Why many ETFs are dumb indexes and value may fail

Bullish — $LITE $AAOI $SIVE $IQE $AXTI Citi, after investor meetings: “CPO is emerging as a major part of the next value-upgrade bottlenecks in AI infrastructure.” The next scarce layer is not another GPU. It’s light: lasers and fiber, packed onto the switch. GPUs don’t scale if the photons don’t. Get positioned before it is too late. We are long and bullish. NFA.

Google's investment portfolio is INSANE! Its SpaceX $SPCX stake alone is valued at $94 BILLION, accounting for roughly 95% of its disclosed portfolio. Here's $GOOGL latest portfolio:

Looks likely Elon is going to use $SPCX as ticker for SpaceX IPO, which he had to acquire from Matt Tuttle, hence the ETF's ticker change shown below. We predicted this could happen in Dec note. Nice catch by Will, who famously 'gave' $META ticker to Zuck. Game recognizing game.

$AXTI THE INP SHORTAGE HASN'T PEAKED! Chinese InP substrate makers hardened their price language over the summer. On July 16, Yunnan Germanium (SZSE: 002428) told investors InP wafer prices had "risen somewhat" and flagged "considerable uncertainty" ahead. By September 11, prices were "rising sharply," with volume and price climbing together, and the warning was gone. More capacity is on the way. Yunnan Germanium is expanding from 150k wafers a year (2 to 4 inch) to 450k (4-inch equivalent) on an 18-month build that began in April, which points to full output around October 2027. In July it signed an InP order worth RMB 570M to 855M, about $85M to $127M, with an unnamed customer, delivering through the end of 2027. Guangzhi Technology (SZSE: 300489) has approval for a 40-ton-a-year InP wafer project and says it holds a 6-inch order from a leading global customer. China gets that capacity first. Every InP export has needed its own license since February 4, 2025, and none of the truce-era suspension notices has covered it. Yunnan Germanium's overseas revenue was 1.45% of first-half sales. AXT also makes its substrates in Beijing and drew $30.8M of its $47.6M Q2 revenue from mainland China, but its sales to Europe rose to $9.2M from $1.6M a year earlier on the export licenses it holds. $LITE has a six-year capacity reservation with AXT, and $COHR prepaid $22.3M for AXT capacity from 2026 to 2028. The buy side is tracking it closely. Three Chinese substrate makers hosted 70 institutional visits in September, and the questions turned to newcomers and oversupply. The "supply-demand gap exceeds 70%" line Bojie (SZSE: 002975) gave on September 22 was its answer to a fund asking whether all the new capacity would end in oversupply. I read the three companies' September IR records on CNINFO against AXT's Q2 10-Q. So, until the new Chinese lines come up in the second half of 2027, the extra substrate available to laser makers outside China runs mainly through AXT's licensed exports and the two Japanese suppliers, JX Advanced Metals and Sumitomo Electric. Outside China, the companies buying those wafers are saying the same thing. Coherent CEO Jim Anderson told investors on August 12 that InP capacity "continues to be our primary constraint." In a Six Five Summit conversation released September 17, he went wider: indium phosphide is "kind of a constraint across the industry right now." $AVGO says its InP, EML, CW and VCSEL fabs are already running at more than three times year-ago capacity, with substantially more to come over the next two years. The fixes all carry 2027 and 2028 dates. Lumentum's UK plant comes online in the second quarter of 2027 and Greensboro in mid-2028. JX Advanced Metals is spending up to ¥120 billion over four years for 7 to 10 times its fiscal 2025 substrate capacity, and Sumitomo Electric is reportedly aiming for 3.1 times its fiscal 2024 level by fiscal 2028. And scale-up hasn't placed its InP order yet. Anderson made the point in the same conversation: the only part of the network still running on copper is scale-up inside the rack, and it moves to optics over the next few years. When it does, that demand lands in the same 2027 window as every expansion above.

By the time people realize this… it will be too late. $SIVE, $AAOI holders prepare for a generational run

$LITE $COHR $AAOI $SIVE $CRDO Scale-Up CPO pushed to 2029, what happens now? Jefferies just took AI transceivers to ~180M units in CY27 (90M 800G + 80M+ 1.6T) and 245M in CY28. Scale-up CPO? Material volume pushed to CY29. As I see it that’s two extra years of front-panel demand before CPO even starts eating share. The market keeps pricing these names like the pluggable cycle is ending. Jefferies just said the opposite. —> $LITE owns the EML bottleneck. Every extra 1.6T module still needs 200G/lane lasers. That’s not optional. —> $COHR is InP + CW + modules with demand still running ahead of supply. Capacity is the constraint, not the order book. —> $AAOI is the US module ramp sitting right in the middle of 90M + 80M units. Their next 24 months is 800G and 1.6T. This just made that window bigger. —> $SIVE is the laser that plays both sides. Jabil/1.6T pluggables now. Ayar/GFS/NVLink ELS when CPO actually shows up. A later CPO ramp doesn’t kill the thesis. It funds the company while the pipeline converts. —> $CRDO wins on volume either way. More 1.6T ports = more DSP, more AECs, more ZeroFlap optics. DustPhotonics now puts the PIC on the same stack. CPO is the 2029 story. 245 million pluggable units in 2028 is the trade in front of you. Bullish! NFA.