An earnings report concerns the "AI conviction," and a speech affects risk appetite! Nvidia and Walsh will decide the fate of the rotation bull market

As global stock markets embark on what is set to be the most decisive week of the summer, two things have become abundantly clear: the current market has moved beyond trading solely around the artificial intelligence computing power theme. Notably, significant sector rotations have emerged, including the rise of precious metals led by gold, the healthcare sector driven by mRNA technology ushering in a new era of cancer treatments, and the resurgence of traditional defensive value sectors like energy, telecommunications, and banking, whose valuations have lagged behind technology stocks for years. These shifts have broken the recent pattern of “AI computing power dancing alone” and ushered in a healthier, broader bull market. Meanwhile, the recent sharp volatility in long-dated U.S. Treasuries (10-year and beyond) has made it clear that catalysts can just as easily fuel bearish forces and trigger further sell-offs as they can drive assets up.

This is the backdrop as we approach this crucial moment at the end of August. On Wednesday, NVIDIA will report earnings—an event widely perceived as the AI computing theme’s closest equivalent to a “quarterly referendum.” Then, on Friday local time, Federal Reserve Chair Kevin Walsh will make his first keynote speech at Jackson Hole. At the same time, fierce debate is swirling in the market over the U.S. Treasury’s actions to arrest the sell-off of long-term U.S. bonds. The publicly announced theme of Walsh’s speech is payment innovation, which at first glance seems unlikely to significantly impact markets. However, he may touch on other topics in his remarks, and there is a risk that the market could over-interpret a single comment.

The Bank of America August Global Fund Manager Survey does not paint a picture of “investor sentiment turning cautious and pessimistic.” Instead, it reveals a contradictory state in which institutional risk perception has risen sharply, yet actual positioning remains highly aggressive, with AI positions crowded: a record 56% of respondents are betting that the U.S. economy will achieve a “no landing” outcome over the coming 12 months, a net 37% expect double-digit corporate earnings growth, and 72% believe the Fed will not hike rates before the midterm elections. Correspondingly, global equity net overweighting has risen to 56%, the highest since November 2021, cash allocations have dropped to the sixth lowest in history at 3.5%, and the net underweight in bonds has widened to 39%.

During the fund manager survey period, another Bank of America research report indicated that as the AI theme faces a deleveraging storm, a forced unwind of extremely crowded long positions, and inflation risks challenge traditional investment portfolios, value stocks, biotech, regional banks, selected credit assets, and commodities all offer attractive investment opportunities. International small-cap value stocks have become more appealing relative to U.S. large-cap growth stocks, while Japanese corporate profitability has reached a record high. The institution also sees listed private equity managers as a contrarian play and prefers high-quality, high-yield bonds over investment-grade bonds.

Wall Street financial institutions including Bank of America are not bearish on the AI theme, but emphasize a clear upgrade in asset allocation: from highly concentrated AI/U.S. large-cap growth trades, toward investment diversification with “retention of structural AI long exposure + an increase in low-correlation, high-cash-flow, low-valuation assets.” As AI moves from a scarce narrative into a trillion-dollar capex realization phase, the deciding factors in excess returns will shift from “do you have AI exposure” to “does valuation, free cash flow, ROIC, and crowdedness align.”

This week is by no means simply NVIDIA earnings week, but a dual stress test of “corporate earnings anchor + global discount rate anchor”: NVIDIA will answer whether token computing power demand can be converted into sustained inferencing revenue, profit, and AI capital returns, while Walsh may reset the valuation baseline for all risk assets by influencing policy rate expectations, the term premium, and dollar asset liquidity.

The correlation between artificial intelligence and non-artificial intelligence sectors has fallen to around -0.6, indicating that this is more a redistribution of funds rather than a wholesale exit; but in a context of insufficient hedging and where the first batch of systematic, fast-money quant selling is only about 4% below current market levels, a “strong NVIDIA report + stable long bonds” could turn sector rotation into a bull market with healthy breadth. Conversely, it could turn a local correction into systemic de-risking.

