Every link inside an AI cluster is a choice about how much of the light's information you bother to recover. direct-detection intensity-modulated optics simply measure how bright the pulse is, which is cheap, simple and short-reach. coherent optical transceivers recover both the brightness and the phase of the light, which multiplies the data a fiber can carry and how far it travels, at the cost of a tunable laser and a power-hungry digital signal processor. The pitch for coherence is reach and spectral efficiency; the catch is that you pay for a decoder whether or not the link needed one.
Coherent optics recover the light's phase and pack far more bits per wavelength and reach kilometers, but the DSP that does it costs several watts whether the link needed it or not.
Part One: What is actually being compared
These two options are usually presented as alternative implementations of one capability. They are better read as two different answers to the question of where uncertainty is allowed to live. coherent optical transceivers absorbs it in one place, direct-detection intensity-modulated optics in another, and a comparison that measures only the average outcome misses precisely the reason a team would accept the extra complexity of either one in intra-data-center optical interconnect.
There is an accounting asymmetry worth making explicit. coherent optical transceivers tends to put expenditure into capital, where it is spread over years and needs an approval process; direct-detection intensity-modulated optics keeps it in operations, where it is easy to approve and difficult to cap. Finance and engineering usually prefer opposite sides of that line, so the decision is often settled by the budgeting calendar rather than by a measurement in intra-data-center optical interconnect.
Part Two: The arithmetic that decides it
The arithmetic is about spectral efficiency and reach. direct-detection intensity-modulated optics are cheap per link but need more fibers and more lanes to hit the same capacity, and they fade past a few hundred meters of demanding fiber. coherent optical transceivers pack far more bits into each wavelength and tolerate dispersion and mixing over kilometers, so the cost per bit falls as the distance grows. The DSP that makes this possible eats several watts, which is invisible on a long link and oppressive on a rack-to-rack hop.
What actually decides the outcome is the slowest link in the chain, not the fastest. coherent optical transceivers improves one step, but if the surrounding steps still dominate the total, the improvement barely moves the wall clock. This is the recurring disappointment in intra-data-center optical interconnect: a component-level win that is real, reproducible and almost invisible to the user because it was applied to the wrong term in the sum.
Deployment changes what gets measured. After coherent optical transceivers goes in, the metric people optimise tends to shift toward predictability; after direct-detection intensity-modulated optics, toward unit price. That shift is why two teams can look at the same production data and reach opposite conclusions about which option won, and why an evaluation should record the metric before the trial rather than after it.
Part Three: Where the simple model breaks
Distance decides the winner more than capacity does. Inside a rack or across a row, the fibers are short and abundant, so the extra reach of coherence is wasted and its watts are pure overhead. Across a multi-building campus or between clusters, where new fiber is a construction project, recovering the phase turns existing glass into far more capacity. The same technology that is pointless at two meters becomes essential at two kilometers.
Inside a rack the fiber is short and abundant, so the extra reach of coherence is wasted and its power is pure overhead on a link where direct detection already wins.
Part Four: The cost that arrives later
There is a cost and calibration bill that scales with the laser. direct-detection intensity-modulated optics tolerate loose optics because they only read brightness, so yield is high and the parts are cheap. coherent optical transceivers need a narrow, stable local oscillator and a DSP tuned to the fiber's imperfections, and as data rates climb both the laser and the processing get dearer. A design that looks elegant at 400G can become a power and cost nightmare at 1.6T if the distance does not justify it.
Adoption is unlikely to be a single flip. The first movers are the cases where the deciding variable is most extreme, the middle of the market follows once tooling and reference designs exist, and the long tail stays with direct-detection intensity-modulated optics until the cost of switching falls below the cost of maintaining an exception. A reasonable reading of the next several years is coexistence with a slowly moving boundary rather than a replacement event.
Part Five: What this changes outside the engineering team
For the operator who already pulled fiber between buildings, the difference is whether a capacity upgrade means laying more cable or simply upgrading the cards at each end. Coherence lets an old plant carry a new load; direct detection means another tray of glass. The daily-life version is a road that adds lanes by repainting versus one that must be widened with concrete, and repainting only helps once the road is full.
There is a mundane public-facing consequence as well: reliability and predictability are what people actually notice. A household, a commuter or a small shop does not experience an architecture, only whether a screen updates, a bill arrives on time or a service works after two years. Those are the terms in which the eventual judgement will be made.
A fair comparison holds distance and capacity fixed and prices the fiber construction and the DSP power against the spectral efficiency gained, including the failure-domain difference. A choice made at rack scale is usually the wrong one at campus scale.
A market sizing sanity check
If AI clusters keep stretching optics across campuses and existing cable plants are reused, the market for coherent transceivers and their DSPs at data-center reach is plausibly $4B to $13B annually by 2029. The low end assumes direct detection keeps winning short links; the high end assumes coherence becomes default wherever new fiber is expensive. The swing factor is average link distance in new builds.
Treat the numbers as a scale model. They are built to show where the curves cross and which term dominates, and they deliberately round the inputs that vary most between deployments. The ranking is robust to those rounds; the exact threshold is not.
The comparison that matters
The comparison that matters is between buying simpler optics and buying a phase-recovering decoder you may not need. Direct-detection wins where links are short and fiber is cheap; coherent optics win where reach and spectral efficiency make the DSP watts worth paying. Most campuses should go coherent, most racks should stay simple.
A change in regulation, in the supply of a single component, or in the tariff a site faces could invalidate the ranking overnight. The comparison is a snapshot of prices and constraints in intra-data-center optical interconnect, not a property of the technologies, and anyone reusing it in a different year should re-run the threshold rather than reuse the conclusion.
Go coherent where new fiber is a construction project, keep direct detection where glass is cheap, and stop applying the campus answer inside the rack.
SOL: HCqPmPJM3NVtS6YrmVaTZYnuP8XZ2PFgQZKT94VK1LkS
If this saved you time, a tip is appreciated.
Top comments (0)