Photonics and the Future of Computing
Photonic computing is like hollow core fibre. The advantages are so great in each case that their long term adoption appears almost inevitable. Logic gates in traditional silicon-based computing rely on using electric charges to represent binary information. The drawbacks are quite clear. The electricity becomes heat. In the presence of high transistor density this translates to calculation errors as well as hardware failure. In turn, high heat requires cooling systems and more electricity. Indeed, electricity is the largest operating expense for a data centre. In 2025 Equinix facilities consumed 8.6 Terawatt hours. A good guess is that at least 60% of the firm's cost of revenues is power. In 2025 the Equinix cost of revenue totalled $4.5 billion so power costs were at least $2.7 billion.
In contrast, photonic computers uses infrared lasers on chips. This consumes a fraction of the power that silicon wafers need. Residual heat is minuscule so the cooling demand drops substantially. The energy savings can be as great as 99% depending on whether a pure photonic setup is used or a hybrid photonic/silicon approach. Another advantage is that photonic computing is much faster for many applications because silicon technology relies on capacitors to induce voltage changes to flip the logic gates. This is much slower than photons traversing fibre optic strands embedded in a chip. Depending on the type of photonic technology, speed gains range from 10X up to a 1000X of its silicon wafer counterparts. Not that computer clock speeds have stagnated for the last 25 years and still remain around 5 GHz because faster cycles require more power than can be effectively cooled. Most computer productivity improvements since 2000 are due to more instructions per cycle and multicore architecture.
In a photonic computer light is transformed in different ways to execute computations. One MIT startup manipulates phase and amplitude (intensity) to calculate whereas other projects involve entangling photons and creating quantum qubits. Photonic computing is far from mature and nonlinear computing is actually quite challenging, but power savings and computation speed acceleration create a strong incentive to solve the current challenges.
See https://thetechcapital.com/photon-fortune-how-the-future-of-ai-may-run-on-light/

Comments
Post a Comment