Token Throne Faces Quarterly Referendum: The AI Computing Theme Is Not Entering a Bear Trajectory, but Undergoing an AI Investment Return Rate Judgment

Before the crucial end of August, it is necessary to accurately assess the current position of the stock market. Although some tech stocks have rebounded, the rotation out of artificial intelligence sectors persists. Over the past five trading days, the best-performing assets have included “high-quality stocks” with traditionally abundant cash flows, baskets tracking so-called “AI victims,” and all value stock assets that are sensitive to inflation data or have relative advantages in stagflation environments. Semiconductor and AI data center infrastructure stocks, on the other hand, have fallen sharply.

As shown above, last week, inflation and interest rate concerns drove market rotations—assets with strong fundamentals, those benefiting from stagflation, and commodities outperformed, while the AI sector lagged.

One particular chart succinctly captures the broader market landscape. The 40-day correlation between Goldman Sachs’ latest broad-based AI stock basket and the S&P 500 index excluding AI-related components has turned sharply negative for the first time, currently around -0.6. This indicates that capital flowing out of the AI sector is not leaving the stock market entirely, but rather is providing large-scale net inflows into all other sectors.

As illustrated above, the correlation between artificial intelligence and non-artificial intelligence sectors has turned sharply negative—the retracement and rotation of popular trades has allowed other non-AI related investment sectors to catch up.

However, none of this means that the “macro AI narrative” has been broken. The technology companies central to the construction of AI infrastructure still account for about half of the S&P 500’s overall earnings growth. Meanwhile, median S&P 500 company profits rose 14% last quarter, meaning the earnings base for the stock market is broadening against the backdrop of a U.S. soft landing.

Some media have previously reported that Anthropic’s run-rate annual revenue has surpassed $65 billion, while OpenAI’s is around $40 billion. These figures offer some comfort to investors concerned about real terminal demand for AI applications. The trickier questions remain: who will ultimately come out on top, can capital expenditures continue to yield robust returns, and just how much is AI computing power actually worth.

A widely watched token cost metric—the Silicon Data AI large language model spend index—has dropped nearly half from its May peak. This has become a hot topic across trading desks and is a frequent subject of social media tips on saving model usage token inference costs. It should be noted that Ornn Market data does not indicate a significant decline in token prices for large AI labs in August. Whatever the case, the index’s popularity alone reflects current market anxiety. At the same time, CDS spreads of hyperscale cloud providers and credit metrics continue to rise, further intensifying concerns about the AI investment return rate and fears that massive capital expenditure could lead to credit defaults.

As shown above, numerous unresolved issues remain with the AI narrative—user volume, laws of scale, pricing, real profits, and sector leaders are still unclear.

AI infrastructure still contributes about half of the S&P 500’s earnings growth, with Anthropic and OpenAI hitting annualized run rates of over $65 billion and around $40 billion, respectively, proving terminal demand is not fictional; but the token cost index has nearly halved from May highs, and cloud giants’ credit spreads continue to widen, meaning the focus has shifted from “who owns the most GPUs” to “who can turn computing power into cash flow.”

What needs to be truly validated in NVIDIA’s results and outlook is no longer just next-gen AI clusters like Vera-Rubin’s resource demand, but whether AI inference-side computing power and revenue growth, pricing power, customer capital returns, and whether AI financing structures using credit bonds and asset securitization can jointly support the next “super rally” for AI computing.

Walsh Holds the Global Discount Rate Switch: 5.30% Long Bonds and $90 Oil Prices Squeeze Tech Stocks

Next comes a slower but far-reaching concern: interest rates/U.S. Treasury yields. Last week, the yield on 30-year U.S. Treasuries briefly hit 5.30%; subsequently, the Treasury announced it would at least double the size of long-maturity bond buybacks, an intervention that caused yields to dip slightly. Yet, by Friday’s U.S. close, yields had recouped nearly all their decline since the buyback announcement. The MOVE index, which measures rate volatility, hovers near 70—well below the March war-induced peak but far from the calm of winter.

The tension around yield expansion remains on a slow boil and has yet to reach a breaking point, just as the new Fed chair is set to deliver a key speech. Overarching all these market factors is geopolitical risk. Brent crude continues to hold above $90 a barrel, up more than 50% this year. This both supports energy stocks and amplifies concerns over inflation, Treasury term premium, and prolongs anxiety over the long-end bond market.

