Markets give you a lot to react to and not much context. This monthly note is our attempt to provide some of that context. Thank you for reading.
If you’re a client, we sincerely appreciate your business. If not, and you would like to learn more about our firm, please feel free to reach out. You’re also welcome to share this update with others who may find it useful.
July 2026 in Review
Call it the summer doldrums. Stocks have had a hard time pushing toward new highs, last seen on June 2. Interest rates aren’t helping, which also meant that bonds struggled. The 30-year U.S. Treasury rate is up 0.37% year-to-date to 5.2%, a 19-year high.
The Dow gained (+0.4%) and the S&P 500 was flat, while the Nasdaq 100 (-6.6%) and small caps lost (-1.9%). The spread came from leadership, not the market. The S&P 500 Equal Weight Index gained about 1.3% and closed at a record. The first half’s 20 best-performing S&P 500 stocks all finished July lower, averaging a 25.3% decline, while 17 of the 20 worst finished higher, averaging a 12.2% gain. Year-to-date, small caps lead (+21.6%) against the S&P 500 (+10.1%).
Global stocks (-0.3%), developed markets (-0.9%), and emerging markets all lost ground (-1.6%). The dollar fell (-0.8%) for the month but is up (+4.2%) for the year. Year-to-date, developed markets are up (+13.8%) and emerging markets (+9.4%), both of which have been running ahead of the S&P 500 all year.
Every category of fixed income was negative. Investment-grade corporates (-2.2%), municipals (-1.6%), aggregate bonds (-1.3%), international bonds (-1.1%), and high yield (-0.2%) all declined. The 10-year Treasury ended the month near 4.75%, its highest since January 2025. High yield’s smaller loss is a duration story rather than a credit one. It carries less interest-rate sensitivity than the investment-grade index. Cash returned (+0.3%).
Commodities had a big jump (+12.0%), the best month on the board. Crude traded above $100 during the escalation in the Strait of Hormuz before easing late in the month. REITs gained (+2.6%) as well as gold (+0.9%).
Bitcoin had the second best month of the asset classes tracked (+7.1%).
Market Narratives
A $45 Billion Fund Became a $10 Billion Fund in Three Weeks
Leopold Aschenbrenner is a 25-year-old former OpenAI researcher who wrote a 165-page essay in 2024 called “Situational Awareness: The Decade Ahead,” arguing that increasingly capable artificial intelligence would drive enormous demand for chips, memory, data centers, and power. He then launched a hedge fund of the same name to trade the thesis, starting with roughly $225 million from a group of prominent technology investors. By the beginning of July, it held about $45 billion and was reportedly up 439% in the first half of the year.
Aschenbrenner was long AI infrastructure (SK Hynix, CoreWeave, Micron, Nebius) and short software, financed with reported leverage of up to 400%. When chip stocks fell in mid-July, the equity cushion shrank, the prime brokers asked for more collateral, and raising that collateral meant selling into the same declining names. By the end of the month, the entire public portfolio had been firesold to Citadel at what he could get. The firm retained some private investments, including a stake in Anthropic reported at roughly $5 billion.
There’s some interesting background and fodder on Aschenbrenner. He started his career at Sam Bankman-Fried’s FTX Future Fund working on “effective altruism.” OpenAI fired him for leaking sensitive information. He was marrying the chief of staff to Anthropic CEO on the week that the fund fell apart. Per the WSJ, “There would also be a pre-wedding colloquium to discuss ideas in panels and breakout sessions. The couple’s only request: no gifts.”
A few things are worth pulling out of this. Aschenbrenner’s thesis may well prove correct over 10 years. But he might not get to fully participate because borrowed money converted a bad three weeks into a forced sale at prices set by a buyer who knew he had to sell. Liquidity, leverage, and concentration kill portfolios and quickly. I like the old quote from Warren Buffett, “Only when the tide goes out do you discover who’s been swimming naked.”
The forced selling also moved the market on its own. A widely followed index of AI momentum stocks fell 37% from its June peak, then rose 14% in a single session on the news of the Citadel sale.
Bespoke drew the obvious historical parallel. If you overlay the Nasdaq since the launch of ChatGPT onto its path after the launch of Netscape in 1994, we would be standing in early August 1998, weeks before Long-Term Capital Management came apart. LTCM was much larger and genuinely threatened the financial system, which this was not. But it is a reminder of how much of a market’s short-term movement can come down to who owns what, and whether they are being forced to sell it.
A New Chinese Model
On July 16, a Beijing-based startup called Moonshot AI released Kimi K3, a 2.8-trillion-parameter model, making it the largest set of openly published AI model weights in the world. Moonshot claimed it performed competitively with the best proprietary U.S. models at a fraction of the price, and on July 27 it published the full weights under a permissive license, meaning anyone can download and run it. Semiconductor stocks sold off hard on the news. Some analysts and pundits are calling it the second DeepSeek shock, after the January 2025 episode that briefly erased several hundred billion dollars of chip market value.
