MTP is an MPO. Not every MPO is an MTP. If a specification says MTP, it is naming a brand; if it says MPO, it is naming the standard.
The second point of confusion is bigger and costs more. Buyers routinely search for a trunk cable by the brand of switch it will plug into, as though a cable were a vendor accessory. It is not. The trunk is passive glass with connectors on each end, and it is specified by the optical interface the transceiver presents, not by whose logo is on the chassis.
What actually determines the cable is the transceiver. A 40G SR4 module uses 8 of the 12 fibers in an MPO, four transmitting and four receiving, with the middle four unused. A 100G SR4 module does the same at higher lane rates. A 100G SR10 implementation uses 20 fibers across a 24 fiber MPO. Two switches from different manufacturers running the same SR4 optics take the same trunk. One switch running SR4 on one side and a breakout to duplex LC on the other does not.
So the right question is never which brand of cable. It is which transceiver, at which end, across which distance, over which fiber type. Answer those and the trunk specifies itself.
Polarity: the Thing That Is Wrong When Nothing Works
Polarity is the discipline of making sure every transmit fiber arrives at a receive port. In a duplex LC world it is trivial, because an installer simply swaps two connectors. In an MPO world it is decided at the factory and designed into the whole channel, and getting it wrong means a link that is physically perfect and optically dead.
TIA-568 defines three methods. Type A uses a straight-through trunk with a key-up to key-down alignment, and the flip happens in the patch cord, which means one standard cord and one crossed cord. Type B uses a fully reversed trunk, with fiber one at one end arriving at position twelve at the other, and identical patch cords throughout. Type C pairs the fibers, swapping them in pairs inside the trunk itself, with standard patch cords at both ends.
None of the three is better. What matters is that the trunks, the cassettes and the patch cords all belong to the same method across the whole channel, and that the method is written on the purchase order rather than assumed. Mixing a Type B trunk with Type A cassettes produces a link that tests as broken with no visible cause, usually at commissioning, usually at night, usually on the day the room is meant to go live.
Two further details belong on the same order. Connector gender, because MPO connectors have pins on one side and sockets on the other and two pinned ends will not mate. And polish type, because APC angled polish and UPC flat polish do not connect to each other.
Polarity is the discipline of making sure every transmit fiber arrives at a receive port. In a duplex LC world it is trivial, because an installer simply swaps two connectors. In an MPO world it is decided at the factory and designed into the whole channel, and getting it wrong means a link that is physically perfect and optically dead.
TIA-568 defines three methods. Type A uses a straight-through trunk with a key-up to key-down alignment, and the flip happens in the patch cord, which means one standard cord and one crossed cord. Type B uses a fully reversed trunk, with fiber one at one end arriving at position twelve at the other, and identical patch cords throughout. Type C pairs the fibers, swapping them in pairs inside the trunk itself, with standard patch cords at both ends.
None of the three is better. What matters is that the trunks, the cassettes and the patch cords all belong to the same method across the whole channel, and that the method is written on the purchase order rather than assumed. Mixing a Type B trunk with Type A cassettes produces a link that tests as broken with no visible cause, usually at commissioning, usually at night, usually on the day the room is meant to go live.
Two further details belong on the same order. Connector gender, because MPO connectors have pins on one side and sockets on the other and two pinned ends will not mate. And polish type, because APC angled polish and UPC flat polish do not connect to each other.
Polarity is the discipline of making sure every transmit fiber arrives at a receive port. In a duplex LC world it is trivial, because an installer simply swaps two connectors. In an MPO world it is decided at the factory and designed into the whole channel, and getting it wrong means a link that is physically perfect and optically dead.
TIA-568 defines three methods. Type A uses a straight-through trunk with a key-up to key-down alignment, and the flip happens in the patch cord, which means one standard cord and one crossed cord. Type B uses a fully reversed trunk, with fiber one at one end arriving at position twelve at the other, and identical patch cords throughout. Type C pairs the fibers, swapping them in pairs inside the trunk itself, with standard patch cords at both ends.
