The week that just ended made a very specific argument: the market is not afraid of one thing, it is afraid of three things arriving at the same time, and it has no idea which one to price first. Yields pushed back toward multi-year extremes. Brent crude held near $90 with Hormuz transit traffic still a fraction of its pre-war norm. And semiconductor stocks spent five days selling off into the single most consequential earnings print on the 2026 calendar, NVDA's August 26 report. The S&P 500 closed Thursday at 7,641, down roughly 0.87% on the session, and Friday's partial recovery left the index nursing a meaningful weekly loss from the prior Thursday's all-time high. The read here is not that the thesis has broken. It is that three macro overhangs converged in one week, and the bill comes due Wednesday night.

The Lead: A Treasury That Couldn't Buy Its Way Out

The most consequential event of the week had nothing to do with earnings or geopolitics. On Wednesday, Treasury Secretary Scott Bessent announced that the department would at least double its long-term debt buyback operations, from $2 billion to $4 billion per operation, targeting 10-, 20-, and 30-year debt. The Treasury confirmed it had doubled the size of its long-term debt buybacks from $2 billion to at least $4 billion per operation. The market's initial reaction was a brief rally in long bonds and a soft dollar. Gold climbed. Bitcoin followed. Then Thursday happened.

Bond yields advanced on Thursday, with the yield on the 10-year Treasury note climbing back above the level it was at just before the Treasury Department said Wednesday morning that it would at least double repurchases of long-dated debt in the next few months. The market sent Bessent a clear message: buying duration from yourself does not change the supply-demand calculus when the primary deficit is what it is. Stocks fell as Treasury yields continued their march higher despite the buyback operation, raising concerns that higher borrowing costs will throw a wrench into the bull market, and the Dow Jones Industrial Average shed more than 700 points, weighed down in part by a steep drop in Walmart shares.

The structural problem the bond market is pricing is not mysterious: elevated energy costs feeding import prices, a fiscal trajectory with no credible consolidation anchor, and a Fed chair in Kevin Warsh who has given no signal he is in a hurry. As one senior market analyst put it, "markets are ending the week on a softer tone after the relative calm of early August was disrupted by renewed pressure in global bond markets, another rise in oil prices and growing uncertainty around the Federal Reserve's next move," with the Nasdaq among the weaker performers as "higher long-term borrowing costs once again challenge elevated valuations." The buyback was a signal of intent. The market read it as desperation.

The 30-year Treasury is not fighting the Fed. It is fighting the arithmetic of the deficit, $90 oil, and a geopolitical premium that has no obvious exit. Wednesday's NVDA print and Warsh's Jackson Hole remarks will not fix that, but they will tell us how much the AI thesis can absorb of it.

The Tape: What Moved, What It Means

AI Chips. The Philadelphia Semiconductor Index spent most of the week under pressure, consistent with the broader long-duration growth selloff driven by the rate shock. NVDA shares closed Friday near $216.85, leaving the stock at roughly 33 times trailing earnings, near a five-year low on that metric, with analysts projecting second-quarter revenue of almost $92 billion, up 97% year over year, slightly above the midpoint of the company's own guidance of $91 billion. We own NVDA. The compressed multiple is the one number in this whole setup that the bears cannot explain away: a company with a disclosed order book of $1 trillion for 2026 and 2027 against trailing twelve-month revenue of $253 billion is not trading at a speculative premium. It is trading below the Nasdaq-100's forward multiple. That is a price setup, not a thesis problem, and Wednesday night resolves it in one direction or another.

What to watch beyond the revenue headline: investors should pay close attention to guidance for the third quarter and commentary on the Vera Rubin platform launch timeline, because the transition from Blackwell to Vera Rubin represents NVDA's next major product cycle and any delays or supply constraints could impact sentiment. Vera Rubin processors are scheduled to begin shipping in the second half of 2026. The guide is the event. Revenue in line with a $91-92 billion consensus changes nothing about the next twelve months. A Vera Rubin update that is late, or guidance that implies any hiccup in the Blackwell-to-Rubin ramp, is the one scenario that turns Wednesday's print from a relief catalyst into another round of multiple compression. We do not expect that, but it is the tail that matters.

We also own AMD and MU, both of which absorbed this week's rate-driven selling. MU has been the more volatile name, and the thesis there is straightforward: memory is the second constraint in the AI compute stack, pricing is recovering, and the company's HBM ramp is a direct lever on hyperscaler capex. The setup into NVDA's print is constructive for the whole cohort if management's language around AI infrastructure demand remains unambiguous.

Power & Grid. GEV and the firm-power names, VST and CEG, held up notably better than the semis this week, which is exactly the rotation texture we have been describing all month. Energy and power were among the few pockets of relative strength in a tape that punished long-duration growth names. This is not coincidental. The power infrastructure thesis is, at its core, a contracted-revenue story with duration more like a utility than a growth stock, which makes it substantially less sensitive to the 30-year hitting 5.3%. The firm-power and grid-hardware names get paid regardless of whether the 10-year is at 4.5% or 5%. The AI capex that drives their backlog does not pause because Walmart missed comparable sales.

