The Loop the Fed Cannot Cut.

Executive Summary

The FOMC voted 12-0 on 16 September to raise the target range 25 basis points to 3.75 to 4.00 per cent, its first hike since July 2023, and the new Summary of Economic Projections shows 16 of the 18 participants who submitted a dot pencilling in at least one further move by year-end, with a median at one and four dots at two. Chair Warsh described the vote as a response to a strong labour market, a resilient consumer, and inflation that "remains elevated"; the statement said the action "will support a timelier return" to the 2 per cent goal. Market reaction was orderly on the surface and revealing beneath it: two-year yields cleared 4.73 per cent, more than 10 basis points above the pre-meeting level; the 10-year eased to 4.97 per cent as the long end absorbed the credibility of the tightening rather than extending it; the S&P 500 closed off 0.4 per cent; gold and silver gave back their morning gains within thirty minutes of the presser opening.

Our controlling thesis for the month: the front end has now caught our July and August calls, but the market is still mispricing what the hike does. Every 25 basis points lands on a funding chain in which the hyperscalers, the Treasury, the consumer, and the equity market are the same trade. We keep no US duration; we add to physical gold on this correction as the largest allocation; we retain the Brazil and Switzerland overweights; we hold the volatility book.

Since last month

The 15 September note made four checkable calls into the FOMC and the scorecard is unusually clean. The hike call landed to the day, at the size and range we forecast (3.75 to 4.00), on a unanimous 12-0 vote. The no-US-duration call landed: the two-year at 4.10 to 4.20 per cent on 14 September cleared 4.73 per cent on the print, and the 10-year, having briefly worn a 5 handle intraday on 14 September, settled 4.97 per cent on 16 September; a duration book taken at face value through the meeting was marked down in both legs. The energy pass-through call landed: retail diesel briefly touched $6.00 per gallon on 12 September per AAA before easing, and the ULSD-WTI crack that we made the analytical centrepiece last month at $108.02 per barrel intraday on 3 September remained the reason a Fed serious about 2 per cent could not make its target with the tool it just used. The gold call is the one to explain honestly: spot has continued from $4,607 on 21 August through $4,322 on 14 September to approximately $4,158 on 16 September post-decision, per Kitco streaming data, a 9.7 per cent drawdown from the August peak. The direction is what we said would happen on a hawkish repricing; the depth is deeper than the $4,250 to $4,400 add zone we stated, and we take that seriously. The volatility call remains unpaid, with the VIX at 15.6 on 16 September. We do not raise it.

Sources: Federal Reserve Board (FOMC statement and SEP, 16 September 2026); Bloomberg reporting of 16 September market moves; S&P Dow Jones Indices; Kitco and LBMA reference data for gold; Cboe Global Markets for the VIX. Data as of dates shown.

1. The hike arrived, unanimous

The Fed raised rates by a quarter point on 16 September, to a range of 3.75 to 4.00 per cent, its first hike since 2023, and the unanimous vote tells you the committee is more worried about inflation than about what dearer money will do to the machinery underneath this expansion. The dot plot placed 16 of the 18 participants who submitted a projection at one or more further hikes by year-end, with a median of one and an average landing near 4.125 per cent by December; only two see no further move in 2026. Chair Warsh, per his stated practice, declined to submit a dot. The statement retained the language that "economic activity is expanding at a solid pace" and added that "productivity growth is strong, and capital investment is robust", both of which our Section 3 will examine as claims. Chair Warsh's press conference offered no forward guidance, criticised the practice of it, and returned the balance sheet to the field of view as a preferred instrument for restraining inflationary pressures.

The market read the release in two motions. The two-year touched 4.73 per cent, its highest since 2024 and more than 10 basis points above the pre-meeting level, per Bloomberg. The 10-year, which briefly cleared 5.00 per cent intraday on 14 September, closed 16 September near 4.97 per cent as the long end absorbed the credibility of the tightening rather than extending it. The S&P 500 closed down 0.4 per cent on the day at approximately 7,588; the Nasdaq was near flat. Gold and silver, which had rallied into the print, gave back the morning gains during the presser. This is what a hawkish delivery looks like when the front end has done most of the work in advance.

2. Every 25 basis points lands on the funding chain

What worries us this month is where those 25 basis points actually settle. That machinery is the AI buildout, and the AI buildout runs on debt. The frontier model companies are unprofitable, the gap between what subscribers pay and what tokens cost to serve is being covered by hyperscaler cheques and venture capital, and the hyperscalers themselves have turned to the bond market to fund their capex. Alphabet reported negative free cash flow in the second quarter of 2026, the first negative quarter in its history as a public company, at approximately $5.9 billion, with quarterly capex at $44.9 billion and full-year 2026 capex guided to $195 to $205 billion; concurrent bond and preferred issuance of roughly $20 billion doubled long-term debt to approximately $98 billion, per Alphabet's Q2 2026 disclosures. Amazon's trailing-twelve-month free cash flow turned negative at approximately $7.6 billion by the end of Q2 2026, per Amazon's filings, on trailing cash capex of roughly $173 billion against operating cash flow of $161 billion. Two of the most cash-generative businesses ever built are now net cash absorbers, and the shortfall is being met in the bond market. Every 25 basis points lands directly on that funding chain.

