The hike was a formality. The dot plot was the event. On Wednesday, the FOMC raised its target range 25 basis points to 3.75%-4.00%, the first increase since 2023 and the first of Kevin Warsh's tenure as chair. The vote was 12-0. The statement was three paragraphs and blunt: prices remain elevated, the action supports a timelier return to 2%, and the committee "will deliver price stability." Then the projections landed, and the market understood the week it was actually in. Sixteen of eighteen participants see at least one more hike this year. Four pencil in two. Futures, which had the September move substantially priced, repriced the December meeting and pushed the implied terminal rate to roughly 4.2% by year-end and 4.6% by September 2027. The 10-year Treasury touched 5.04% on Monday before the meeting, its highest intraday level since 2007, then closed Friday back at an even 5.00% as traders chewed through what the new dot map means for everything with a duration.

The Repression Thesis Meets Its Falsifier

This letter has been transparent about the central tension in the book since August. The fund is built on a financial-repression frame: a debt stock too large to service at genuine real rates, which eventually pressures the Fed to hold short rates below inflation while Treasury manages the curve with buybacks. That was the thesis. This week, the tape argued the other side out loud.

Real short rates are now demonstrably positive. Three-month bills sit near 3.91% against a 3.4% headline CPI print (August, released September 11). That spread isn't wide, but the direction matters. A Fed that hikes unanimously, publishes a dot plot skewed toward more, and lets Chair Warsh say nothing at all in his own dot, is not a Fed that is being pressured to capitulate. It's a Fed that is, for the moment, winning. The repression frame doesn't break overnight. But the scenario in which it simply never arrives, in which real rates stay positive long enough to reset duration expectations, is now the live competing hypothesis and not a tail risk.

The most important line from the Warsh statement wasn't a sentence. It was the absence of one. Warsh hasn't submitted a dot since taking the chair, which means the committee's forward projections are his colleagues' views, with no constraint he has placed on himself. That matters. It leaves him free to move faster or slower than 16 of 18 dots suggest, and markets have almost no ability to model which direction that asymmetry cuts. The dollar index pushed back toward 100 in the hours after the announcement, with EUR/USD near 1.1470 and GBP/USD near 1.3385. That's the first real dollar firmness since late July, and it doesn't help the hard-asset sleeve.

The dot plot was a warning. The one-and-done narrative is already dead on arrival.

So where does that leave conviction? On the AI infrastructure pillars, it reinforces the case for contracted, supply-constrained compute over long-duration software multiples. On the hard-asset sleeve, it forces an honest reckoning. On the rate-sensitive financials, it splits the baby in a way the market hasn't fully processed yet, and that split is worth naming precisely.

What Moved, and Why It Matters for the Pillars

Monday was the week's sharpest single-day stress test. The Philadelphia Semiconductor Index dropped 5.9%, its worst session in months, as the 10-year briefly broke 5% and a separate, stranger catalyst landed simultaneously: executives at Anthropic and OpenAI called publicly for a slower pace of AI development on safety grounds. NVDA fell 3.4% on the day, MU dropped more than 5%, and AMD shed more than 4%. By Friday, NVDA had clawed back most of it, closing up 1.33% and finishing as one of the Dow's top gainers on the week.

The AI safety commentary deserves the same reading the thesis gets. Dario Amodei and Sam Altman don't control hyperscaler capex budgets. The companies that are actually committing to $1.3 trillion in aggregate AI infrastructure spend next year, the top-five hyperscalers whose guidance underpins NVDA's revenue outlook, are not announcing slowdowns. They're signing contracts. The Monday selloff was a sentiment event triggered by a rate shock dressed up with a narrative. The thesis didn't move. The price did.

MU reports September 30. That print will be the first hard data on whether HBM pricing is holding into a rate-tightening environment. Watch the gross margin line, not just the revenue beat.