None of the three is better. What matters is that the trunks, the cassettes and the patch cords all belong to the same method across the whole channel, and that the method is written on the purchase order rather than assumed. Mixing a Type B trunk with Type A cassettes produces a link that tests as broken with no visible cause, usually at commissioning, usually at night, usually on the day the room is meant to go live.
Two further details belong on the same order. Connector gender, because MPO connectors have pins on one side and sockets on the other and two pinned ends will not mate. And polish type, because APC angled polish and UPC flat polish do not connect to each other.
Counts, Breakout and How to Choose
With the interface and the polarity settled, the remaining decisions are about shape.
Trunk fiber count follows the zone. Between the main distribution area and the horizontal zones, counts of 144 to 864 are normal, landing on 1U rack panels that hold around 144 fibers each, which is what makes a high density row physically possible. Within a row, 12 and 24 fiber trunks are the working unit. The underlying standard for all of it is TIA-568.2-D.
Fiber type follows distance. OS2 singlemode carries the long runs between buildings and across a campus. OM4 multimode remains common on short reaches inside a row, where the optics are cheaper. The cost crossover between the two has been moving toward singlemode for several years as speeds rise.
Breakout follows the equipment. A harness assembly takes one MPO at the trunk end and fans out to duplex LC at the other, which is how a trunk feeds equipment that has no MPO port. Cassettes do the same job inside a panel and are easier to rearrange later. Harnesses are cheaper and tidier for a fixed layout; cassettes cost more and survive change better.
Where duct or pathway is the constraint rather than rack space, rollable ribbon is increasingly what feeds these trunks from outside the building, because it carries far more fiber in the same diameter.
The mistakes repeat across every project. Ordering by switch brand instead of by transceiver type. Leaving polarity method off the purchase order and discovering it at commissioning. Mixing genders. Specifying trunk length to the drawing rather than to the real cable route, which is always longer. And ordering the trunks on the schedule while the breakouts and panels go out for quote later, which is how a room ends up with cable it cannot terminate.
With the interface and the polarity settled, the remaining decisions are about shape.
Trunk fiber count follows the zone. Between the main distribution area and the horizontal zones, counts of 144 to 864 are normal, landing on 1U rack panels that hold around 144 fibers each, which is what makes a high density row physically possible. Within a row, 12 and 24 fiber trunks are the working unit. The underlying standard for all of it is TIA-568.2-D.
Fiber type follows distance. OS2 singlemode carries the long runs between buildings and across a campus. OM4 multimode remains common on short reaches inside a row, where the optics are cheaper. The cost crossover between the two has been moving toward singlemode for several years as speeds rise.
Breakout follows the equipment. A harness assembly takes one MPO at the trunk end and fans out to duplex LC at the other, which is how a trunk feeds equipment that has no MPO port. Cassettes do the same job inside a panel and are easier to rearrange later. Harnesses are cheaper and tidier for a fixed layout; cassettes cost more and survive change better.
Where duct or pathway is the constraint rather than rack space, rollable ribbon is increasingly what feeds these trunks from outside the building, because it carries far more fiber in the same diameter.
The mistakes repeat across every project. Ordering by switch brand instead of by transceiver type. Leaving polarity method off the purchase order and discovering it at commissioning. Mixing genders. Specifying trunk length to the drawing rather than to the real cable route, which is always longer. And ordering the trunks on the schedule while the breakouts and panels go out for quote later, which is how a room ends up with cable it cannot terminate.
With the interface and the polarity settled, the remaining decisions are about shape.
Trunk fiber count follows the zone. Between the main distribution area and the horizontal zones, counts of 144 to 864 are normal, landing on 1U rack panels that hold around 144 fibers each, which is what makes a high density row physically possible. Within a row, 12 and 24 fiber trunks are the working unit. The underlying standard for all of it is TIA-568.2-D.
