July’s semiconductor selloff wiped hundreds of billions off chip stocks in a matter of days. But one class of Wall Street institution earned record fees whether AI infrastructure spending kept climbing or investors rushed for the exits.
July 2026 delivered one of the most volatile months global markets have seen since the pandemic. Chip stocks fell nearly 30% from their June peak, South Korea’s exchange triggered back-to-back circuit breakers for the first time in its history, and the US Federal Reserve delivered a rate decision that satisfied almost no one. Days later, the same tech names that had cratered clawed back hundreds of billions of dollars in market capitalisation.
None of this settled the underlying question of whether artificial intelligence delivers on its promised returns. What it did settle is a narrower, and arguably more useful, question for anyone trying to gain exposure to the AI capital cycle without betting on a single chipmaker: which institutions get paid regardless of which way the story breaks.
The Structural Case for the Middlemen
Investment banks such as Goldman Sachs don’t manufacture GPUs, build hyperscale data centres, or train frontier models. Instead, they arrange the capital that flows into and out of those projects, taking a fee each time money moves in either direction.
The logic resembles an auction house: a percentage is taken when a collection sells at a record price, and a percentage is taken again when the same assets are restructured or liquidated during a downturn. During a boom, banks earn advisory and underwriting fees as hyperscalers raise debt and equity to fund data centre buildouts, regardless of which chip supplier or cloud platform ultimately wins market share. During a drawdown, the same banks earn fees again, this time from restructuring, mergers, acquisitions and private debt financing as distressed or discounted assets change hands.
Goldman Sachs’s own second-quarter results illustrate the boom side of that equation. Reporting on 14 July, roughly two weeks before the worst of the July selloff, the bank posted diluted earnings per common share of $20.98 and an annualised return on average common shareholders’ equity of 23.5% for the quarter. That EPS figure was nearly double the $10.91 reported a year earlier and beat the Zacks consensus estimate of $14.47 by a wide margin.
The revenue drivers were concentrated in exactly the businesses that benefit from a capital-raising boom. Investment banking fees climbed 55% to $3.40 billion, supported by growth in advisory work, equity underwriting and record debt underwriting revenues, while record equities revenue of $7.42 billion jumped 72% year over year. On the deal-advisory side, the bank’s investment banking backlog rose to its highest level in five years, driven by robust strategic M&A activity, and management pointed directly to the AI buildout as the source: CEO David Solomon said the AI infrastructure build-out is in its early stages and is expected to drive strategic activity and financing opportunities going forward.
What Actually Happened in July
The month that tested this thesis began in Korea. On 13 July, SK Hynix, Nvidia’s key supplier of the high-bandwidth memory used in AI accelerators — tumbled by a record 15% in a single session, part of a broader Kospi selloff that triggered a market-wide circuit breaker after the index plunged more than 8% amid geopolitical tension in the Middle East.
Things escalated further at the end of the month. On 29 July, SK Hynix reported its most profitable quarter on record and still missed what investors were pricing in, contributing to a second consecutive circuit-breaker halt on the Kospi, the first time that had happened in the exchange’s history. Combined with China’s ChangXin Memory completing an $8.6 billion IPO and reports of new domestic lithography capacity, the pressure spread across the broader semiconductor complex: the Philadelphia Semiconductor Index closed the month roughly 28.6% below its late-June peak, based on figures reported by BeInCrypto.
Nvidia itself was not spared. The stock fell nearly 5% on 27 July, briefly ceding its position as the world’s most valuable company. More striking than the equity move was the reaction in credit markets, where the cost of insuring Nvidia’s debt against default spiked sharply, a signal that credit desks, which had treated the chipmaker as a low-risk name, abruptly repriced that assumption.
