too new, my rule of thumb is 10-20 years for real mass production real advantage of it is large single mode emission area, great for FSO. DFB lasers are born for datacom, and has been grinded to be almost dirt cheap. CPO needs arrays of lasers on the wavelength grid (see oci-msa), no need for extra high power or large emission area.

ELSFP pain points: - too much IL to the optical engine, could be > 3 dB. - not enough volume for polarization maintaining fibers (PMF) and PMF too expensive - clean env for the packaging. High power too easy to burn shit.

TECs used in ELS alone are gonna be a multi hundred million dollar market lol, guess who makes most of Bismuth and Tellurium.. funny the majority of tellurium goes into solar panels, and almost entirely to First Solar, so you can ask them to spare you some tellurium. most of micro TECs are still going into pluggables tho, total market might shrink as one TEC controls more lanes and more powerful lasers.

also all the big western equipment makers are on a countdown in general, prepare yourself from competition from China or start lobbying for regulations. Not just fab equipments, all kinds, from the low end and slowly eating upward. You've seen plenty of Rigol, you'll start seeing more Lianxun or Ceyear. AMAT and LAM are well known, so will naura and AMEC, for various reasons. the capital market and competition there are far from healthy or even sane, and it brews the most vicious monsters with almost human sacrifice ritual like operations. It's not a world most people would want to live in, but also at the same time grinds down the price of many seemingly precious or meticulous things. The cost is real innovation. Everyone asks you whats your unit per hour, no one would ask where is your 100x or 1000x compared to your competitor or the incumbents. The questions you get in the US is totally flipped, uph? commercialization? We'll figure it out later. But wheres your 1000x? Why would I bet on you? Perfect curation for larpers. It's almost a formula for many startups, western larpers commanding or wrapping eastern grinders. Dont get me wrong there are also eastern larpers and western grinders, I'm just talking about what you'd get if you draw a few random samples. life at big equipment co is mostly very comfortable. One thrives in adversity and perishes in laxity.

this somehow reminded me that there was a surge in immersion cooling for bitcoin miners in 2021 that helped developing of the tech and supply chain involved, because mining got banned in China and immersion is way less noisy than air cooled. And later cold plate development because then you can hide miners in heaters.

we are dreaming too small. The total number of agents will double every 9 months, and number of agents per user will double every 2 years, millions of agents in 2024, trillions of agents in 10 years. We'll need EDA tools for harness architecture. the problem is most people just consume. Do you need hundreds of agent swarm to consume? I guess if you consume things generated on the fly will create that much demand.

the more I chat with packaging people the more respect I have for cerebras, their IO in and out of the wafer sucks, but their power delivery absolutely mogs CoWoS or EMIB-T, so much cleaner than these clown show with amateur clowns. Good luck to those powerpoint warriors that dream up as many dies and as much power without any basic thermal and PI checks, and then scrambling to try fixing their broken supply chain with broken capabilities.

i was looking at some ultra processed dessert, and awed by how many parameters it was optimized with, and it then occurrd to me that LLM outputs are literally ultra processed information.

- Grutter2020: [Design Challenges for Socket Level Photonic I/O]( CPO numbers are from Lee2025: [Photonic Interconnect for Next-Generation AI Systems]( I wonder how much these foundry numbers have updated from 2020 ( Existing PIC foundries (semiconductor foundries in orange) )