As above, option skews are rising, credit spreads are elevated—some market indicators show that investors remain cautious about taking risks.

August positioning also tells its own story. In the middle week of the month, hedge funds were net buyers of U.S. equities every trading day; the prior three weeks of buying were the largest since March 2020. Yet last week, hedge fund buying momentum faded instead of accelerating. Data from Goldman Sachs’ prime brokerage desk shows global equities saw their fastest net selling in two months last week. S&P 500 call option volumes have returned to normal from July’s frenzy; current buying looks like cautious re-entry, not the breathless, one-way melt-up chase of the past two years.

As above, the call-options-driven chase of a 4-million contract rally quickly normalized—latest daily volume shows no “risk at any cost” market mood.

Over $1 Trillion in Buybacks Builds the Floor, but $150 Billion of Programmatic Selling Lurks Behind a Trapdoor

As markets enter this week, a solid floor of support remains—but the wrong combination of outcomes could suddenly open a “trapdoor” below. About 96% of S&P 500 components are currently in corporate share repurchase windows, and over $1 trillion in buyback authorizations is providing steady buying support.

Trend-following funds (such as CTA strategy funds—known as “fast money”) hold about $140 billion in global equities, with their baseline scenario turning modestly positive. CTA and similar systematic trend followers are often considered “fast money” that quickly adjusts portfolios mechanically according to price signals.

However, risk structure is not entirely symmetric. Should the market see a significant pullback over a month, such systematic funds could trigger short-term global selloffs exceeding $150 billion in equities, with initial sell triggers just about 4% below current benchmark equity levels. The major events of this week will determine which side of this risk structure plays out.

Ultimately, fund flows continue to provide support, options gamma is stabilizing, and the rotation out of pure AI trades—especially into value and healthcare sectors that have long underperformed—is proving very healthy. More defensive asset allocation indicates some tempering of risk appetite; a steeper options skew curve suggests investors are adding partial hedges for volatility in rates/yields and uncertainty in AI’s intrinsic value. Should the market remain in summer’s quiet period, this setup could persist.

However, a negative news cycle would hit a market where both hedge funds and traditional asset managers are near this year’s lows for hedging, and various buffers remain untested. The floor is still there—until the trapdoor unexpectedly opens beneath it. This week, we may find out whether it can bear the weight.

With over $1 trillion in buyback authorizations, trend strategies holding about $140 billion in equities, and stabilizing options gamma, the market enjoys short-term buffers. However, global equities have flipped to net selling, and the July call option frenzy has clearly cooled. If this week’s two major events deliver a favorable combination, capital could flow back into the AI computing supply chain and spread even more healthily into high-quality cash flow stocks, energy, commodities, and stagflation beneficiaries—driving a broadening bull market. But if earnings and rates both disappoint, a roughly 4% index drop could trigger systematic strategies to sell over $150 billion—making the market floor appear solid, but hiding untested trapdoors below.

Bank of America’s strategy team is urging investors to step away from the crowded AI computing power trade. The firm believes that as the AI theme faces a deleveraging storm, increasingly challenging earnings expectations, forced unwinds of extremely crowded positions, and inflation risk challenges traditional portfolios, value stocks, biotech, regional banks, selected credit classes, and commodities all offer highly attractive investment opportunities.

As the AI bull market enters the stage of “high valuations + crowded trades + high capital consumption,” BofA advocates shifting marginal capital away from the priciest AI beta into “cheaper earnings growth, real cash flow, and anti-inflation assets”—this is not the end of the AI bull, but a rebalancing from a single tech narrative toward broad market earnings and high-quality cash flow. For example, legendary billionaire hedge fund manager Bill Ackman’s Pershing Square, as AI computing stocks undergo deleveraging and clearing of crowded positions, has switched its investment framework to “buying fundamentally high-quality stocks at distressed/discounted prices”—including new positions in digital payments and card network giants Visa (V.US) and Mastercard (MA.US), as well as four other companies with historically high cash flow quality and low portfolio concentration.