Critical Minerals. MP Materials is a position of ours, entered on a technical reclaim earlier in August, and the week was a strong one for the name. In the last few weeks, MP ran from a close near $38 on July 29 to about $60.05 on August 21, a move driven by the fundamental re-rating that followed Q2 results and sustained by the broader geopolitical narrative around Western critical mineral supply chains. Q2 revenue jumped 89% year over year to $108.5 million, topping the consensus, while the company posted an adjusted loss just at expectations. Management reported that upstream rare earth oxide production reached 13,500 tons in a single quarter, representing roughly 60-70% of the Upstream 60K target, and noted the company has advanced quickly with very limited capital spending relative to what was originally anticipated.

The thesis on MP is not a commodity trade. China produces roughly 90% of the world's rare-earth magnets and has used export controls to increase pressure on U.S. companies and supply chains, and beginning January 1, 2027, U.S. defense sourcing rules extend restrictions on Chinese-origin NdFeB magnets across the full supply chain from mining to finished production. That regulatory clock is the structural floor on MP's demand, and the company has guided Q3 NdPr production above 1,000 metric tons with realized pricing in the high $90s per kilogram, and expects first commercial magnet shipments to General Motors in Q4 2026. The stock has moved substantially from our entry. We are watching it on a disciplined basis for any thesis-relevant news rather than price noise, but the fundamental story here is intact and arguably gaining momentum as the defense-supply-chain mandate timeline draws closer.

Gold. GLD is a deliberate strategic position in our book, and this was the week it earned its keep most visibly. Gold futures climbed to a three-month high Friday on a soft dollar as investors weighed the Treasury's liquidity move to double bond buybacks, with the announcement initially pushing yields and the dollar lower and lifting gold prices, with December gold hitting $4,569.40 per ounce, its highest since mid-May. The contract is on track for a fifth straight weekly gain, its longest winning streak since October 2025, and gold is up nearly 5% for the week. The structural bid here is central-bank accumulation, fiscal debasement pressure, and a dollar that weakened on Bessent's buyback move. None of those forces disappear next week.

The Geopolitical Layer: Hormuz and the $90 Floor

The 60-day U.S.-Iran ceasefire memorandum expired Monday and Trump closed the door on extension. What followed was a week of oil near $90 Brent and strait traffic that remains a shadow of its pre-conflict norm. Shipping in the strait, a conduit for about one-fifth of global oil supplies before the war, has effectively collapsed since the conflict began in late February, and between eight and fifteen vessels crossed the strait on the most recently tracked days, a fraction of the roughly 130 transits before the conflict.

The U.S. Energy Information Administration, in its latest market outlook, said it did not expect oil production in the Middle East to return to near pre-conflict levels until early 2027, and for the price of Brent to average $87 a barrel in 2026. The EIA's $87 average implies the current $90 handle is already above their central scenario, which means any further diplomatic deterioration or shipping incident has an asymmetric upside effect on prices. The only partial offset this week was a report that Iran had finalized a preferential trade agreement with Oman, which some traders read as a back-channel that could eventually ease strait conditions, though the terms involve no firm commitments on transit rights for non-Iranian vessels.

For the fund, sustained $85-90 oil is a mixed signal. It supports our power pillar, where high energy prices increase the urgency around domestic firm-power capacity. It pressures input costs broadly. And it feeds the fiscal narrative that keeps the long end of the Treasury curve elevated. The oil risk here is not that prices spike above $100 again; it is that they stay exactly where they are, a persistent headwind to sentiment and consumer confidence, without being dramatic enough to force a diplomatic resolution.

The Setup: What Wednesday Decides

Next week contains more consequential events per square inch of calendar than any week since early July. Three of them will almost certainly determine the tone for September.

Wednesday, August 26, NVDA earnings after the close, plus GDP and core PCE at 8:30 a.m. The data arrives first. Core PCE is expected to rise 0.2% month-over-month in July, leaving the annual rate at approximately 3.3%, and a hotter-than-expected print would lift the probability of a September rate hike. If PCE comes in hot and NVDA delivers an ambiguous guide, Wednesday becomes a very difficult day. If PCE is in line and NVDA's Blackwell demand commentary is unambiguous, the setup for a Q4 re-rating in the AI pillar is real. The GDP second estimate for Q2 is also on the tape Wednesday morning, and the annual revision to nonfarm payrolls lands the same day. NVDA's results will remain a bellwether for global AI demand, while personal income and spending, PCE prices, and durable goods data for July, in addition to the annual revisions to nonfarm payrolls, are all awaited Wednesday.

Thursday, August 27, Chair Warsh speaks at Jackson Hole. The markets will focus primarily on the Jackson Hole Economic Symposium, where Federal Reserve Chairman Kevin Warsh plans to speak on monetary policy, with the main topic of concern for market participants being the likelihood and extent of further monetary policy tightening by the Fed this year and next. Warsh has been more hawkish in his public posture than Powell ever was, and the bond market's continued refusal to rally after the Treasury's buyback announcement suggests the market wants to hear directly from him on whether the fiscal-inflation dynamic changes his calculus. A data-dependent non-answer buys time. A signal toward additional tightening on top of the $90 oil backdrop would be a genuine shock. We are not positioned for the shock scenario, but we are watching NVDA's Wednesday print before drawing any conclusions about what Thursday's setup looks like for the growth names.

