AI capital is tracking time-to-compute as SMCI and Lumentum earnings highlight new infrastructure bottlenecks
WhiteLine Daily, published by WuBlockchain, argues that recent earnings from CoreWeave, Supermicro and Lumentum point to the same shift in AI infrastructure: demand has not cooled in a meaningful way, but the main constraints on bringing capacity online are no longer just GPUs. The bottlenecks are moving toward financing, power, cooling, networking and optical interconnects. For Supermicro, the report says demand remains firm. Fourth-quarter revenue reached $11.1 billion, up about 93% year over year, new orders topped $60 billion, and the company guided for $65 billion to $72 billion in FY2027 revenue. Management said some revenue delays were tied not to GPU or server readiness, but to customer-side power, cooling and networking conditions. WhiteLine Daily said the next key variable is margin stability, with Q4 gross margin recovering to 17.5% from 9.9% in the prior quarter while FY2026 gross margin stood at 10.8%. Lumentum’s results showed a similar pattern from the optical side. Fourth-quarter revenue rose to $1.01 billion, up 109% year over year, while non-GAAP gross margin climbed to 50.4%. The note said tight supply in AI optical interconnects is now showing up not only in orders but also in profitability, with Sivers and IQE identified as downstream and upstream names to watch.








