Thread Guy

Thread Guy × Bubble Boi

Kimi’s linear-attention architecture may keep most model memory fixed as context grows, weakening Bubble Boi’s memory-and-capex thesis for AI winners.That could reduce future DRAM and rack demand while improving premium Nvidia GPU economics; he favors selectively priced CXMT, remains bullish on Intel’s 14A customer commitment, and is watching neoclouds, flash-based inference, and Qualcomm’s unproven throughput claims.

Thread Guy × Bubble Boi

Intel’s turnaround thesis is shifting from fabs alone to advanced packaging, with improved 18A yields and a possible 14A leapfrog over TSMC if execution continues.EMIB could enable denser, more repairable multi-die systems, while token demand and semiconductor capex support at least three more years of AI infrastructure spending; execution and speculative bottleneck trades remain the risks.

Thread Guy × SuspendedCap

Oil, not battlefield headlines, sets the market boundary: above roughly $90, expected rate cuts could disappear and equities lose monetary-policy support.SuspendedCap is holding roughly 45% cash, capped-downside options, and selective NVIDIA, software, consumer, and European exposure while warning that memory’s commodity cycle may break within a year.AI demand may persist, but the cycle’s catalyst is capital discipline: if markets stop funding customer capex, semiconductor equities could still fall 30-40%.

Thread Guy × Ejaaz

Ejaaz has shifted from expecting an imminent AI crash to believing geopolitical competition with China will keep funding compute, energy, and robotics despite unclear near-term returns.Reported costs of $4.6 million for Kimi K2 Thinking versus rumored $2.4 billion for GPT-5, alongside oversubscribed GPU clouds, challenge brute-force scaling and the stranded-chip bear case; watch for falling purchases or genuine excess capacity.