The power names are in the most mechanical bind in the book right now. A higher discount rate compresses the present value of long-dated contracted cash flows, and higher financing costs make new capacity additions less attractive at the margin. Neither VST nor CEG is structurally broken by a 25-basis-point hike, but the dot plot's implication, a terminal rate potentially 75 basis points above where we sat a month ago, is a meaningful headwind for names that trade partly on the duration of their PPA cash flows. Both are in repair mode. The AI power demand thesis remains intact in the physical sense, 24/7 firm power is still the datacenter bottleneck that panels can't solve, but the rate backdrop has put the pillar on its back foot and that's where it sits until the dots move or the 10-year finds a ceiling.

The broad market's weekly scorecard looks more constructive than the week felt: the S&P 500 finished up 0.7% on the week and the Nasdaq gained 2.6%, with the Dow the outlier, shedding 733 points as rate-sensitive and old-economy names led the selling. The split matters. Tech broadly recovered from Monday's shock; the Nasdaq's week-over-week gain came entirely from the back half of the week after the hike was absorbed and investors concluded the AI buildout isn't a Fed casualty. AMZN reports October 29. MSFT and NOW don't report until late October either. The platforms pillar is effectively in a six-week earnings vacuum, which makes the rate path and the 10-year the daily score until then.

One note on the Nasdaq that's worth flagging: Friday's close triggered the index's quarterly rebalance, with SpaceX's weighting rising to roughly 2.82% from 1.28%, generating an estimated $15.5-22 billion in programmatic buying into the close. That flow is not a signal; it's index mechanics. Don't mistake the late-day Nasdaq strength for genuine breadth.

The financials sector lagged the broader market this week, and the reason is the nuance we've been flagging for several weeks. A Fed-driven rate hike flattens the curve rather than steepening it, which is the opposite of what bank lending spreads want. BAC took an additional hit after CEO Brian Moynihan told investors Wednesday that quarterly sales and trading revenue would come in "flat." That's not a disaster, but it's not the number that justifies a re-rating either. The brokerage side of this pillar, where cash sweep revenue accrues to the rate level rather than the curve shape, is the cleaner expression of higher-for-longer. The bank side is working against its own macro tailwind right now.

Bitcoin closed Friday at $81,057, above the $78-80K shelf that failed twice in the sessions preceding the FOMC. Gold finished at $4,419.90. Both are above the levels that had us concerned heading into Wednesday, which is a reasonable outcome given what the dot plot delivered. But the structural argument is unchanged: positive real short rates are the single condition under which the hard-asset sleeve faces real headwinds. The repression frame, in which bills are a parking spot paying below inflation, is the world where these positions earn their keep. If December delivers another 25 basis points and the terminal rate reaches 4.5%, that frame gets harder to maintain with a straight face. The honest position is that this sleeve is earning its insurance premium even while the short-term rate environment argues against it.

Oil ended the week around $95 for WTI, off the mid-week spike but still elevated versus the August baseline. Geopolitical supply risk from the Middle East remains the operative variable, and Saudi Arabia's efforts to resume pipeline flows provided some short-term relief. COP's thesis, Permian volumes plus durable free cash flow plus a nominal escalator against inflation, doesn't change at $95 oil. It gets better. The energy pillar is one of the few places in the book that benefits from exactly the environment the Fed is fighting.

What's on Deck: The Week of September 22

The most consequential item on this list is the August PCE print on Friday, not any earnings report. August CPI ran 3.4% and core came in a tick above consensus. If PCE confirms that core services inflation isn't decelerating, the December dot becomes a near-certainty in the futures market, and the 10-year will test 5% again before October is out. The September CPI release on October 14 lands two weeks before the next FOMC meeting (October 27-28), which carries no updated projections. The December meeting does. That sequencing matters: the market will need to price December's potential hike with only two more CPI prints and one PCE in hand. The October 14 number is therefore the next true regime-change event.

MU on September 30 is the near-term AI-chip data point we don't have to wait until November for. NVDA's next report isn't until November 25. MU's HBM demand signal and margin commentary will be the proxy read on whether the compute buildout is absorbing higher financing costs without blinking.