Fiber type follows distance. OS2 singlemode carries the long runs between buildings and across a campus. OM4 multimode remains common on short reaches inside a row, where the optics are cheaper. The cost crossover between the two has been moving toward singlemode for several years as speeds rise.
Breakout follows the equipment. A harness assembly takes one MPO at the trunk end and fans out to duplex LC at the other, which is how a trunk feeds equipment that has no MPO port. Cassettes do the same job inside a panel and are easier to rearrange later. Harnesses are cheaper and tidier for a fixed layout; cassettes cost more and survive change better.
Where duct or pathway is the constraint rather than rack space, rollable ribbon is increasingly what feeds these trunks from outside the building, because it carries far more fiber in the same diameter.
The mistakes repeat across every project. Ordering by switch brand instead of by transceiver type. Leaving polarity method off the purchase order and discovering it at commissioning. Mixing genders. Specifying trunk length to the drawing rather than to the real cable route, which is always longer. And ordering the trunks on the schedule while the breakouts and panels go out for quote later, which is how a room ends up with cable it cannot terminate.
What to Settle Before the Order Is Placed
What to Settle Before the Order Is Placed
What to Settle Before the Order Is Placed
The whole specification fits on one page: transceiver type at each end, fiber type, trunk count, polarity method, gender, polish, length as routed, and the breakout format. Settled together, a factory-terminated channel installs in an afternoon. Settled separately, it becomes a commissioning problem.
Vocom International supplies that channel rather than manufacturing it. Its data center and indoor range covers factory-terminated MPO and MTP trunks from 12 to 864 fibers in Type A, B and C polarity, harness breakouts to duplex LC, modular cassettes and adapters, and OS2, OM3 and OM4 patch for the final rack hop, sourced build-to-order through tier 1 manufacturing partners on Fujikura preform glass. Standard lead time is 6 to 8 weeks, expedited production is available, and custom builds run to 10 weeks.
If a room is going live in the first half of 2027, the trunk schedule is worth pricing now, while the polarity method is still a decision rather than a discovery. Talk to a specifier.
The whole specification fits on one page: transceiver type at each end, fiber type, trunk count, polarity method, gender, polish, length as routed, and the breakout format. Settled together, a factory-terminated channel installs in an afternoon. Settled separately, it becomes a commissioning problem.
Vocom International supplies that channel rather than manufacturing it. Its data center and indoor range covers factory-terminated MPO and MTP trunks from 12 to 864 fibers in Type A, B and C polarity, harness breakouts to duplex LC, modular cassettes and adapters, and OS2, OM3 and OM4 patch for the final rack hop, sourced build-to-order through tier 1 manufacturing partners on Fujikura preform glass. Standard lead time is 6 to 8 weeks, expedited production is available, and custom builds run to 10 weeks.
If a room is going live in the first half of 2027, the trunk schedule is worth pricing now, while the polarity method is still a decision rather than a discovery. Talk to a specifier.
The whole specification fits on one page: transceiver type at each end, fiber type, trunk count, polarity method, gender, polish, length as routed, and the breakout format. Settled together, a factory-terminated channel installs in an afternoon. Settled separately, it becomes a commissioning problem.
Vocom International supplies that channel rather than manufacturing it. Its data center and indoor range covers factory-terminated MPO and MTP trunks from 12 to 864 fibers in Type A, B and C polarity, harness breakouts to duplex LC, modular cassettes and adapters, and OS2, OM3 and OM4 patch for the final rack hop, sourced build-to-order through tier 1 manufacturing partners on Fujikura preform glass. Standard lead time is 6 to 8 weeks, expedited production is available, and custom builds run to 10 weeks.
If a room is going live in the first half of 2027, the trunk schedule is worth pricing now, while the polarity method is still a decision rather than a discovery. Talk to a specifier.
