Temple 8 Research
AI Buildout, Macro & Energy Infrastructure Research
MACOM's product catalogue covers much of the optics area that could benefit from an AI-driven shift in network bottlenecks. Siliconcodesign expects coherent and coherent-lite optics to play a bigger role as AI usage shifts the bottleneck from scale-up bandwidth to data-center networks and internet traffic. This is quite bullish for MACOM longer term. x.comÂ
MACOM is a concentrated photonics pick for terrestrial high-speed optics plus emerging space optical links. Capital was moved into MACOM and Tower Semiconductor after a full exit from Lumentum, as part of rotating toward less obvious, higher-conviction AI bottlenecks while long-term conviction in the AI buildout stays unchanged. x.comÂ
MACOM's analog chips sit inside every 800G and 1.6T optical module Nvidia's Rubin needs, so modules cannot ship without it. MACOM is the volume supplier of drivers, TIAs and lasers. Lumentum ships the 200G-per-lane lasers, but MACOM owns share in the driver and TIA beside them, so a Rubin Ultra ramp in H2 2026 scales its optical line without new design wins. At 26x EV/sales and 156x earnings it looks rich, but it alone ships optical, RF and defense analog in volume. x.com x.com x.comÂ
MACOM is priced for perfection after going all-in on the second half. A $100M+ drop in cash went into a big inventory build of finished goods for 800G and 1.6T shipments, so if hyperscaler demand from Meta or Google softens even 5%, it could be left holding expensive specialized chips, and Q3 EPS guidance of $1.31-1.37 is wound tight. The quarter did show real execution: 800G LPO adoption has begun, data center guidance rose to about 40% growth, and its analog moat looks wider than feared. x.comÂ
Orbital AI data centers favor optical suppliers with space heritage, making MACOM the top pick, Tower Semiconductor a buy, and STMicroelectronics and Coherent holds. MACOM gains from heritage and content per terminal that grows superlinearly toward 6.4T-12.8T. STMicroelectronics owns the permanent PAT control-loop socket. Coherent gets incremental free-space optics demand, though not transformative. Tower wins on radiation tolerance, with rising value per PIC. x.comÂ
Optics and photonics should be among AI's biggest winners as faster GPU networks need optical connections that suppliers cannot expand quickly enough. Lumentum says some components are sold out through early 2029; Sivers is investing $30M to target more than 100M lasers a year, with operations expected in Q4 2027. Potential US restrictions on Chinese transceivers could reshape supply in 2027. Applied Optoelectronics and Sivers are the favorites. x.com x.comÂ
MACOM is a hold amid the AI sector sell-off: its investment in IQE de-risks operations a lot, but Coherent and Lumentum are better ways to play the 800G and 1.6T transition. x.comÂ
MACOM is a key optical component supplier for Google's TPUs at a time when Google wants as many TPUs as it can get. The whole supply chain keeps beating quarter after quarter, underpromising and overdelivering on capacity increases, while MACOM is under the radar, underowned and still guiding very conservatively. A high P/E is justified if the market is underestimating future growth, and the stock is setting up to smash; pullbacks are buying opportunities. x.com x.com x.comÂ
Networking is the next AI hardware trade after memory, whose share of rack value rose from 15% to 35%. Optics is moving inside the server, as Nvidia's Feynman architecture builds in co-packaged optics and silicon photonics. Scale-across networking could grow 30% a year just linking existing data centers, as inference moves closer to users and training spreads across sites. Supply is tight: Lumentum's CEO said it can meet only 30% of expected 2027 demand and sees optics tighter than memory. x.comÂ
Higher optical-fiber use per GPU could expand demand for Applied Optoelectronics. Corning said fully optical scale-up would raise fiber count from about 16 to 160 per GPU; that 10x case is hypothetical for now, with a mix of technologies likely near term, but it shows the direction of travel. x.com x.comÂ
Inference's move toward edge devices and dense, memory-hungry racks should drive demand for specialized chips, links, power delivery and cooling. McKinsey projects inference at 60% of AI computing demand by 2030. Ambarella makes on-device AI chips; Astera Labs links GPUs, memory and networks; Credo supplies short copper connections; Rambus earns from memory interfaces; Navitas reduces power waste; Modine cools dense racks; and Super Micro builds inference-specific server racks at volume. x.com … Show more
