Here is what this week proved: the AI supercycle is not monolithic, and the market has started pricing the seams. The Dow closed at a record 52,900 on Thursday. The Nasdaq dropped 0.8% on the same session. The divergence is the signal. When 57,000 jobs print against a 115,000 consensus, the rate-sensitive, old-economy side of the market gets a tailwind and the momentum trade gets a repricing. That's exactly what happened, and chips were the epicenter of the revaluation, down nearly 4.5% on the SMH in a single session with the sector averaging a -10.5% week. Layer in Meta's Wednesday bombshell, a move into third-party compute sales that sent META up nearly 9% and CRWV down more than 15%, and you have a week that demands honest pillar-by-pillar accounting. The thesis is not broken. But the terrain shifted, and pretending otherwise is how you get caught wrong-footed into the late-July capex binary.
"The Dow at a record and the Nasdaq at a weekly low is not a contradiction, it's a message. Read it correctly: the market is repricing risk concentration, not abandoning the AI megatrend."
Pillar I: AI Chips / Hardware, The Profit-Taking Corrects, Not the Thesis
NVDA closed the week near $194, down roughly 7.7% for the four-day holiday-shortened stretch. AVGO shed approximately 6.9%. The Philadelphia Semiconductor Index, via the SMH proxy, averaged a -10.5% weekly move across the sector. The tech-heavy Nasdaq fell as investors dumped semiconductor names, taking profit after the swath of stocks surged more than 80% in the first half of 2026. That is the key context. This is not a thesis break, it is a regime where four months of 80%-plus gains invite half a week of institutional rebalancing when a soft macro print gives cover.
The non-consensus read here: the NFP miss is actually constructive for the chip thesis on a 90-day horizon. Warsh has maintained that artificial intelligence will ultimately have a disinflationary impact on the economy as rising productivity helps ease the cost of goods and services. A Fed chair who believes AI is inherently disinflationary, and who just lost his strongest argument for a July hike, is a Fed that just extended the duration window for every growth asset in the portfolio. The selloff is the market working through the quarter-end rotation. The fundamental AI capex cycle has not changed by a single dollar.
The critical pending data point: NVDA is expected to report earnings on August 26, 2026. AVGO is expected to report on September 3, 2026. Every week between now and late August is noise relative to those two prints. The thesis does not change on noise.
Pillar IV + V: The Meta Compute Story Splits Two Pillars at Once
Wednesday's Bloomberg report was the week's most consequential single event, and the market's split reaction made the thesis implications crisp. CoreWeave stock closed meaningfully lower on July 1 following a Bloomberg report that Meta Platforms is building a cloud infrastructure business to sell its excess artificial intelligence computing capacity to third-party customers. Meta shares surged nearly 9% on the same news, as investors reframed the company's massive capital expenditure program, guided at $125 billion to $145 billion for 2026, as a potential new revenue stream rather than a pure cost center. That single data point is the letter's most important asymmetry.
For META (Pillar IV: AI Software / Platforms), the market just repriced the capex narrative from liability to asset. A $125–$145 billion CapEx commitment that could generate third-party cloud revenue is a different multiple than one that doesn't. News broke Wednesday that Meta is going to start renting out some of its computing infrastructure, pitting the social media company against smaller cloud providers. JPMorgan is cautioning against the move, saying it would prefer that Meta boost its own capabilities on query response, also called "inference", to develop its advertising business. JPMorgan's objection is reasonable at the margin. But the market's 9% response reflects something different: the realization that Meta has built one of the world's most capable AI compute platforms, and the optionality around that asset is only beginning to be priced.
For CRWV (Pillar V: Specialized GPU Cloud), the same event is a legitimate competitive threat, but the selloff is at least partially overdone on first-principles analysis. Despite reporting strong Q1 revenue growth of $2.08 billion and a substantial revenue backlog of $99.4 billion, investor sentiment has shifted negatively. The company's net loss widened to $740 million, driven by rising infrastructure costs and aggressive expansion spending, leading to profitability concerns. The backlog of nearly $100 billion provides contractual revenue visibility that a nascent Meta compute-rental business cannot immediately threaten. The company was built from the ground up specifically for machine learning, training, inference, and agentic AI workloads, giving it a developmental head start over Meta's nascent third-party infrastructure business. Meta Platforms has not officially confirmed the plan either, and the details around pricing, launch timelines, and target customers remain unspecified.
The non-consensus read: Meta entering the compute market is a demand signal, not just a supply signal. The world's largest social media company is telling you AI inference demand is so massive that even after stuffing their own workloads, there is surplus to monetize. That is a bullish statement for total addressable compute demand. The risk for CRWV is customer concentration and leverage, not the existence of demand.
Pillar II: Nuclear / Power, The Quiet Pillar in a Noisy Week
Power infrastructure had the week's cleanest tape relative to expectations. VST held near the critical $150–$151 guardrail level, the technical line between "thesis intact" and "thesis in question." VST did not flush. That matters: in a week where the Nasdaq shed nearly a full percent and chips fell nearly 7%, the power infrastructure pillar showed exactly the non-correlation the thesis requires.
The structural driver has not changed. CoreWeave surpassed 1 gigawatt of active power capacity and secured major financing support, including an $8.5 billion loan facility. Every gigawatt of active AI compute that CoreWeave brings online, and the broader hyperscaler buildout across AWS, Azure, and Google, is incremental demand for reliable baseload power that the grid cannot currently supply at scale. That demand flows directly to VST and CEG.
The LEU DOE two-year option remains unconfirmed. Nasdaq officially announced that SpaceX will be added to the Nasdaq-100 index before the market opens on July 7. The LEU mid-July recheck is the next binary for the nuclear sub-pillar. The DOE follow-on contract status should confirm by approximately July 13, and that is the catalyst that resolves the thesis one way or the other.