Forward Thesis by Pillar

AI Chips / Hardware. We own NVDA, AMD, MU, and TSM. Conviction remains high on the contracted compute thesis. Monday's selloff was a rate-shock and sentiment event, not a demand revision. The hyperscaler capex cycle doesn't have an off switch at 4% fed funds; it has contracts. MU on September 30 is the next data gate. What would change conviction: a genuine demand revision from one of the top-five hyperscalers, a capex cut rather than a capex delay. We haven't seen that signal. A safety speech from an AI CEO is not that signal.

Power and Grid. VST and CEG are part of our power book. The thesis on firm 24/7 generation for datacenter PPAs is physically intact. The rate environment is not the thesis; it's a valuation headwind on the duration of contracted cash flows. Higher financing costs at the margin slow new capacity economics. That's a real force. Both names are repairing rather than leading, and that's the correct characterization of where they sit. What would change conviction in the other direction: a return to a flatter rate path after December, or a major new PPA announcement that validates the demand premium.

Platforms and AI Software. MSFT, AMZN, and NOW are held. The platforms thesis, hyperscaler capex owners on the other side of the chip trade, doesn't require a benign rate environment to work. It requires AWS reacceleration and cloud consumption to hold up against higher corporate borrowing costs. The AMZN print on October 29 is where that thesis gets graded. NOW is on the shortest leash in the book; enterprise software spending is the first thing CFOs revisit when the cost of capital rises. Six weeks until we know.

Financials / Rate-Regime. BAC and SCHW remain positions. The pillar is working, but the internal split is widening. A flattening curve on the back of a Fed-driven hike squeezes lending spreads and makes the bank side of the trade less clean. The brokerage and cash-sweep side is the truer expression of higher-for-longer, and that's where the rate thesis compounds. BAC's flat trading revenue quarter doesn't break the thesis. It complicates the timing. Watch the October earnings call for net interest margin guidance before resizing any conviction here.

Hard Assets. GLD and IBIT are held. The honest framing: this sleeve is currently on the wrong side of positive real rates. Bitcoin recovered to $81,057 by Friday's close, above the shelf that failed twice, and gold is holding. But the structural argument for this sleeve depends on the repression frame reasserting itself, which requires either inflation re-accelerating past the Fed's ability to respond, or fiscal pressure forcing a change in the rate path. Neither has arrived. The sleeve earns its insurance premium precisely in conditions where it's uncomfortable to hold. That's where we are.

Energy. COP is a position. WTI near $95 and persistent Middle East risk are the environment this thesis was built for. Real assets with nominal escalators, record Permian volumes, and durable free cash flow at current oil prices. The pillar is one of the few outright beneficiaries of the inflationary regime the Fed is now actively fighting. That tension resolves one of two ways: either the Fed wins and oil comes in, taking COP's tailwind with it, or the Fed doesn't fully win and oil stays elevated. We know which scenario the book prefers.

VRTX and TSLA. Vertex remains the deliberate off-AI diversifier, generating cystic-fibrosis cash flows and funding a genuine pipeline in non-opioid pain and type-1 diabetes. Uncorrelated with everything above. TSLA reports October 28; autonomy and physical AI remain the long-horizon thesis, and the quarter's EPS print is less interesting than any commentary on the robotaxi rollout and the energy storage business at current power prices.

The week ahead is a PCE print and a rates repricing exercise. The six weeks after that are an earnings gauntlet. Warsh delivered what he said he would. Now we find out if 4% was the destination or just a waypoint.

This document is for informational purposes only and does not constitute investment advice or an offer to buy or sell any security. Nova One Capital, LLC and the Nova X Convergence Fund hold positions in securities discussed in this letter, including but not limited to NVDA, AMD, MU, TSM, VST, CEG, MSFT, AMZN, NOW, BAC, SCHW, GLD, IBIT, COP, VRTX, and TSLA. Past performance is not indicative of future results. All investments involve risk, including the possible loss of principal. This letter reflects the views of the author as of September 19, 2026, and those views are subject to change without notice.