The week gave you everything at once: a Federal Reserve that held rates but nearly didn't, with three regional presidents voting openly for a hike; a bond market that briefly treated the dissenters' view as the new baseline and pushed the long end to multi-year highs; and then, directly contradicting the gloomy rate narrative, an Amazon earnings print that showed AWS growing at its fastest pace in 18 quarters, faster than Azure, faster than Google Cloud a week prior, and backed by $220 billion in 2026 capital spending. The week ended with the Dow and S&P each up roughly 1% and the Nasdaq recovering nearly all of a brutal Monday-through-Wednesday drawdown. The surface read is "AI wins again." The more honest read is that the market is now navigating two distinct risks in parallel: a Fed that may tighten further, and a labor market that is cooling in ways that could make cuts necessary for the wrong reasons. That tension, cuts versus recession, is the organizing frame for everything that matters next.

"AWS at 37% growth, the fastest in 18 quarters. Amazon raised capex to $220 billion and said even that won't meet demand. The AI infrastructure funding debate is settled. What isn't settled is what a softening labor market does to it."

The Fed's Family Fight Has Real Consequences

The FOMC voted 9-3 on Wednesday (Jul 29) to hold its benchmark rate in the range of 3.5% to 3.75%, with three regional presidents: Beth Hammack of Cleveland, Neel Kashkari of Minneapolis, and Lorie Logan of Dallas, dissenting in favor of a quarter-point hike, citing inflation that has remained above the Fed's 2% target for more than five years. That is not a routine dissent. Three hawks openly calling for tighter policy, in a room chaired by Kevin Warsh, tells you the internal debate is genuinely unsettled, not a performative "family fight" the chairman can wave away with a memorable phrase.

When a reporter asked Warsh whether this should be characterized as a "pause," he pushed back directly: "I wouldn't characterize what we did as anything like a pause. I would characterize what we did as a rigorous review of the economic situation... a view of the big hard questions." That framing matters because it signals the committee sees itself as genuinely in motion, not waiting out the data, but actively weighing it. Recent price pressures have reflected both tariffs imposed under the current administration and higher energy costs tied to the Iran conflict. The inflation problem is not clean; it has a structural component that rate hikes can barely touch, and a cyclical component that the labor data may now be resolving on its own.

The Fed's next FOMC meeting is scheduled for September 15-16, with the rate decision due on the second day. Between now and then, the committee will see the July jobs report (due August 7), one more CPI print, and Jackson Hole, where, per CNN's reporting, Warsh may stay on the sidelines rather than use the platform to signal a major course correction. If he declines to speak, that silence will itself be a signal, one the bond market will interpret aggressively. The two-year yield's behavior the week of August 10 will tell you more than any Fed speech.

What Moved, What It Means, and Where the Thesis Stands

The hyperscaler prints this week are the most consequential data points for the chip pillar since NVDA's last earnings report, and they cut in a specific direction: the demand is structural, the capital commitments are real, and custom silicon is becoming a growth line in its own right. Amazon CEO Andy Jassy described AWS as "booming," growing 36.7% year-over-year in Q2, its fastest growth in 18 quarters, with its AI and chips businesses each exceeding annual revenue run rates of more than $25 billion. That last number matters: Amazon's in-house Trainium and Graviton chips are now a meaningful business, which the market has been slow to price into the broader custom-silicon thesis that runs through AVGO's networking and custom-accelerator franchise.

Alphabet reported Google Cloud growth of 82% in its prior quarter, while Microsoft's Azure cloud revenue rose 43% during its fiscal fourth quarter. Three clouds, accelerating simultaneously, that is not a share-shift story, it is a market-expansion story, and it requires the same physical substrate: NVDA GPUs, MU HBM memory, and AVGO networking silicon. The chip pillar's conviction is intact. The master risk tell remains a decisive weekly break of NVDA's long-term trend line, not a headline, not a single session.

