AI hardware demand remains contract-backed, but capital discipline is now the test: memory and WFE evidence is firm while hyperscaler cash returns, future supply and SSD substitution divide experts. Disclosed positioning favors optics, tactical crude and existing software—not indiscriminate dip-buying.
Aswath Damodaran says Apple may win by waiting while hyperscalers risk having committed too much capital; Neil Campling says Meta's free-cash-flow measure could turn negative. An unaudited X calculation puts Google's incremental ROIC at 16%. The supplier counterargument is that lower inference cost broadens usage rather than reducing total infrastructure demand.
Daiwa memory analyst SK Kim calls China-supply concern overdone and points to hyperscaler demand, while KLA call analysis raises the WFE path to the low-$150 billion range for 2026 and about $190 billion for 2027. Jim Cramer counters that parabolic memory stocks historically turn before business data, and an unverified Reddit author treats current profits as a capital-cycle peak.
Seagate call analysis says customers will pay above contract price for incremental exabytes, with product commitments through 2027 and allocations into 2028. A separate SK Hynix read-through identifies AI data lakes and HDD replacement as QLC enterprise-SSD applications, while preserving HDD advantages in cold storage.
Serenity, Alpha Score rank 3 with 44 smart followers, retained full conviction in SIVE and AAOI. Crux Capital, followed by 21 smart accounts, argues that scaling from thousands to millions of wafers limits near-term overbuild; its substrate, epitaxy, tool and test names are a beneficiary map, not disclosed positions. KLA's broader equipment cycle supports capacity spend but not every vendor equally.
Serenity explicitly retained full conviction in these optical names; the post did not disclose fresh purchases.
Temple Eight opened a crude-futures long targeting the $90s, then kept long crude paired with short BE after temporarily closing QQQ puts; the rationale was thin.
Demand and pricing evidence is strong, but new WFE capacity, crowded sell-side optimism and early equity turning points argue against an indiscriminate basket.
Ansem retained a $600 target but softened the timing; the transcript did not establish a new purchase.
>> Um I do think I do think 600 is the target though.
AskLivermore said these software positions were still held; the reply disclosed neither a fresh add nor a new thesis.
Cramer argues shortage-driven memory equities can fall before fundamentals; strong spot data alone is not a timing signal.
The Google 16% incremental-ROIC figure is an author's naive calculation; validate definitions, depreciation and timing before using it.
Crux mapped AXTI, VECO, AIXA and IQE to bottlenecks but disclosed no holdings; keep them on a conditional watchlist.