The same logic applies to Washington, which has to refinance a large stock of maturing debt into 2027 while the deficit is still widening. Coupon maturities alone in calendar 2027 run near $4 trillion per public reporting of the maturity wall, and combined gross issuance (coupon and bill rollovers plus net new borrowing into a still-widening deficit) will run materially higher, in our estimation above $7 trillion. A higher policy rate means more interest, more borrowing and more supply pushed into a market that is already charging a premium to absorb it. On 9 September the 10-year auction cleared at 4.834 per cent with indirect bids at 79.2 per cent, close to a record indirect share; on 16 September, after the hike, the 10-year sat at 4.97 per cent. That is a market that will still be bid at the right yield; it is also a market unwilling to rally on the news that the Fed intends to bring inflation to heel.

3. The circle: missing input, mispriced risk

Here is the loop the Fed did not break on 16 September. Roughly 58 per cent of the jobs created in the United States in 2026 year to date, in our estimation from the BLS Employment Situation series, came from three sectors: government (predominantly state and local, offsetting a federal drawdown of approximately 348,000 since October 2024), healthcare, and social assistance. These are the sectors funded by the state or by state-supported financing, not by output per worker. Chair Warsh's statement on 16 September cited productivity growth as strong; the composition of the 2026 jobs print is not the composition of a productivity boom. It is, empirically, the composition of a transfer-financed expansion.

Consumption is holding up, but it is holding up because the top decile of the household distribution is carrying an outsized share of the marginal spending, per the BEA income-and-outlays series and Federal Reserve Bank of San Francisco decomposition work published in 2025 and 2026. That top-decile spending is driven by portfolios; those portfolios are driven by the seven names whose capex plans, financed as described in Section 2, are the entirety of the AI trade. The circle closes on itself. The Fed is tightening into an economy whose reported growth, reported employment, and reported consumption all lean on the same leveraged trade. Consensus reads the four series (payrolls, retail sales, S&P 500 level, capex intentions) as independent evidence of resilience; they are not independent, they are the same equity, expressed through four labels. Every 25 basis points is heavier than it looks because the hike does not encounter four resilient economies, it encounters one leveraged trade, four times.

4. The buyer base is thinner

Abroad, the buyers who used to absorb this debt are stepping back. The joint US-Japan intervention to support the yen on 1 August, per the Ministry of Finance in Tokyo and confirmed by the US Treasury on 3 August, was the first coordinated yen-buying operation between the two countries since 1998; it stabilised the currency for a few weeks and Japan continued to trim its US Treasury holdings on the margin in the months that followed, per subsequent TIC releases. The World Gold Council's July central-bank statistics record official-sector accumulation running at roughly 1,000 tonnes per year over the past four years against a 500-tonne average over the preceding decade; the 2026 Central Bank Gold Reserves Survey reports a record 45 per cent of central banks planning to add to gold reserves over the next 12 months and 74 per cent expecting a moderate or significant decline in dollar reserves within five years. Secretary Bessent had one job on taking office, which was to bring the deficit down; instead the deficit is larger, the policy rate is higher, and the buyer base is thinner. None of this is a timing tool. Valuations, dollar hegemony, and the size of the deficit can stay stretched for years. It is, however, a strong reason to hold assets that do not depend on the circle staying closed.

5. Portfolio positioning

The direction is unchanged from July, August and last month. We keep adding on the correction.

Gold remains the largest single allocation. Spot has drawn from $4,607 on 21 August through $4,322 on 14 September to approximately $4,158 on 16 September post-decision, per Kitco and LBMA reference data, a 9.7 per cent drawdown from the August peak. Our stated $4,250 to $4,400 add zone was exceeded to the downside; we add through the range and below it, as the reasons that drove the July and August adds have concentrated rather than weakened. A Fed committed to price stability, in a statement acknowledging that inflation "remains elevated", that cannot lift real yields far enough to keep gold under $4,200, is a Fed whose credibility on the target rests on delivering more than one further hike; the metal reprices to the credibility, not to the level.

Equities. The US market remains crowded. The S&P 500 closed approximately 7,588 on 16 September, down 0.4 per cent on the day and up 11.1 per cent year to date on a price basis, with concentration in the AI-related names unchanged. We stay underweight. The Swiss Market Index closed 13,775 on 11 September, up 15.3 per cent year to date; the overweight is retained. The Ibovespa closed 187,207 on 12 September, up 16.2 per cent year to date, with the Selic at 14.00 per cent and Copom easing while the developed world hikes; the overweight is retained and remains our preferred developed-vehicle for real-currency duration. China remains structurally underweight.