GEV reported the week prior (Jul 22), but the numbers belong in this letter because the bond-market reaction to the FOMC held the stock back from what the fundamentals warranted. Here is what actually happened: Q2 orders reached $24.2 billion, up 88% year-over-year, with a book-to-bill ratio above 2x. Revenue increased 22% year-over-year to $11.1 billion. Free cash flow for the quarter was $5.1 billion, up from $194 million in Q2 2025, with backlog at quarter-end reaching $176.3 billion, up 37% year-over-year. A company generating $5 billion in free cash flow in a single quarter, with a backlog that grows by $13 billion sequentially, is not a company facing a thesis problem, it is a company the bond market temporarily mispriced because rising yields compressed multiples across the capital-intensive sector.

GEV raised its 2026 revenue guidance to $45.5-$46.5 billion (up $1 billion from prior outlook) and increased full-year free cash flow guidance to $11.5-$12.5 billion, up from $6.5-$7.5 billion. The quarter was mixed, adjusted EPS of $2.47 missed Wall Street's $3.04 forecast, while revenue of $11.1 billion topped expectations of $10.73 billion. The EPS miss was the headline; the 88% order growth and the $176 billion backlog were the thesis. We treat the pullback as a price disconnect, not a structural break. GEV now expects at least 125 GW of gas equipment under contract by year-end 2026, with annual gas turbine output on track to reach 20 GW in Q3 2026, scaling to 24 GW in 2028 and 30 GW in 2030. The ramp is not slowing. It is accelerating.

The firm-power and wires thesis, VST and CEG on the merchant side, the grid infrastructure layer behind them, is the current lead theme for a straightforward reason: most of the increase in hyperscaler capital spending will support AI infrastructure, and that infrastructure runs on continuous 24/7 power that intermittent sources cannot provide. The binding constraint in the AI buildout has shifted upstream into transformers, switchgear, and high-voltage transmission, a 2-3 year lead-time bottleneck that no software update resolves. GEV's 88% order surge is the physical embodiment of that constraint clearing the pipeline.

Apple and Amazon reported on July 30 and received very different verdicts: Amazon's shares rose in late trading while Apple's fell. Apple dropped roughly 7% on the day, after chip shortages raised costs and reduced June-quarter production, weighing on earnings and guidance. CEO Tim Cook stated that Mac and iPad prices will increase due to the shortage and that iPhone prices may increase in the future. Apple is not in our book, but its supply-constraint problem is diagnostic: the same memory and advanced packaging shortage that Apple cannot route around is the shortage that supports pricing power for the memory and chip architecture names we own. The scarcity that hurts device makers is the scarcity that sustains margins for infrastructure suppliers.

LLY's Q1 2026 EPS of $8.55 beat estimates by 22.67%, with revenue of $19.8 billion exceeding forecasts by 12.5%. Mounjaro and Zepbound generated combined revenue of $12.8 billion, fueling 56% year-over-year growth in the cardiometabolic portfolio. The company raised full-year 2026 revenue guidance to $82-$85 billion. Q2 is the next chapter: LLY is scheduled to report before market open on August 5, with the consensus penciling in a step-up in the GLP-1 franchise. The Q1 beat-and-raise cadence sets a high bar, and given that the stock has rallied substantially since April, any guidance miss on volume will be punished swiftly. A name we own; we are watching the GLP-1 volume trajectory, not the headline EPS, as the tell.

Centrus (LEU) reports this coming week, the earnings date falling directly in the window when the domestic enrichment and HALEU thesis is most visible. The frame has not changed: this is a reshoring and national-security story, a bet on domestic uranium enrichment supply chains that the federal government has already underwritten through contracting. The HALEU program is a strategic imperative that survives any single SMR timeline slippage. We expect the call to provide updated production status on the Piketon HALEU demonstration; that is the detail to watch, not the quarterly revenue number in isolation.

The Apple Miss Is the AI Infrastructure Bull Case in Disguise

Here is the read you will not get from the standard earnings recap: Apple's chip-shortage warning and its ~7% single-session decline on Friday (Jul 31) are not a cautionary tale about the AI trade, they are the structural argument for owning the infrastructure layer instead of the device layer. Both Apple and Amazon depend on scarce chips, memory, and data-center equipment. Yet the same shortage affects them differently, Amazon buys that equipment to create capacity that customers can rent repeatedly. Apple sells a device; if the device cannot be made, the sale is lost. Amazon builds infrastructure; if the infrastructure is supply-constrained, the demand queues up and the pricing power accrues to the supplier.

