Thursday plan: Tilt AI exposure toward the compute suppliers
add AVGO and press NVDA while trimming the capex-heavy compute buyers as 'AI spend overshadows growth' bites
Wednesday's tape split AI in two: semis suppliers (NVDA +2.3%, AVGO +2.67%) were bid while mega-cap compute buyers were punished for capex after Alphabet and Tesla. Into Thursday the plan is to raise NVDA to 21% and introduce AVGO at 7%, funded by trimming an oversized META to 20% and the laggard AMZN to 8% plus a one-point LLY trim, with JPM held at full weight as the cyclical anchor. VIX eased to 16.64 but a live Hormuz energy premium keeps the tail open.
Accountability: Wednesday's plan leaned into semis and it half-worked — the NVDA up-weight caught a +2.3% day, but pressing META to 24% was punished with a -2.58% drop, and the book returned -0.4% versus SPY's -0.12%, lagging by a touch. The read carried into Thursday is that within AI the compute suppliers are being rewarded and the heavy-capex compute buyers are now on a shorter leash.
Since Last Session
Wednesday, 22 July, was a session of dispersion rather than direction. The book returned -0.4% against SPY's -0.12%, a modest underperformance. NVDA led the holdings at +2.3%, confirming that semiconductor leadership is intact, while JPM added +0.86% as financials stayed firm. The drag came from the internet names: META fell -2.58% and AMZN slipped -1.09%, and the quality anchor LLY gave back -1.05%. The net was almost entirely a function of our oversized 24% META weight meeting a soft tape for mega-cap AI spenders — the same tape that greeted Alphabet's imperfect-but-solid quarter and Tesla's margin miss with skepticism about how much of the capex is translating into growth.
Plan for Thursday
Into Thursday the intent is to sharpen the book's AI exposure toward the part of the theme the market is paying for. I plan to raise NVDA to 21% and introduce Broadcom (AVGO) at 7%, building a two-name compute-supplier core out of the day's strongest movers (NVDA +2.3%, AVGO +2.67%). Those adds would be funded by trimming an oversized META down to 20%, cutting the laggard AMZN to 8%, and taking a single point from LLY to 24%. JPM would hold at 20% as the cyclical anchor.
The logic is a bifurcation trade, not a wholesale reset: within AI, favor the suppliers of compute (NVDA, AVGO) over the heavy-capex buyers (META, AMZN) that are now being asked to justify their spend. For the open to confirm the tilt I want to see semis hold their bid and the mega-cap software/internet complex stabilize rather than accelerate lower; a further leg down in QQQ that drags the suppliers with it would argue for patience on the AVGO entry.
Positioning
The book stays fully invested and growth-tilted, with caution expressed through name selection and sizing rather than cash or hedges. The center of gravity is a compute-supplier core (NVDA 21%, AVGO 7%) plus a cyclical earnings anchor (JPM 20%) and a quality, non-AI ballast in LLY (24%) whose 60-day trend (+6.69%) is the book's strongest structural signal. META (20%) and AMZN (8%) remain live but deliberately smaller, sitting on the scrutinized side of the capex debate. There is no energy, gold, or duration position: the Hormuz premium is real but I express it by owning durable earnings power, not by rotating into a geopolitical hedge.
Institutional Signals
The filings corroborate the shape of the book more than any single trade. Dalio holds NVDA (3.7%) and AMZN (4.1%) inside a diversified sleeve; Burry's Scion runs a concentrated AI-adjacent book (PLTR 66%, NVDA 13.5%) — together they signal that AI conviction persists but is being expressed through concentration in the leaders. Ackman's concentration in AMZN (17.4%) and MSFT (15.3%) supports keeping AMZN live rather than exiting on price weakness. Buffett's quality-and-cash-flow tilt (AXP, KO, BAC 9.5%, CVX) reinforces the JPM financials anchor and the decision to respect the energy backdrop through discipline rather than chasing it. Net: institutions imply AI leadership is intact but should be owned through the highest-quality suppliers, not through indiscriminate mega-cap capex exposure.
What Could Break It
The thesis inverts on a genuine hyperscaler capex cut — that would hit the suppliers (NVDA, AVGO) directly and would not be price noise. It also weakens if upcoming mega-cap prints show ad or AWS deceleration that validates the 'spend overshadows growth' fear at META and AMZN. On the macro side, a Hormuz closure or a disorderly oil spike could flip the contained energy premium into a broad risk-off that drags growth indiscriminately, and a credit-spread widening beyond 25-30bps would impair the JPM leg. A VIX break back above roughly 20 from the current 16.64 would be the cleanest signal that the constructive-but-selective read is wrong and the book needs to de-risk within its growth mandate.