Onto Innovation looks good around current levels because it is cheaper than many of its peers. The comparison comes in reply to @Opu61659893, who argued that a structural shift toward heterogeneous integration and advanced packaging should favor names like Amkor, Onto and Entegris, and that an AI capex unwind would slow but not kill it. Amkor could also work from here, though ASE is a little cheaper. x.com 

Onto Innovation looks cheap against its earnings, trading at 31x forward earnings versus the roughly 38x chip-equipment makers usually get. Applying 38x to next year's forecast $9.95 a share, give or take 12%, puts fair value at about $333–423, versus $308.30 now. The case leans on that forecast, with forward EPS 102% above trailing, and rests on valuation levels from reported earnings and estimates, not fundamental work. x.com 

Applied Optoelectronics' projected $5.65 billion in 2027 revenue against a $6 billion market cap drives its appeal. It may be the best opportunity in the market right now. x.com 

KLA trades at a premium valuation despite lower growth and less specialization in advanced packaging, the hottest AI segment. Camtek and Onto Innovation often win specific inspection and metrology sockets with better throughput, cost of ownership and packaging-optimized 2D/3D performance, while KLA, dominant overall, is sometimes less specialized for the hardest HBM and hybrid bonding steps. Onto shows the strongest Rule of X at the lowest cash-flow multiple. x.com 

AXT makes the indium phosphide substrates needed for high-speed data center connectivity, making it an AI infrastructure and optical connectivity play. EPS is expected to grow from $0.86 in 2026 to $3.29 by 2028, nearly 4x in two years. The chart is building the right side of a cup base. x.com 

Applied Optoelectronics is the US-built option in 800G datacenter transceivers, where there is almost no domestic supply. It is scaling its line in Sugar Land, Texas, while larger Innolight and Eoptolink build in Asia and Lumentum and Coherent are crowded large caps. It also makes its own lasers, which Goldman flags as the binding shortage through 2027. Revenue rose 86% last quarter, with net cash. Risks: a -12% operating margin, concentrated customers, and whether in-house lasers hold at 1.6T. … Show more

Applied Optoelectronics is a domestic AI data-center bottleneck winner whose results show exponential operating leverage. EPS jumped 186% quarter over quarter and 138% year over year on an 86% revenue surge, as 800G transceiver economics improve with its Texas and Taiwan fabs reaching prime yields. With the US tightening restrictions on Chinese optical supply chains and the FCC banning new Chinese data-center hardware, hyperscalers are flooding it with high-margin orders. x.com 

AXT remains a vital InP substrate bottleneck, and keeping some exposure makes sense because a ton of AI capex flows directly to it, though much is priced in. Transceivers moving from 800G to 1.6T and 3.2T need lasers built on InP substrates, which AXT and Sumitomo supply; demand exceeds supply even as AXT doubles capacity by end-2026. Management sees CPO inflecting late 2027 and beyond. Earnings should improve as backlog converts, though another parabolic move looks unlikely. x.com  x.com 

Interconnect, not compute, is now the binding constraint on AI clusters. Copper, pluggable optics and CPO will coexist for years, so the strongest names span both paths or own the tightest bottleneck. Top five: Credo, the fastest-growing full stack; Applied Optoelectronics, the 800G/1.6T ramp; Marvell, broadest but richly priced; Lumentum, the laser chokepoint; Tower, the foundry bet. Coherent, Amphenol, Fabrinet, GlobalFoundries, TSMC and Broadcom also gain. Astera Labs is pricier and narrower. … Show more

Applied Optoelectronics has interesting risk-reward around these levels given laser shortages and the scope of the 800G and 1.6T ramp it plans over the next 12-18 months. Its history still makes it a name with real risks. x.com