The winners in a supply-constrained chip cycle are the ones with contracted capacity and multi-quarter lead-time protection, which is precisely the thesis behind the AI chip names in our book and behind GEV's 88% order surge in turbines and electrification gear. The hyperscalers' earnings week demonstrated that the market can no longer accept capex as demand unproven; Microsoft and Amazon proved the demand justified the capex and were rewarded, while Meta and Alphabet were punished for not proving it as convincingly. The market is now grading on a sophisticated rubric: show us the demand verification, not just the spending. Amazon passed. Apple, as a device company caught in the same supply crunch, showed what happens when you are on the wrong side of the constraint.

The non-consensus implication: the supply crunch Apple is navigating is a persistent tailwind for MU's HBM pricing and for the custom-silicon franchise at AVGO, where every dollar of memory and networking silicon that goes into a hyperscaler data center is a dollar that is not available to device manufacturers at competitive prices. Scarcity is a friend to infrastructure suppliers. It always has been.

What's On Deck: The Week of August 3

The calendar next week is as consequential as any we have seen this earnings season. Three distinct catalysts matter for the pillars.

LLY (August 5, before open). Eli Lilly reports before market open on August 5. Mounjaro and Zepbound generated combined revenue of $12.8 billion in Q1, fueling the cardiometabolic portfolio's 56% year-over-year growth. The Q2 bar is high. What we are watching is not headline EPS but GLP-1 unit volume, specifically whether the sequential volume ramp from new manufacturing capacity is accelerating or plateauing. A volume shortfall against supply build-out would be the tell on near-term pricing pressure. The stock has moved substantially since April; the options market is pricing an 8%+ move in either direction.

LEU (this week). Centrus's earnings call will be the first public update on HALEU production status at Piketon since the large federal contract was confirmed over the summer. The question is not quarterly revenue, it is whether the demonstration program is on track to scale, and whether any additional government offtake commitments are visible on the horizon. A clean call with production-on-track language is the bull case; any delay language would move the stock in the opposite direction quickly given the thin float.

July Jobs Report (August 7, 8:30 a.m.). The Employment Situation for July 2026 is scheduled to be released on August 7 at 8:30 a.m. Eastern Time. This is the most consequential macro print between now and the September FOMC. The June print came in at 57,000, well below the downwardly revised 129,000 in May and forecasts of 110,000, the lowest job gain in four months. A second consecutive soft print would shift the September debate from "hike risk" to "cut or hold," accelerating the two-year yield's decline and providing a tailwind to rate-sensitive infrastructure valuations. A surprise beat, on the other hand, hands ammunition directly to Hammack, Kashkari, and Logan. The three dissenters are not going away, they are waiting for data.

Jackson Hole (August 21-23). Far enough out to be in the planning window rather than the immediate calendar, but Warsh's posture at the symposium, whether he speaks, and what he signals, will set the regime narrative for September. The current reporting suggests Warsh may stay on the sidelines at Jackson Hole. If he does, the two-year yield will interpret his silence as optionality preserved, probably to the downside.

Technical levels to watch: The S&P 500 closed the week around 7,437, recovering above its 50-day moving average after the mid-week FOMC dip. The S&P 500 Equal Weight notched a fresh all-time high this week, even as the market-cap weighted index dropped below its 50-day SMA at points, a dynamic that speaks to the continued broadening of the rally beyond mega-cap tech. That breadth signal is constructive and reduces the tail risk of a market break driven purely by AI multiple compression.

