AI inference is hitting a memory limit, and CXL switching, which pools memory across servers, is where the value map changes. HBM is costly and limited per chip, and bigger models outgrow one server. Marvell is the bet on switching as the control point, via XConn and Structera S, now in production. Penguin Solutions makes CXL deployable in its KV cache server. Micron's CXL DRAM is a structural tailwind still tied to DRAM cycles, and Credo sits in the competitive transport layer. x.com  x.com  … Show more

Networking, power and compute look more durable than cyclical memory in the AI buildout. Lumentum's revenue compounded year after year across the 2010s and 2020s, unlike Micron's; AWS revenue also compounded. Lumentum, Nebius, Bloom Energy and Credo should heavily outperform SanDisk and Micron from here, though AI itself remains heavily doubted. x.com  x.com 

Astera Labs' CXL integration thesis faces a threat if Nvidia builds a CXL-attached memory architecture: Jensen would not use Astera for it. Such an architecture could make training and data sharing across nodes easier. Nvidia's Feynman might also get a better I/O die to avoid switch hops, as Google did with TPUs. x.com  x.com 

Agentic AI's repeated context favors cache memory and storage demand for Micron, SK hynix and Sandisk over fresh GPU computation. Per @demian_ai, about 95% of input tokens for DeepSeek V4.1 Flash on OpenRouter were cache hits; storage matters unless routing and caching keep almost everything in GPU memory. New products such as 3D DRAM and HBM-on-logic could be the final trigger for a full memory-stock re-rating. x.com  x.com 

SanDisk CEO David Goeckeler's report of customer signals beyond fiscal 2030 adds storage to the long-duration AI hardware-demand case. He says those signals come through new-business-model partnerships; Micron and SK hynix in memory and Lumentum in photonics were already cited as signs of demand beyond 2030. x.com 

Astera Labs carries a huge premium for flawless execution against a CXL market that is small near term. Switch-based CXL memory pooling capturing $15-16B in 2026-2028 is low probability: Marvell guides its XConn switch business to only about $100M by fiscal 2028, PCIe included. Pooling software likely stays immature into 2028, per a Samsung white paper, and Nvidia is contesting the warm KV-cache tier. Astera does CXL best, but SK hynix and Micron offer better risk-adjusted returns. x.com  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

NAND-based memory hierarchies that handle ever larger AI context and per-organization model tuning are driving Kioxia and SanDisk higher, and Phison, which plays well in the space, could follow. Tiered architectures are already happening. People use Phison's aiDAPTIV with SSDs for post-training, and its NAND is built for massive endurance, slow but long-lasting. Some data must be instant while other data can wait seconds or longer, so memory needs span a wide continuum below HBM. x.com  x.com 

Penguin Solutions buyers chasing Marvell photonic memory are likely to be disappointed, because Penguin is a Super Micro-style integrator, not an IP holder, hence its low price-to-sales ratio. It builds the 2U chassis inside the rack around photonic memory IP from Celestial AI or Marvell, which own the high-margin foundational IP, and that work is still in development collaboration, not even qualification. The base business is a legitimate idea, with material upside if it gets past sampling. … Show more

Penguin is in pole position to bring optical memory systems to customers because it was already developing a Photonic Memory Appliance with Celestial AI. Marvell is showing an HBM-and-DDR5 module that creates 33TB memory pools over optical links up to 50 meters, letting memory capacity scale independently of GPUs. x.com