The competitive pressure is real. If capability close to the frontier is available as a free download, the companies charging for access to frontier models have a harder time defending their pricing. Artificial Analysis ranked K3 fourth overall on its Intelligence Index, third on the GDPval-AA v2 agentic benchmark, and second on AA-Briefcase, in each case behind the frontier models from Anthropic and OpenAI but by narrow margins.
The July earnings reports sorted out perceived winners and losers. And the market was more discriminating than it usually gets credit for. Alphabet raised its capex guidance and fell 7% the next day, dragging the group down with it. Meta guided revenue light and sold off through a nine-day losing streak. Microsoft disclosed that Azure had crossed $100 billion in annual revenue and rose 16%. Amazon raised 2026 capex to $220 billion, reported cloud growth strong enough that nobody minded, and rose 15%. Spending was rewarded where it came attached to visible revenue and punished where it did not.
Viktor Shvets at Macquarie described the environment as a series of rolling bubbles, with capital inflating and deflating across AI-adjacent themes in sequence rather than one bubble that pops once. Two things are worth holding alongside it. This is the second time in 18 months that a Chinese lab has moved markets by delivering capability at a fraction of the cost, and both shocks were about price rather than existence. And semiconductors remain cyclical. Margins are extraordinary when demand outruns capacity and slim quickly when it does not. Chip stocks have roughly doubled this year even after July’s selloff. The buildout can be real, and the stocks can still be priced for a cycle that does not turn.
The Fed Held and Three Officials Voted to Hike
The Federal Reserve’s Federal Open Market Committee left the federal funds rate at 3.50%–3.75% at their latest meeting on July 29, while the market had priced a 30-40% chance of a 0.25% increase.
However, the vote was 9–3, with all three dissents preferring a 0.25% increase. That is somewhat rare. It’s only the sixth time in 30 years that three members have dissented, although dissent typically clusters in periods when inflation is running above target.
The Fed statement was unchanged from June apart from a verb tense. There were no new projections, and Chair Warsh declined to explain the dissents or say much about the path ahead.
The market reaction was the interesting part. No rate change, no new statement language, no projections. Yet stocks still fell hard that afternoon. The Dow dropped 1,153 points for its worst day since April 2025, and the Nasdaq 100 finished more than 10% below its peak. The yield curve produced its sharpest steepening in a year. Two-year yields fell while the 30-year rose to its highest level since 2007, which is the market saying it thinks policy is too easy now and will have to be tighter later. Markets have a way of testing a new Fed chair, and the S&P 500 has now fallen more than 1% at each of Warsh’s first two meetings.
The inflation backdrop is genuinely mixed. The June CPI came in at 3.5% year over year, down from 4.2% in May and the first decline in five months, with core at 2.6%. Almost all of that improvement came from energy prices falling after the June ceasefire, and energy is still up 15.7% over 12 months. Chair Warsh was direct about not treating one cool print as a victory. Markets now put the odds of a September hike at roughly two-thirds.
Practically, this is a year that opened with most forecasters expecting cuts and now has hikes priced in. It is why bonds are slightly negative year-to-date and why cash is still earning close to 4%. The damage is concentrated in longer bonds.
From here, it’s a toss-up. You have a new Fed Chair that desperately wants to be dovish. It’s difficult to assign inflationary pressures from Iran energy shocks as worthy of a hike. In my opinion, the base case is zero to one rate hike through year-end.
Closing Thoughts
We’ve talked about a rotation of leadership in a couple of monthly reports this year, and that continued this month. We would argue that it’s healthy and positive for long-term growth.
On the positive side of the ledger for investors is improving stock breadth and momentum in corporate earnings. On the negative side are ongoing uncertainties about the AI business model, the Middle East war, the persistence of inflation, and the Fed’s reaction. The net result last month was a sideways market with some volatility.
The sentiment data is worth reviewing. The CNN Fear and Greed Index finished July in fear territory, and the AAII survey recorded one of the largest single-week drops in bullishness on record, leaving bulls at just 31%, with the S&P 500 only 2% below an all-time high. Those are typically bullish contrarian signals.
Ask a longer-horizon question and the answer inverts: 52.4% of Consumer Confidence respondents in July expected stock prices to be higher a year from now, against a long-run average of 35.6%. The Investors Intelligence survey also sits above its historical average. People are nervous about the next few weeks and optimistic about the next year. None of that is a forecast; it’s a reminder that how a market feels and what a market is doing are two different data sets.
Also, one counterweight to the 1998 and 2000 comparisons that appear in this report as we talk about AI and hyperscalers. The S&P technology sector’s forward price-to-earnings ratio is currently 20.0, against roughly 19.4 for the S&P 500 as a whole. At the 2000 peak, that gap was 30 points. Whatever this market is, it’s not that one.
As always, please reach out if you’d like to talk through how any of this affects your specific situation.
Thank you for reading.