Also on deck: CRWD earnings Thursday. Wednesday's earnings lineup includes Salesforce, Synopsys, HP, and NVDA, while Thursday brings CrowdStrike, Marvell Technology, Workday, and Autodesk. CrowdStrike is not in our book, but its print is read directly into the enterprise security spending environment that supports PANW, which is a position of ours. A strong CRWD print with durable net new ARR tends to lift the cohort. Marvell's guidance will also be read as a secondary data point on AI chip demand after NVDA sets the tone Wednesday night.

The Read: Pillar Conviction Heading Into the Week

AI Chips (NVDA, AMD, MU). Conviction intact, price setup the most compelling it has been since the May correction. NVDA's forward earnings multiple of roughly 25 sits slightly below the Nasdaq-100's 26 times, though on a price-to-sales basis the stock still trades around 21 times trailing revenue. The multiple compression this week is entirely rate-driven, not demand-driven. NVDA delivered a record $81.6 billion in Q1 FY2027 revenue with data center solutions anchored by accelerated computing and AI platforms driving 92% growth, fueled by surging demand for Blackwell architecture in hyperscale and enterprise deployment. The Q2 print needs to clear a $91-92 billion bar and give the market a clean Q3 guide above $103 billion to restart the multiple expansion conversation. What would change this: a Vera Rubin delay of more than one quarter, or guidance that implies any softening in hyperscaler pull-through. We do not see either as the base case.

Power & Grid (GEV, VST, CEG). Thesis strengthened this week, not weakened. The rate shock and oil volatility that crushed long-duration growth names left this pillar relatively intact, exactly as the contracted-infrastructure argument predicted. GEV carries a backlog that does not care about the 30-year yield. VST and CEG provide firm power at contracted prices into a grid that is increasingly supply-constrained by AI data center demand. What would change this: a hyperscaler capex reduction announcement, or a meaningful Hormuz resolution that brings natural gas prices down sharply. Neither is imminent.

Critical Minerals (MP). The thesis is a chokepoint made into a contract: DoD price floor, Apple offtake, first GM magnet shipments in Q4. MP Materials is expanding a U.S. mine-to-magnet supply chain with Pentagon backing and a $500 million Apple agreement as demand grows across robotics, EVs, aerospace, and defense. The January 2027 defense sourcing rule change is the structural deadline that concentrates urgency. The stock's move from the high $30s to $60 in less than a month reflects a re-rating of that probability, not speculation. What would change this: a meaningful U.S.-China diplomatic détente that eases rare-earth export controls, or a 10X facility construction delay that pushes the GM shipment timeline beyond Q4.

AI Software / Security (PANW, NOW). PANW has been a strong performer and remains a core position. The CRWD print Thursday is the next read on enterprise security spending velocity. NOW remains a proxy for AI-agent infrastructure adoption in the enterprise. Both are sensitive to rate-driven multiple compression, but less so than the semiconductor names given their recurring-revenue characteristics. What would change the conviction: back-to-back quarters of deceleration in net new annual recurring revenue across the security cohort.

Fuel Cycle / Energy Security (LEU). Repairing. The HALEU enrichment and DOE award thesis has not changed, and the broader energy-security narrative is, if anything, stronger with Hormuz functionally constrained. The name requires patience and a long enough time horizon to let the domestic enrichment policy play out. It is a hold, not a headline.

GLP-1 / Healthcare (LLY). Unchanged and performing its intended function as an off-AI diversifier. LLY is a position we hold, and the metabolic-drug pipeline thesis is independent of rates, oil, and semiconductor cycles. The Q2 report established the baseline for the Zepbound volume ramp. The next meaningful catalyst is Q3 guidance, not this week.

Financials (BAC). The rate environment is, counterintuitively, supportive for net interest income at a name like BAC. A steeper long end benefits the bank's asset repricing profile. The risk is that 30-year yields above 5.3% create credit stress in commercial real estate or high-yield that shows up in reserve builds, but that is a Q3 earnings story, not this week's.

Gold (GLD). The structural bid is intact and the week confirmed it. Central-bank accumulation, a dollar that the Treasury just intervened to weaken via the buyback announcement, and a fiscal trajectory with no credible consolidation path all argue for a continued premium on the debasement hedge. We hold GLD as a deliberate, strategic uncorrelated position, and this week made the case for that allocation more clearly than any week this month.

Nova One Capital and the Nova X Convergence Fund hold positions in securities discussed in this letter, including NVDA, AMD, MU, GEV, VST, CEG, MP, PANW, NOW, LLY, BAC, and GLD. This standing disclosure applies to all named holdings referenced above.

This document is for informational purposes only and does not constitute investment advice, a solicitation, or an offer to buy or sell any security. The views expressed are those of Nova One Capital as of August 22, 2026, 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. Nova One Capital and the Nova X Convergence Fund may hold positions in the securities discussed. This letter is intended for sophisticated investors and is provided for informational and educational purposes only.