Jul 24, 2026
CautiousMacroRegime · AI leadership keeps bifurcating: compute suppliers (Intel posting its fastest revenue growth in ~15 years, semis, NVDA/AVGO momentum) are rewarded while heavy-capex buyers are repriced — Alphabet and Tesla shed hundreds of billions post-earnings, yet Google Cloud's 50% customer-spend growth confirms the supplier demand is real, not narrative. Layered on top is a hawkish shift: the 10-year yield hit its highest since January 2025 and Fed rate-HIKE odds surged as oil ripped on supply friction (jet-fuel logistics stress, tanker/Hormuz risk), with Japan core CPI creeping up on higher oil. VIX firmed to 18.7. Higher real rates pressure long-duration tech multiples and hit the highest-multiple capex spenders hardest, while financials (JPM 60-day +9.59%) benefit from higher NII and elevated trading volumes.

Friday plan: Press the compute suppliers and exit AMZN as rising oil, higher

yields, and the 'AI capex overshadows growth' repricing punish the heavy spenders

Into Friday the book leans harder into the AI names being paid — NVDA and AVGO — validated by Google Cloud's 50% spend jump and Intel's fastest revenue growth in 15 years, while it exits AMZN and trims META, the capex-heavy buyers caught on the wrong side of both the Alphabet/Tesla repricing and a hawkish rate shift. JPM and LLY stay as rate-beneficiary and quality anchors. Fully invested, no hedges.

Alphabet and Tesla lost hundreds of billions post-earnings, deepening the 'AI capex overshadows growth' repricing — but Google Cloud's 50% customer-spend growth validates that supplier demand is real10-year yield hit its highest since January 2025 and Fed rate-hike odds surged as oil ripped, a hawkish shift that pressures long-duration tech multiples and the highest-multiple capex buyers mostCompute suppliers are being paid — Intel posted its fastest revenue growth in nearly 15 years riding the AI boom, and semis/custom-silicon momentum held up against the broad tape

Accountability: Thursday's plan — tilt into the compute suppliers, add AVGO, press NVDA, and trim the capex-heavy buyers as 'spend overshadows growth' bites — was directionally right. The book fell just -0.87% versus SPY's -1.23%, a 36bp cushion, and the split played out exactly as framed: the capex buyers META (-3.36%) and AMZN (-4.57%) were the drags while LLY (+1.97%) and JPM (+0.49%) held green. The one imperfection was that the suppliers didn't rally on the day — NVDA (-1.56%) and freshly-added AVGO (-1.09%) dipped in the broad risk-off — so the thesis was validated on the short side more than the long side.

Since Last Session

Thursday, 2026-07-23 was a broad risk-off session: SPY fell -1.23% and the book fell -0.87%, outperforming by 36 basis points. The dispersion inside the book was the story. Eli Lilly led at +1.97% and JPMorgan added +0.49%, the two names with the least exposure to the AI-capex debate. On the other side, the heavy compute buyers were punished — Meta dropped -3.36% and Amazon fell -4.57% — as Alphabet and Tesla shed hundreds of billions in post-earnings value and the 'AI capex overshadows growth' repricing widened. The suppliers were not spared in the tape either, with NVDA off -1.56% and AVGO off -1.09%, but their weakness looked like beta rather than a thesis break. Underneath, the 10-year Treasury yield rose to its highest since January 2025 and odds of a Fed rate hike surged as oil ripped higher on supply friction.

Plan for Friday

Into Friday the plan is to lean the book further toward the part of AI the market is actually paying, and to reduce exposure to the part it is punishing. The intended moves: press NVDA to 24% and deepen AVGO to 13% as the two supplier legs, fund those adds by exiting AMZN entirely and trimming META to 18%, nudge JPM up to 22% as the rate-and-volatility beneficiary, and hold LLY at 23% as the quality anchor. The read-through justifying the supplier tilt is concrete — Google Cloud reported customers spending 50% more and Intel posted its fastest revenue growth in nearly 15 years, both confirming that the infrastructure demand behind NVDA and AVGO is real earnings, not narrative. AMZN leaves the book because it sits on the wrong side of both live themes: it is the weakest structural name (60-day -7.67%), the worst single-day mover Thursday, a capex-heavy buyer in the penalty box, and a long-duration name exposed to the rate shift. The open should confirm that suppliers are stabilizing relative to buyers and that a further leg higher in yields is not forcing indiscriminate de-grossing across all of tech.

Positioning

The book stays fully invested with no cash and no hedges, expressing caution through name selection rather than by retreating to defensives. It concentrates in five high-conviction ideas: two validated AI suppliers (NVDA 24%, AVGO 13%), a quality growth anchor insulated from the macro cross-currents (LLY 23%), a rate-and-volatility beneficiary (JPM 22%), and a right-sized ad-franchise holding whose capex is still awaiting validation (META 18%). This is a deliberate continuation of the live book, not a reset — the same core names carried forward, with capital rotated from the deteriorating capex buyer (AMZN) and from a reduced META into the suppliers being paid. The tilt deliberately favors exposures that can survive a bad AI-capex headline: LLY's earnings are non-AI, JPM actually benefits from the rate move, and the suppliers are being paid on real cloud demand.

Institutional Signals

The filings reinforce rather than contradict the tilt. Buffett's continued heavy financials exposure (BAC 9.5%, AXP 17.4%) corroborates the quality-cyclical JPM position, and his Chevron stake reflects the same energy-inflation backdrop now bidding oil. Dalio's book holds NVDA (3.7%) alongside broad index exposure, consistent with keeping the supplier leg. Burry's Scion remains concentrated in NVDA (13.5%) and PLTR, a conviction read on the compute-supplier side. Ackman's large AMZN position (17.4%) is the one cross-current — it argues the AWS franchise is durable long-term — but 13F data is slow-moving positioning context, not a daily instruction, and it does not override AMZN's clear near-term breakdown on both the capex and rate axes. Nothing in the filings argues for adding defensives or hedges here.

What Could Break It

The view is invalidated if the rate shift turns into an indiscriminate de-grossing that drags the suppliers down with the buyers — a sign the market is repricing all of tech's multiple, not just capex efficiency. A genuine hyperscaler capex cut (not price noise) would break the NVDA/AVGO supplier thesis directly. On the other side, if oil rolls over and Fed hike odds fade fast, the punished mega-cap buyers could snap back hard and the underweight to META plus the AMZN exit would cost relative performance. For JPM, the risk is a disorderly credit-spread widening beyond 25-30bps rather than the rate move itself. For LLY, an adverse CMS action or a credible tirzepatide biosimilar filing would be the trigger — neither is present today.