Integrated photonics narrows the gap between laboratory result and manufacturable part
Foundry access and standardised process design kits are doing for photonics roughly what they did for electronics decades ago.

Photonics research has never lacked impressive demonstrations. What it has historically lacked is a reliable route from a working device on an optical table to something a company can order in quantity. That route is now being built, and it looks familiar to anyone who watched semiconductors industrialise.
Foundries and process design kits
The central change is the availability of shared foundry processes for photonic integrated circuits. A research group or startup can design to a documented set of building blocks — waveguides, couplers, modulators, detectors — with characterised performance and tolerances, and have the design fabricated alongside other users' work.
The process design kit is what makes this more than a fabrication service. By fixing what is available and specifying how each element behaves, it lets designers simulate a circuit with reasonable confidence that the manufactured device will match. That single abstraction is what turned electronic design from a craft into an industry, and it is now doing similar work in photonics.
Where the platforms diverge
Silicon photonics has the deepest ecosystem, inheriting mature CMOS fabrication and offering dense integration with electronics. Its weakness is well known: silicon is a poor light emitter, so lasers usually have to be bonded on rather than grown in place, and that hybrid integration step is a persistent cost and yield concern.
Silicon nitride offers exceptionally low propagation loss and broad transparency, making it attractive for sensing, frequency combs and quantum work, at the cost of weaker electro-optic modulation.
Thin-film lithium niobate has moved quickly from a promising material to a serious platform, combining strong electro-optic response with low loss and enabling modulators at bandwidths difficult to reach otherwise.
Indium phosphide remains the platform that emits light natively, and retains a clear role wherever monolithic sources matter most.
What still limits adoption
Two constraints dominate. Packaging and fibre coupling frequently cost more than the chip, because aligning a fibre to a submicron waveguide with the required precision is slow and equipment-intensive. And testing is difficult to parallelise: optical measurement at wafer scale is far less mature than electrical probing.
Neither is a physics problem, which is why both are attracting sustained engineering investment.
The signal to watch
The meaningful indicator is no longer record-setting device performance in isolation. It is how quickly a design moves from a foundry run to a qualified, packaged, testable product. That interval is shortening, and it is the number that decides which laboratory results become products.
Related reading

Metasurface optics are running into the manufacturing question
Flat lenses built from sub-wavelength structures work. Producing them at volume, at yield, with the tolerance…

Photonics found its foundry model, and the effect resembles electronics in the 1980s
Shared process design kits and multi-project wafer runs let a small team design a photonic chip without ownin…

Frequency combs are becoming components rather than experiments
A comb links optical frequencies to microwave references with extraordinary precision. Microresonator version…
