Serenity spent Tuesday chewing over a single industry headline, but one he thinks carries more signal than most: a report (via Digitimes) that the chairman of Largan Precision cast doubt on whether MicroLED technology will ever be viable for AI data-center interconnects. Largan is a Taiwanese optics specialist embedded in TSMC's COUPE co-packaged-optics supply chain, making fiber-array units, microlenses and other passive components — which, Serenity argues, makes its skepticism relatively neutral rather than self-serving.

To unpack the stakes: MicroLED has been floated as a longer-dated light source for optical links, a possible rival to the laser-based approaches Serenity has built much of his book around. The Largan chairman reportedly flagged power consumption, size and transmission efficiency as reasons MicroLED is unlikely to commercialize. For Serenity, that is indirectly bullish for the laser roadmap he already backs — VCSELs, and especially the CW (continuous-wave) lasers that act as the light source for silicon photonics and co-packaged optics (CPO), the technique of placing the optical engine right next to the switch or GPU die to save power and lift bandwidth.

His read of the broader landscape: he thinks AMD, NVIDIA and the wider industry are visibly scrambling to lock up CW laser supply this year. AMD (an AI accelerator and CPU maker he treats as a demand engine rather than a personal holding) and NVIDIA (the roadmap he uses as a telegraph for where the next bottleneck lands) matter here because their CPO ambitions pull through the upstream laser suppliers he actually owns. Last year the market's attention was on EML (electro-absorption modulated) lasers for 800G/1.6T links, which played to Lumentum and Coherent — both disclosed holdings he treats as quality large-cap photonics anchors, Lumentum at a 5% weight tied to Google TPU optics and Coherent at 2.5% as a vertically integrated, lower-beta compounder.

MicroLED, by contrast, was the 2028-and-beyond wildcard. Serenity noted that Credo — the connectivity chipmaker he holds as a datacenter beneficiary — had leaned into the space via its Hyperlume acquisition, that AMS Osram (an Austrian-German optical-semiconductor firm making LEDs and sensors) was a popular name working with Credo, and that hyperscalers like Microsoft were developing in the area too. Microsoft, for him, is less a position than a demand signal: a top-tier cloud spender whose capex underwrites his real photonics longs. TSMC, meanwhile, sits as the indispensable foundry center of the whole trade.

The practical takeaway Serenity drew: if Largan's chairman proves right, the center of gravity stays with lasers — VCSEL, then CW and silicon photonics — reinforcing the thesis that his upstream chokepoint names are where the AI optical buildout ultimately flows. It was a quiet, read-through day, but one that nudged his conviction in the laser-first path rather than changing his positioning.