Silicon Photonics' optical transceiver market share is estimated to exceed 50% by 2027. $TSM VP of Advanced Packaging said today that silicon photonics is becoming the mainstream form thanks to the structural transformation of optical transceivers. He also stated that the real bottleneck for large-scale deployment lies in lasers, optical fibers, fiber optic connectors, and product testing. If I was to pick the "main" players instead of just name-dumping: - $LITE / $COHR: InP lasers - CW for SiPh, ELS for CPO, 200G EMLs - $AEHR: SiPh WLBI testing - $GLW: PM fibre, fibre arrays & MPO/MTP connectivity - Sumitomo Electric (5802): InP lasers + substrates Obviously you've got a ton of other names from: $AAOI for InP lasers, $IQE for InP base epi, $BESI for hybrid bonding and $KEYS / $VIAV for optical test. Feels like TSMC have been running the same narrative alongside the COUPE roadmap for a while now though. But always good to get some re-confirmations on where everyone's favourite bottlenecks sit / narratives coming from TSMC themselves.

TW silicon photonics names are unstoppable! 🥳 Based on $TSM’s current guidance, 2027 EPS is estimated at around NT$200. At 20–25x P/E, that’s NT$4,000–5,000/share, or roughly US$3.24–4.05tn market cap. Assuming Xi doesn’t invade us first 😂 Can you imagine how much liquidity could get sucked into the Taiwan stock market if $TSM actually trades there? US$4tn would put TSM alone close to the size of the entire UK stock market today. And TW photonics micro caps can easily 5–10x riding this wave, I think.

As silicon photonics moves toward volume production, designing the PIC is only part of the challenge. Designs become more complex once the optical engine is tied to a foundry process, packaging strategy, and system architecture. On @SemiEngineering:

$GFS $TSM $INTC Schulz Class #10#: silicon photonics, or drawing light's paths into a chip Class #9# ended on one fact: silicon can't make laser light, but it's see-through to it. That turns out to be enough to build a business on. GlobalFoundries expects about $500M of silicon photonics revenue this year, up from over $200M in 2025. What it is A classic optical module is put together a bit like a camera: a laser, lenses, a modulator, more lenses, then the fiber, each a separate piece lined up by hand or machine. Silicon photonics prints most of that onto one chip. Thin strips of silicon called waveguides act as wires for light, split one beam into several lanes, and carry it through modulators on the same chip that write data onto each lane. The chips are made on the same 300mm wafers and tools used for computer chips. The laser still comes from indium phosphide, either as a separate part or bonded onto the silicon. Why it matters Chip factories are built to make the same part millions of times. GlobalFoundries said on Nov. 4 that it now expects to exit 2027 at an annualized silicon photonics revenue run rate of about $1B, a year earlier than it planned, and to roughly triple its silicon photonics capacity next year. It has more than 50 silicon photonics customers, including 7 of the top 10 optical transceiver makers. One laser can also feed several lanes, which is how the 800G chip in Class #8# got by with two lasers instead of eight. What it means for stocks GlobalFoundries gives the clearest read because it puts a number on it. NVIDIA's co-packaged optics switch, developed with TSMC, uses TSMC's COUPE platform, according to TrendForce. Intel went the bonded route: by June 2024 it had shipped over 8 million silicon photonics chips carrying over 32 million on-chip lasers. Move that chip out of the plug-in module and right up against the switch chip, and you get co-packaged optics. That's the next class. We looked at what GlobalFoundries' faster ramp means for each share earlier this week. What would you like the next class to explain?

$INTC Claims Industry’s Only High-Volume 300mm InP Silicon Photonics Capability At OCP 2026, Intel claimed it is the only foundry capable of supplying this technology at scale to the broader industry. Its platform integrates InP-based lasers into 300mm silicon photonics manufacturing, aiming to improve production efficiency and support next-generation near-package and co-packaged optics for AI infrastructure. As NVIDIA and Broadcom advance CPO commercialization, testing remains a major constraint. Fully testing a single photonic chip can still take over 100 seconds. TrendForce sees 2027 as pivotal for integrating CPO testing systems, followed by a critical assessment of mass-production readiness in 2028. $INTC $NVDA $AVGO $TSM