Lightwave Logic Inc (LWLG)

[MarcoPolo4]

From Rajiv Pancholy head of Hyperscale Strategy and Optic Systems Division Broadcom
As generative AI systems move to unsupervised, transformer-based parallel architectures, there is less time for resending packets due to data transmission errors. Improved bit error rates are thus required to reduce training times while higher interconnect bandwidth and data rates are needed to support larger GPU clusters. These compute networks are already moving to 224 gigabit PAM4 well before the previous generation at 112 gigabit PAM4 was allowed to reach hyperscale deployment volumes.

He is telling folks that open eyes are clearly needed in modulator transmission and that means a low error rate and less resending of data He is also saying we are not far from moving to next generation bandwidth. A large error rate lessens the efficiency of greater bandwidth. AI will not like that.. So low optical loss matters. Michael Lebby has been touting the low optical loss with polymers and the open eyes in transmission. TFLN has a significant optical loss problem now that will only be magnified with increased demands for next generation speed. He might as well be saying polymers will win out over TFLN.