SemiAnalysis

Using Coding Agents at SemiAnalysis

Indium Phosphide (InP) LASERs are suddenly catching everyone's interest since it suddenly became one of the more strategic components in computing. As optical connectivity moves ever closer to the package, every optical engine and laser source inevitably runs on InP LASERs. But supply isn't keeping up with demand, particularly for Continuous Wave (CW) Distributed Feedback (DFB) LASERs which enable near-package optical connectivity. So why can't the industry just pour money at the problem and solve this supply scale issue immediately? What stops this much needed ramp up? Making a CW DFB InP LASER is far from simple - it is like baking a complicated multilayer birthday cake (and don't forget the candles), and with the III-V industry living in the metaphorical stone age with 2" to 4" wafer production standard compared to CMOS with its 12" wafer processes, supplying emerging scale-up and scale-out applications is like trying to cater a wedding using your apartment's tiny kitchen. Does trawling through incredibly technical papers on LASER fabrication make you feel like your brain is getting a quantum cavity? Stay tuned! Our upcoming note on LASERs has the answers.

Semiconductor industry right now be like

SemiAnalysis joins @TuckerCarlson on Fox News to ask "what are we even doing" with regard to ai stocks.

Short Statement, More Dissents, New Normal Warsh claimed tightening of long rates as market success. And no trade off between inflation and employment. So how do rates impact the economy?

Ever wonder why PCB folks throw around "M8" or "M9" like a spec sheet magic word? The "M" traces back to Panasonic's Megtron series — long the benchmark for low-loss, high-speed copper clad laminate (CCL). Each generation from Megtron 4 to 8 progressively lowered the dissipation factor (Df), meaning less signal loss at high frequencies. Because Megtron set the bar, it became industry shorthand for a performance tier rather than a Panasonic-exclusive part number. Say "M8 grade" today, and it simply means "meets Megtron 8 electrical specs" — whether it's made by Panasonic, EMC, or Doosan. As AI servers demand ever-higher SerDes speeds, high-tier CCL has become the critical bottleneck material. That makes it one of the tightest, fastest-growing segments in the entire hardware supply chain.

The Wild Wild West Of LEGO Datacenters Everyone Says They're Modular, Do The Vendor Claims Hold Up? Zuck's Tents, AWS's Houdini, 60GW+ Modular Capacity Tracked, Full Vendor Landscape Mapping, Vertiv's 2x Content Uplift Per MW https://lnkd.in/gWXn_CTn

Trading Houses For Datacenters

The Street models 2028 WFE around $190–200B. If the Top-5 equipment makers simply sell out planned capacity at today's prices, we think WFE lands above $230B. Capacity plans support it: AMAT says it can double output with floor space ready to go, Lam has nearly doubled manufacturing capacity in four years, TEL targets 1.8x by FY2029, and ASML turned pre-emptive on capacity after 2Q26. Prices won't stay at today's levels. Like-for-like increases are already on the table. ASML flagged them at 2Q26 earnings, TSMC cited tool price inflation as a major driver of its +15% capex guide, Chinese customers are reportedly agreeing to +10% on DUV, and channel checks suggest TEL is discussing ~30% increases. The reason they can: Top-5 SemiCap revenue is at a cyclicals-lows of big-4 chipmaker revenue (TSMC, Samsung DS, SK Hynix, Micron). Chipmakers' profit pool has never been this deep relative to what the tools cost them. Price hikes flow ~100% to toolmaker gross profit. Their suppliers have no pricing power. A decade of data shows Ichor's gross margin banded at 8–17% and Ultra Clean's at 11–21%, with no breakout even in the tight 2017 and 2021–22 markets. Toolmakers reprice customers annually while their own input costs stay flat. A 10% like-for-like increase would put SemiCap gross margins 1–5pts above historical peaks. Volume up, price up, margins to new highs.