Forward Thesis by Pillar: Conviction and What Would Change It

AI Chips / Hardware. Conviction is high and the week's prints reinforced it structurally. Amazon's total Q2 net sales reached $200.6 billion, with AWS revenue at $42.2 billion, an annualized revenue run rate of $169 billion. AWS growth accelerated from 28% in Q1 to 37% in Q2, and the AI and chips divisions each brought in over $25 billion in annualized revenue, more than doubling from a year ago. These numbers validate the contracted-capacity thesis: the hyperscalers are not pausing, they are building faster than they can source the compute. What would change conviction: a decisive multi-week break of NVDA's long-term trend line, or a hyperscaler that cuts guidance citing demand softness rather than supply constraint. We have seen neither. The risk remains the macro overlay, a recession would reduce enterprise AI software spend, though it would take longer to dent the physical infrastructure already under contract.

Power & Grid. GEV's 88% order growth and $176.3 billion backlog are not statistics that deteriorate in a quarter. GE Vernova achieved 61% year-over-year growth in adjusted EBITDA, generated approximately $10 billion in free cash flow year-to-date, and raised both revenue and free cash flow guidance for 2026. The refined thesis, firm power and the wires, not panels, is being confirmed by every hyperscaler capex announcement: the physical bottleneck is electrical infrastructure with 2-3 year lead times, not solar panels with 6-month lead times. VST and CEG are the merchant-firm-power expression of that constraint on the grid demand side. What would change conviction: a sustained 30-year Treasury yield above 5.5% that structurally reprices infrastructure multiples, or a FERC Order 1920 reversal that removes the transmission investment mandate. Neither is in view.

GLP-1 / Healthcare. LLY is an off-AI diversifier where conviction is moderate pending Tuesday's print. The Q1 beat-and-raise cadence is extraordinary: EPS of $8.55 beat estimates by 22.67%, revenue of $19.8 billion exceeded forecasts by 12.5%, and guidance was raised to $82-$85 billion for the full year, but the stock has reflected much of that upside. The August 5 print is the inflection point: if volume is accelerating and the manufacturing ramp is showing up in the numbers, the guidance raise will extend the thesis through year-end. If volume disappoints against the supply build-out, we get a repricing that the options market is already pricing at roughly 8%.

Fuel Cycle / Energy Security. LEU is a reshoring and national-security story, and the government contracting backstop is the floor under the thesis. The Piketon HALEU demonstration program is the operational catalyst, not an SMR-timeline bet, not a uranium spot price trade. What would change conviction: a formal program delay announced by the Department of Energy, or a federal budget rescission of HALEU appropriations. The former is a real risk; the latter is not plausible in the current defense-and-energy-security political environment.

Financials / Rate-Regime. BAC is the deliberate uncorrelated leg in the book. The current rate environment, held at 3.5%-3.75% with three dissenting hawks and a September meeting that is live for either a hike or a cut, is the kind of two-sided ambiguity that suits a large money-center bank with a clean quarter behind it. If September resolves dovish (cut), the steepening yield curve is a net interest income tailwind. If it resolves hawkish (hike), the short end rises further but the bank's asset repricing follows. The macro uncertainty is a feature, not a bug, for this position.

The overarching read: The week confirmed the demand thesis for AI infrastructure and revealed a Fed that is more hawkish in its internal distribution than the headline hold suggests. Those two facts are not in direct conflict, a higher-for-longer rate environment slows enterprise software spending at the margin but does not cancel gas turbine orders that are already on the books with 2-3 year lead times. The hyperscalers will spend $220 billion (Amazon alone) regardless of where the two-year yield settles in September. The physical infrastructure gets built. The power and compute layers get paid. The thesis is the thesis. What the cuts-versus-recession tension adds is a dose of intellectual honesty about the macro backdrop: the labor market is cooling, and if it cools further, the market's concern will shift from inflation to growth. We treat that as two-sided and do not chase the whipsaw in either direction.

Disclosure: Nova One Capital and the Nova X Convergence Fund hold positions in securities discussed in this letter, including names in the AI infrastructure, power and grid, GLP-1, and energy security pillars. This document is for informational purposes only and does not constitute investment advice or a solicitation to buy or sell any security. All views reflect the author's analysis as of August 2, 2026. Past performance is not indicative of future results. Investors should conduct their own due diligence and consult a qualified financial advisor before making investment decisions.