Friday, July 17

Equities turned decisively risk‑off with tech and semiconductors taking the brunt: the S&P 500 closed down 1.01% at 7,457.69, the Nasdaq fell 1.40% to 25,520.24 and the VIX jumped 12.19% to 18.77. Market commentary tied the pressure to an unwinding of crowded AI‑infrastructure bets—investors questioned whether anticipated hyperscaler capex can justify stretched multiples—and to company news that punctured sentiment (a disappointing Netflix guide amplified selling). The macro‑news backdrop amplified flows: yields that had traded around 4.60–4.62% earlier in the week cooled, changing funding and carry dynamics for leveraged positions. Fresh product and model launches acted as an accelerant: China’s Moonshot AI released Kimi K3 (top‑tier coding performance, open‑weight), while OpenAI and new entrants rolled out enterprise and open models, reshaping perceived moats. Reports that Apple quietly surpassed Nvidia as the world’s most valuable company punctured the single‑name AI narrative. Implication: de‑risking is repricing cyclicality in semiconductor earnings and forcing a valuation reset for the largest AI bets, with volatility now a more persistent headwind for leveraged tech strategies.

A softer core PCE print (2.8% y/y, down from 3.2% in May and below the 3.0% consensus) materially revised rate expectations: futures pushed the probability of a 25bp Fed cut in September to roughly 72% from 58% a day earlier, while CME’s FedWatch shows July hike odds sliding to about 11% from ~45% at the start of the week. Traders now overwhelmingly expect the Fed to hold at the July 28–29 FOMC (priced ~95–96% as a no‑change outcome) and instead begin easing later in the year. That re‑pricing has been visible in the Treasury market—10‑year yields rose into the mid‑4.6% area earlier in the week before drifting lower—and in a softer dollar. Regulators have simultaneously tightened supervisory practices (joint Fed/FDIC/OCC statement on handling sensitive supervisory data) and updated bank provisioning guidance (OCC ALCO booklet v2.0), signaling readiness for a cycle turn. Implication: fixed‑income positioning and carry trades are recalibrating toward earlier easing, but the path is fragile—oil or renewed inflation prints could reverse the repricing quickly.

At the World AI Conference in Shanghai, President Xi gave a high‑level keynote and backed a new global AI governance body with 29 founding countries—an explicit move to make China a convener in international AI rule‑making. That diplomatic thrust landed alongside an inflection in model economics: Moonshot AI’s Kimi K3 (open‑weight) and Western entrants like Thinking Machines’ Inkling (975B parameters) and DeepSeek V4 highlighted an emergent open‑weight cohort that lets firms run and customize large models on their own infrastructure, contesting closed‑model business models. Regulatory pressure compounded the competitive shift: the European Commission’s Digital Markets Act will force Google to open key Android and search capabilities to rival AI assistants and provide anonymized search data under defined terms, while leaders at DeepMind, OpenAI and Anthropic publicly converged on independent pre‑release testing for frontier models—DeepMind proposed an industry‑funded, federally supervised evaluator. Implication: expect a faster fragmentation of standards, greater policy risk around model deployment and data access, and renewed antitrust/regulatory scrutiny of hyperscalers that could alter long‑term competitive dynamics and valuation assumptions for AI incumbents.