SpaceX's $40B financing for GPU procurement adds another large, funded source of AI-infrastructure demand, with spending flowing into GPUs, memory, networking and optics. At the same time, InP substrates are seriously short, and Morgan Stanley identifies InP, EML and CW lasers as primary bottlenecks; suppliers that secure raw materials should gain pricing power. Any slowdown looks unlikely. x.com 

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

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.com 

Sivers is less exposed to tighter AI financing because its 700M SEK raise funds a buildout largely reliant on external foundries. It is qualifying with Ayar, Jabil, GlobalFoundries and O-Net for potential tier-one hyperscaler pluggable, CPO and NPO deployments. Higher rates may strain neoclouds needing billions for GPUs, but optical content and aggregate demand should still rise even if hyperscaler AI capex falls, especially as new players start qualification. x.com 

Optical networking's broader rollout will be slow because scale-up and scale-out require orders of magnitude more connections than today's limited optical distance links. Supply-chain discussions echo that slow pace. NPO/CPO starts in switches and reaches XPUs and GPUs much later, with new challenges. Optical backplanes will be necessary as racks move to optical, though copper will remain in many implementations. x.com  x.com  x.com  x.com 

More lasers per module and new 400G DSPs could keep the pluggable-optics runway open for years, supporting Lumentum, Coherent, Marvell, Applied Optoelectronics and Sivers even as co-packaged optics approaches. NVIDIA and Cisco co-designed a module with four times the bandwidth in the same slot; fewer modules per GPU look bearish, but the added optical content points the other way. x.com 

Semiconductor demand could withstand slower OpenAI and Anthropic growth if open-source models and post-training add more chip demand than it removes. Falling token prices need not erode per-GPU economics as cost per token falls and output per GPU rises, potentially keeping rental rates firm. The balance is unresolved; the lean is bullish, but markets may focus on the easier-to-see slowdown. Shares could chop until numbers defeat the bears, or a well-founded bear case could trigger heavy selling. … Show more

Broadcom, Arista, Celestica, Credo and Astera Labs should inherit AI network spending as flatter data-center designs shift dollars from optical transceivers to silicon, integration and copper. B. Riley warns Amazon's RNG and OpenAI's MRC protocols could cut transceiver demand 40-50%, though the facility-wide drop looks nearer 25-35%. Broadcom ASICs run MRC routing, Arista enables flat clusters, Celestica builds custom racks, Credo copper replaces short optics, Astera retimers ease bottlenecks. … Show more

Scale-up moving to optics adds demand for optical suppliers even if new AI architectures need fewer switches. Bigger data centers and more GPUs add to that need; Jensen Huang expects future optics requirements to be several orders of magnitude above what NVIDIA ships today. This challenges the optics bear case for Lumentum, Coherent, Tower Semiconductor and Credo. x.com 

Credo trades at 15-18x forward earnings despite 200%-plus accelerating growth because the market treats it as legacy connectivity while its TAM has multiplied. It is taking scale-out share from copper and fiber and is poised for scale-up, and its ALCs match copper's reliability at half the power of laser optics. New products like ZF Optics and OmniConnect CXL memory links, which management expects at $1,000 per GPU, could add about $1B on 1M GPUs, versus $1.3B of mostly AEC revenue this year. … Show more