Macro: The Fed's July Problem Just Got Smaller
The June NFP print, covered in depth in Thursday's special edition, did more than kill the July hike. It recalibrated the entire second-half rate path. The FOMC's June 16–17 meeting left the federal funds rate unchanged at 3.50%–3.75%, consistent with the pause maintained since the December 2025 cut, while the updated dot plot shifted hawkishly: the median year-end 2026 projection rose to 3.8%, implying limited room for easing or even a possible hike. That 3.8% median was contingent on a labor market that continued to run hot. 57,000 jobs in June, with prior months revised down a combined 74,000, is not a hot labor market. The hike arithmetic no longer closes.
Traders are now focused on upcoming data releases, including the July CPI and the July 28–29 FOMC meeting, for signals on whether inflation momentum will ease enough to reopen the door to cuts later in the year. The July 28–29 FOMC meeting is now the single most important date on the macro calendar, and it arrives the same week as the hyperscaler capex prints that drive the AI infrastructure complex. The next three weeks are about data collection ahead of that confluence.
The SpaceX / Nasdaq-100 Rebalance: What It Actually Means for Holdings
Monday's open brings a structural event that has nothing to do with AI earnings and everything to do with passive fund mechanics. SpaceX will join the Nasdaq-100 Index before the market opens on July 7, 2026. Analysts estimate the inclusion could trigger around $4.3 billion in passive inflows as index funds are required to buy the stock.
Here is the part CNBC will not tell you: to fund that $4.3 billion purchase, every QQQ and QQQM manager in the world must proportionally reduce their existing positions. To fund that purchase, QQQ must sell proportional slices of every current constituent. Apple, Nvidia, Microsoft, Amazon, Alphabet, Meta, Broadcom, all are being reduced slightly to make room for SpaceX. That mechanical selling creates a technical headwind for NVDA and AVGO on Monday and Tuesday, on top of the profit-taking pressure already in the tape from this week. This kind of index-rebalancing flow is mechanical noise, not a fundamental signal about the chip names.
SpaceX's public float today is only 3%–5% of total shares outstanding. Starting around Q2 earnings expected mid-July to September 2026, a staggered lockup expiration schedule begins releasing insider and pre-IPO investor shares into the market: approximately 20% of locked shares released at Q2 earnings. The lockup calendar is the next structural risk for SPCX after the rebalancing sugar rush fades, and it is worth watching from the sidelines.
What to Watch and Why It Matters
| Date | Event | Fund Relevance |
|---|---|---|
| Mon Jul 7 | SpaceX (SPCX) enters Nasdaq-100 at open, $4.3B forced passive inflows | Mechanical rebalancing pressure on large-cap names, a technical flow rather than a fundamental signal. |
| Mon Jul 7 | Markets reopen post-Independence Day holiday | First read on where the chips sit versus their long-term trend after the holiday. |
| ~Mid-week | Fed speakers post-NFP (watch for Warsh or vice chairs) | Any pivot language on July hike probability is a direct read for the duration-sensitive pillars. |
| Wed Jul 9 | PepsiCo (PEP) earnings, first major Q2 print | Not a portfolio name, but a consumer pricing-power read-through for inflation trajectory and Fed path. |
| ~Jul 13 | LEU DOE 2-year option confirmation window | Hard recheck date. Confirmed exercise validates the thesis; non-renewal plus a broken commercial path breaks it. |
| Jul 28–29 | FOMC meeting: Warsh's second as chair | Master binary. A hike pressures duration and multiples; a hold with a dovish lean is a green light for the AI infrastructure trade into late-July capex prints. |
| Late July | Hyperscaler Q2 earnings + NVDA capex commentary | The key checkpoint. AWS, Azure, Google CapEx guidance governs the chip and power pillars for H2. |
The week ahead is not about trading, it is about gathering information. Three technical signposts to watch: (1) NVDA and AVGO both above their 200-day moving averages; (2) VST holding above $150; (3) no acceleration of CRWV's slide toward the $70 area. If all three are intact by Friday close, the thesis enters the capex binary on solid technical footing.
Pillar I: AI Chips / Hardware (NVDA, AMD, AVGO)
The late-July capex print is the next decision gate for the chip complex. The SPCX rebalancing headwind Monday is passive noise, not a thesis event. Patience is the discipline here.
Pillar II: Nuclear / Power Infrastructure (VST, CEG, LEU)
The $150 guardrail on VST is the technical line between thesis intact and thesis in question. CEG absorbed its post-lockup dip without breaking its structure. LEU is a high-risk, high-reward nuclear-fuel play riding into the mid-July DOE confirmation.
Pillar III: Digital Finance Rails
The GENIUS Act passage remains the structural catalyst for this pillar. Nothing from this week changes that calculus.
Pillar IV: AI Software / Platforms (META, NOW)
The cloud compute announcement is a thesis amplifier: the read is that Meta's CapEx is transitioning from cost center to revenue asset, which is structurally bullish for the AI monetization thesis. NOW needs to reclaim its 50-day with conviction to confirm its enterprise-software setup.
Pillar V: Specialized GPU Cloud (CRWV)
The Meta compute overhang, combined with CRWV's leverage and customer-concentration profile, makes this the highest-risk name in the group. The $99.4 billion backlog is a real asset; the competitive narrative is a real risk.
This document is for informational purposes only and does not constitute investment advice. The Nova X Convergence Fund letters reflect the views of the fund's management team as of the date of publication and are subject to change without notice. Past performance is not indicative of future results. All investments involve risk, including the possible loss of principal. Nothing herein should be construed as a solicitation or offer to buy or sell any security. July 4, 2026.