Tulip King
Head of Strategy at CounterpartyTV, Crypto Investor
The AI trade is probably over because every layer is heading toward surplus. Chinese models are undercutting American ones; memory companies capture most profits, so much of the semiconductor complex is rallying around bringing more memory online over the next 18 months; and Google is pivoting from its own models to renting cloud capacity to frontier labs. Fatigue with AI lab leaders' doomsday talk could also lead to data center bans. x.comÂ
Semiconductor hardware faces pressure as AI shifts from racing on capex to competing on capital efficiency. Meta selling surplus compute amid slower-than-expected agent progress raises doubts about returns on AI spending, while Anthropic is reportedly exploring custom chips with Samsung. Nvidia moving to revenue sharing and Meta renting out capacity squeeze resellers like CoreWeave. Makers with capacity but no pricing power will be hit first if demand growth slows. AI demand has not collapsed. … Show more
Semiconductor trade faces a capex growth ceiling even if AI delivers what bulls expect. Spending cannot compound at 70-100% a year indefinitely, and compute per dollar roughly doubles every two years, so the same budget buys more capacity later. Today's chip valuations also demand years of 20% AI earnings growth beyond 2028: roughly 10.9 extra years for AMD, 8.9 for ASML, 7 for Marvell and 6.4 for TSMC, versus 1.6 for Broadcom and 0.8 for Nvidia; Micron needs none under the chart's assumptions. … Show more
Semiconductors could stay undersupplied for a long time as more capable AI per dollar creates more uses and drives compute demand. Micron and NVIDIA trade at low forward P/Es because investors fear peak cyclical earnings, yet current earnings show no slowdown. Shortages may worsen as AI adoption spreads, though the industry could remain cyclical; a reversal of the forced rollover trade is hoped for, not assured. x.com x.com x.comÂ
Semiconductor stocks' easy-money phase is over as the sector shifts from hype to show-me-the-money. The SOX index is down over 15% from its June peak, chip funds saw $11B of outflows in one week, the largest this century, short interest is at a three-year high and earnings growth is projected to slow by 2027. Memory and logic chips remain commodity-like. This is cyclical cooling, not a death knell; the AI build-out continues, but expect more turbulence and favor companies with defensive moats. … Show more
A worsening memory shortage in 2027 could favor Micron, SanDisk and SK hynix as AI stocks. Samsung warns that the shortage will worsen then. x.comÂ
China's domestic market lets YMTC and CXMT threaten Micron, Samsung Electronics and SK hynix without entering US products. China receives 15-25% of smartphones and PCs and is on a path to at least 20-30% memory share over the next several years; current leaders are much too optimistic. History suggests at least a near-term semiconductor rally even if the AI trade has topped. It likely has not: agentic AI should give the trade room to run into 2027. x.com x.com x.comÂ
Semiconductor stocks could take a short-term hit because the market's premise that frontier models will handle most valuable AI work is likely wrong, which might be a small shock. As models improve fast, frontier models will be too costly for most tasks, so AI labs' growth and margins slow and fundraising gets harder. Compute they reserved at hyperscalers and neoclouds could be released, slowing capex growth. Long term this is not bad for semis, but the market is unprepared for the gap. x.comÂ
Semiconductors may be nearing peak AI-infrastructure euphoria as hyperscalers appear to have overspent on capex. Per a Bloomberg leak, Meta plans to sell excess AI capacity to third parties, entering cloud; one of the world's largest GPU buyers placing surplus capacity is a first warning. The same session showed a textbook rotation, with semis (SOXX) down 6.4% and cloud software (WCLD) up 3.6%. AI is here to stay, but innovation doesn't equal investment success. x.com x.com x.com x.comÂ
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
Most AI chip companies are unlikely to be good investments over the next decade, because so much capital floods in that excess returns get competed away. Chipmakers need massive capital and constant reinvestment just to stay relevant, a treadmill rather than a compounding machine. Value concentrates only in customer relationships, lowest-cost production or toll-bridge bottlenecks. When returns miss high expectations, repricing is fast and brutal. Winners are uncertain, but most won't be. x.comÂ