Fixed income. We continue to hold zero US duration. The buyback backstop did not hold, the 10-year briefly wore a 5 handle on 14 September and closed 4.97 per cent on 16 September. Every 10 basis points above the CBO's 4.3 per cent trend assumption, applied to the roughly $32.4 trillion of debt held by the public, adds approximately $32 billion a year to interest cost. The two-year at 4.73 per cent post-decision is priced approximately to the median dot; if the September or October CPI prints surprise higher, or if the four-dot camp expressing two more hikes gains adherents at the December SEP, the two-year has further to rise.

Volatility. The VIX at 15.6 on 16 September, on the day the Fed delivered its first hike since 2023 accompanied by a hawkish dot plot into a market at 11.1 per cent year to date, is not priced for what our thesis implies. The call has not paid for three cycles; we retain it and do not raise it. Selling strength and carrying index hedges remain the operational expressions.

Risk management. If October and November CPI prints firm the median dot toward the four-dot two-hike camp, we see a US equity market at 11.1 per cent year to date and a VIX near 15 exposed to a correction of at least 10 per cent, in our view. If instead the diesel crack collapses on a rapid sanctions resolution and core continues to decelerate, gold, the Swiss Market Index, the Ibovespa and long duration all rally, and in that world we add duration through the Brazil real book, not through US Treasuries. We size the equity hedge to the first scenario, hold cash to add on the second, and we do not remove the gold position in either.

6. Conclusion

The August note said the hike was the correct next move; the 15 September note said the market was pricing the wrong cure; on 16 September the Fed delivered the cure the market wanted, and by the close of the presser the market disliked the delivery. Our controlling thesis for the month is that neither the front end's re-pricing nor the equity market's shrug has yet reckoned with the loop underneath. For our loop thesis to be wrong from here, three things must resolve together. Hyperscaler cash flow must turn positive on unchanged capex plans, which requires monetisation of the frontier models at rates their current pricing does not deliver. The composition of US labour-market growth must swing back to the goods-producing and business-services sectors that carry productivity, and the top-decile share of marginal consumption must revert toward the historical distribution. And the foreign official sector must resume net accumulation of Treasuries at a pace that closes the gap now being filled by private yield-seekers, cooling the term premium the market imposed even as the Fed hiked. Each is possible; in our view their joint probability is materially below what a market recovering from a 0.4 per cent day is treating as consensus. We are positioned for the alternative: gold added into the correction as the largest allocation, non-US equity retained with Brazil and Switzerland preferred, no US duration, and the volatility hedge held through the October CPI print and the November FOMC.

Important Information

This document is a research note prepared by the Neovision Group for institutional and professional clients. It is not investment advice, an offer, or a solicitation to buy or sell any security or financial instrument. All figures cited are sourced from public information.

The Neovision Group comprises Neovision Wealth Management Limited (regulated by the Financial Services Regulatory Authority in the Abu Dhabi Global Market), Neovision Investment Fund Management (NIFM, regulated by the Capital Markets Authority in the onshore UAE), and Héritage Riviera SA (regulated by FINMA in Switzerland).

Sources referenced in this month's note include: Federal Reserve Board (FOMC statement, Implementation Note and Summary of Economic Projections released 16 September 2026; Chair Warsh's 16 September press conference; Chair Warsh's keynote at the 2026 Jackson Hole Economic Policy Symposium of 28 August 2026); US Bureau of Labor Statistics (Employment Situation and industry payrolls series; CPI and PPI releases for August 2026); US Bureau of Economic Analysis (Personal Income and Outlays); Federal Reserve Bank of San Francisco (consumption decomposition research); US Department of the Treasury (Daily Treasury Par Yield Curve; 10-year auction results 9 September 2026; TIC data through June 2026); Congressional Budget Office (February 2026 Budget and Economic Outlook); US Energy Information Administration and AAA (retail fuel data); DTN Progressive Farmer reporting of CME ULSD-WTI settlements; World Gold Council (July 2026 central-bank statistics and 2026 Central Bank Gold Reserves Survey); Kitco and LBMA reference data for gold; S&P Dow Jones Indices for the S&P 500; SIX Swiss Exchange for the Swiss Market Index; B3 for the Ibovespa; Banco Central do Brasil for the Selic; Alphabet Inc. and Amazon.com Inc. Q2 2026 quarterly disclosures; Bloomberg, CNBC and Reuters reporting of 16 September FOMC market reaction; Ministry of Finance of Japan and US Treasury for the 1 August 2026 joint yen intervention; Neovision Group internal analysis for the funding-chain composition, the labour-market-share estimate, and the sensitivity of federal interest cost to the trend-to-market yield gap.

Market data as of close, 16 September 2026, unless otherwise stated. Forecasts and views expressed represent the opinion of the Neovision Group at the date of publication and are not guarantees of future outcomes. Past performance is not indicative of future results.