The Fed Added to the Pressure, Not the Relief
Markets went into the Federal Reserve’s late-July meeting already on edge, and the decision did little to calm things. The Federal Open Market Committee voted 9-3 to hold the federal funds rate at 3.5%–3.75%, with all three dissents coming from regional bank presidents, Beth Hammack of Cleveland, Neel Kashkari of Minneapolis and Lorie Logan of Dallas, who wanted to raise rates rather than hold. It marked the largest bloc of same-direction dissents on the committee since 2016.
Markets did not take the hold as dovish reassurance. The Dow fell 1,153 points that day, its worst single session since April 2025, while the S&P 500 dropped about 1.5% and the Nasdaq fell roughly 1.7%. Fed Chair Kevin Warsh, only weeks into the role, offered little forward guidance on the central bank’s next move, leaving investors to price in real uncertainty ahead of the Jackson Hole symposium and the next scheduled decision in September.
Weighing the Fee-Engine Thesis Against Its Weak Points
The case for institutions like Goldman is a relative one, not a defensive one, and it comes with real caveats.
It is not insulated from the selloff. Bank stocks fell alongside the broader tech and AI complex in late July. If AI-linked equities keep falling, bank share prices are likely to fall with them — the thesis concerns the durability of the underlying fee income through the cycle, not share-price protection from volatility.
It is not a neutral bystander. As a major prime broker and lender to leveraged market participants, an investment bank’s balance sheet is directly exposed to the same leverage that amplifies both the boom and the bust. Extending credit into a leveraged market is a materially different risk than simply collecting a toll on transactions.
It is not cheap. Bank valuations were trading near record levels relative to tangible book value heading into August, and some analysts argue this reflects a cyclical peak in fee income rather than a sustainable new baseline.
Restructuring fees have historically not offset a genuine credit event. In both 2008 and 2022, underwriting and advisory revenue fell together rather than one offsetting the other. The two-sided fee model cushions a slowdown; it does not neutralise a full-blown credit crisis.
Even Goldman’s own leadership has struck a more cautious tone than the headline numbers might suggest. Solomon has repeatedly noted that enterprise adoption of AI is proving slower and more complex than markets initially expected, even as the infrastructure investment cycle itself remains in its early stages.
The Bigger Picture for August
None of this amounts to a claim that bank stocks preserve capital during a broad market drawdown — high-beta financial stocks offer no such shelter. The more precise argument is about diversification of risk within a multi-trillion-dollar capital cycle: for investors seeking exposure to AI-driven capital formation without concentrating that bet in a single chipmaker exposed to sudden shifts in memory pricing or export policy, a business that captures fees on the way up and the way down offers a materially different risk profile than owning the underlying hardware names outright.
Whether current bank valuations represent a sensible entry point, against a backdrop of a divided Fed and a September rate decision still very much in play, remains a separate and much harder question. With AMD’s earnings landing on 4 August and further binary events likely through the month, position sizing, not conviction , may be what matters most in August.
This article is for general informational and educational purposes only and does not constitute financial, investment, legal or tax advice. Mentions of specific companies or tickers do not constitute a recommendation to buy or sell any financial asset. Readers should conduct independent research or consult a licensed advisor before making financial decisions.
Sources
- Goldman Sachs, Second Quarter 2026 Earnings Results
- Zacks Investment Research, Goldman Q2 Earnings Beat on Solid Trading & IB Revenues, Shares Rise
- GuruFocus, The Goldman Sachs Group Inc (GS) Q2 2026 Earnings Call Highlights
- Advisor Perspectives, SK Hynix Shares Plunge Most on Record in Deepening Korea Selloff
- Bloomberg, Korean Stocks Plunge 16% in Two-Day Burst of Retail Selling
- TechTimes, KOSPI Triggers Historic Back-to-Back Circuit Breakers as AI Memory Rally Faces Structural Reckoning
- CNBC, Fed rate decision July 2026: Divided Fed holds interest rates steady
- CNN Business, Fed holds interest rates steady after cliffhanger meeting, but three officials dissent
- Bloomberg, Fed Holds Rates Steady as Three Officials Dissent in Favor of a Hike
