Yields Hit 5.20%|Who Sells NVDA Tonight?

Yield Strike

The S&P 500 just lost its third straight session into Nvidia's print, and the obvious read is that traders are derisking ahead of earnings. The 30-year Treasury yield tells a different story. It pushed to 5.20% on Tuesday, the highest since the eve of the 2007 financial crisis, and the move was not driven by oil — crude actually drifted lower on the day. That detachment is the signal. When duration sells off without an energy catalyst, the market is repricing the discount rate itself, and growth multiples sit at the front of that line. The Nasdaq 100 dropped 1.4% in the morning, marking its third consecutive decline, and the index is already losing altitude from its overbought reading on the relative strength index. The capital that left was not parked in cash. Trading volume in 10-year futures ran near twice the recent average with sizable block trades, which is institutional repositioning out of long-duration equity exposure into shorter-dated paper, not retail panic. Bank of America's May survey caught the setup in advance — long global semiconductors is now the most crowded trade in survey history at 73%, up from 24% in April, and cash levels fell to 3.9%, below the 4% threshold that historically triggers a sell signal. The structural read is that yield discipline is doing the work that valuation discipline stopped doing months ago. Fed funds futures now price a 58% chance of a December hike, a complete reversal from three expected cuts when the Iran war began in late February, and Yardeni is already calling for a July move. The chapter's real question is not whether NDX corrects from here. The 50-day moving average sits 9% below current levels and lines up with the October 2025 and January 2026 highs as a confluence resistance flip. The question is what kind of buyer steps in if Nvidia delivers tomorrow and yields refuse to back off — because the holders forcing this selloff are not the names usually cited.

Short Book

That holder question has a name on it now, and it complicates the bullish read on Nvidia's print. Leopold Aschenbrenner's Situational Awareness fund, run by a former OpenAI alignment researcher, disclosed roughly $1.56 billion in put options on NVDA in its latest 13F, plus more than $2 billion of bearish positioning tied to the VanEck Semiconductor ETF, with additional put exposure on AVGO, ORCL, AMD, TSM, ASML, and INTC. The mechanism deepens what chapter one only described — the yield-driven selloff is not anonymous flow, it is overlaid by a concentrated institutional short book that profits as the crowded trade unwinds. Options markets are now pricing an 8.65% post-earnings move in either direction on Nvidia, well above the roughly 6% the Investopedia data implied earlier in the day, and that gap is the volatility premium for exactly this asymmetry. A clean beat sends NVDA back toward the $233 area near its record. A miss, or in-line guidance against $83-84 billion implied by Arya's beat-history model, sets up gamma unwinds that the short side has already positioned for. Sentiment confirms the shift in marginal buyer. Evercore's survey shows only 35% of investors picking Nvidia as their top AI logic pick, half the 70% reading from July 2025, and NVDA is up just 19% year to date against AMD and Micron above 100%. The capital flow trace here is specific — discretionary AI-overweight funds are being pushed out of the most crowded single name by valuation-sensitive shorts and by passive rebalancing into laggards like MU, where Citi just nearly doubled its target to $840 on DRAM pricing. The supporting case for Nvidia's print remains intact: hyperscaler capex commitments of $200B at Amazon, $190B at Microsoft, $185B at Alphabet, $135B at Meta, and Nvidia's own supply commitments doubling from $50.3B to $95.2B between fiscal quarters. The counter-signal is what BofA's Arya flagged as the unresolved overhang — only 47% of free cash flow from 2022-2025 went to dividends and buybacks versus an 80% peer average, leaving Nvidia owned by just 16% of equity income funds against a 32% peer average. The marginal buyer that absorbed every prior dip — the income-oriented fund pool — is not in this name at scale. If the print disappoints and yields stay above 5.2%, the natural absorber is missing, and that is the asymmetry the short book is paid for.

IPO Drain

If the AI trade is losing its marginal buyer in public markets, the question becomes where that capital actually goes — and the IPO calendar is the answer the data points to. Cerebras priced its IPO at $185 on May 14 and opened at $350, hitting $385 intraday on a 20-times-oversubscribed book before settling near $280 by Friday, briefly carrying a roughly $100 billion valuation on 2025 revenue of $510 million. The institutional bid did not value current cash flows. It valued positioning ahead of the next wave, and the next wave has a date. SpaceX is reportedly targeting a Nasdaq listing as early as June 12 under SPCX at a $1.75 trillion valuation and a $75 billion raise, which would exceed Saudi Aramco's $29.4 billion debut by more than 2.5 times and stand as the largest IPO on record. The transmission into existing tech is direct. Alphabet's 6.11% stake from a 2015 Series F at a $12 billion mark would be worth around $107 billion at the IPO valuation — that is balance sheet appreciation that does not depend on Gemini 3.5 monetization, which still slipped to next month at Tuesday's I/O. Nvidia's exposure is forward-looking through the xAI Colossus data center with 220,000 GPUs plus a Saudi facility carrying 600,000 GPUs, and SpaceX's $75 billion cash raise becomes incremental AI capex demand that flows back into Nvidia order books. The capital flow trace is asymmetric — institutional money that exits crowded public AI names through Aschenbrenner-style structural pressure does not leave the AI theme, it rotates into private-market beneficiaries via IPO allocation. Polymarket and Nasdaq Private Market just launched private-company prediction markets on Tuesday, formalizing that rotation channel. The leaning tilts toward an asymmetric reaction tomorrow. A Nvidia beat with strong Vera Rubin commentary and any capital-return signal closes the equity-income-fund gap that chapter two flagged, and yields can back off without sending NDX to its 50-day. A Nvidia print that confirms hyperscaler capex but offers no buyback acceleration leaves the short book paid and the 30-year yield as the binding constraint, with the SpaceX June 12 window absorbing the capital that exits. The benchmarks to watch by close Wednesday are concrete — 30-year yield holding above 5.15%, NDX failing to reclaim the prior day's high, and the implied move on NVDA options decompressing toward 6% rather than the current 8.65%. If yields break back below 5.10% on the print, the bearish read collapses fast. That is the condition that would prove the short book wrong.