Reversal Day: Chips Give Back Yesterday's Surge, Three Hormuz Tanker Strikes Hand the Oversold Oil Its Catalyst, Samsung's Record Profit Meets a 6% Drop

This report is based on intraday data as of 12:31 PM ET and does not reflect closing prices. Markets may have moved since publication.

Recap of Yesterday’s Call

Yesterday’s (7/6) back-from-the-long-weekend risk-on bounce read last week’s “hardware bloodbath” as a violent snap-back — semis surged across the board, VIX slammed down to 16, and the market shrugged off even a real negative (NVIDIA’s next-gen Kyber rack slipping to 2028). Oil, meanwhile, was still pinned at an RSI-16 extreme oversold, waiting for a catalyst that could break the “linear normalization” extrapolation. Today both lines reversed: semis gave the entire surge back (Samsung’s record profit but 6% drop was the trigger), and oil finally got the catalyst it was missing — three commercial vessels struck in the Strait of Hormuz in a single day. All last week we stressed that “extreme oversold is not a reversal condition — it’s waiting for a catalyst.” Today that catalyst arrived.

Today’s Core Call

Today is yesterday’s mirror image: a full round of risk appetite got taken back. Semis sold off across the board inside Samsung’s “record profit but -6%” print, erasing yesterday’s surge and reconfirming that efficiency-vs-compute is violent rotation within AI, not an exodus from it. And the oil mispricing that had been priced at zero all last week finally bounced +2.8% on the catalyst of three Hormuz vessel strikes (the most concentrated since the ceasefire) — but with RSI still buried at 26, the market is reading it as “tension escalation,” not “blockade return.”

Macro & Geopolitical Deep Read

The cleanest mispricing of the past week finally got its catalyst today — but the shape of that catalyst decides whether it’s a “pulse” or a “repricing.” WTI printed $70.44 (RSI 26.56), +2.8% intraday, pushing the prior $68-69 extreme-oversold level up a notch — driven by geopolitics, not fundamentals. Within 24 hours, three commercial vessels were struck in the Strait of Hormuz: the Qatari LNG carrier Al Rekayyat was hit on the bridge by an “unidentified projectile” about 8 nautical miles east of Limah, Oman, sparking an engine-room fire; the Saudi-flagged supertanker Wedyan was damaged off the Omani coast; and a third unidentified vessel took projectile damage with “structural damage” (no casualties on any of the three). This is the most concentrated wave of attacks since the Hormuz ceasefire. Axios, citing two U.S. officials, says the IRGC fired at least two missiles on the night of 7/6 and that the U.S. is weighing retaliation; Iranian state media has not formally claimed responsibility but implied Al Rekayyat was struck after “ignoring warnings” — the direct fulfillment of the earlier threat that ships “not using the designated lane will meet a forceful response.” Trump, en route to the NATO summit on 7/7, escalated his rhetoric — “either we make a deal, or we finish the job… within an hour we can take out their bridges, destroy their energy supply” — with Iranian FM Araghchi retorting “Honor your signature.” This is exactly what the “zero-diplomacy + multi-party military warnings + highly sensitive funeral mobilization” window we named on 7/3 looks like when its latent catalyst gets lit. But read the price honestly: oil is only +2.8% and RSI is still 26, meaning the market is pricing a “tension pulse,” not a “blockade return” — because physical normalization is genuinely progressing this time (transit volumes recovering, the dark fleet adapting), and the fee/lane dispute remains a binary, event-driven tail. Which way it breaks still comes down to the far end of the Brent curve: a lift toward $80+ = the market starting to bet on a Day-60 showdown; a push back to $65 = a bet on a full return to pre-war. This attack lands just 4 days before the 7/11 Islamabad talks restart, ratcheting up the urgency of the Day-60 (~mid-August) fee dispute another notch.

Samsung cast the heaviest vote in the semiconductor efficiency-vs-compute debate with a “record profit but -6%” print. Samsung’s 7/7 preliminary Q2 results: operating profit of KRW 89.4 trillion (roughly 19x YoY, a record quarter), but revenue of KRW 171 trillion came in slightly below the KRW 172.18 trillion consensus — record earnings, yet the stock fell over 6%, dragging SK Hynix down 6% alongside, while Intel, AMD, and Applied Materials fell roughly 10%, 8%, and 8%+. That “profit at an all-time high, stock still down” pairing is essentially the market pricing memory-cycle-peak plus capex-digestion worries: the good number gets read as “this is as good as it gets.” It’s the mirror image of last week, when Micron’s beat was treated as single-name alpha that failed to spill over into a whole-chain re-rating. SMH is -3.3% today, taking yesterday’s +3.5% bounce back whole. Put last week’s “bloodbath,” yesterday’s “snap-back,” and today’s “re-sell” in the same week: this is not capital fleeing AI, but continuous violent repricing of leadership within AI (the compute/infrastructure leg and the platform/efficiency leg taking turns getting hit and chased) — efficiency-vs-compute is still an active fault line, but it’s rotation, not a retreat.

Chinese open-source models are pushing from the periphery toward the frontier, adding geopolitical fuel to the “compute narrative.” CNBC reported on 7/7 that AI startup Lindy has shifted 100% of its traffic from Anthropic Claude to DeepSeek; Zhipu’s Z.ai GLM 5.2 became the fastest-adopted model on Vercel in 2026, and Chinese models have entered the Top 5 on several developer platforms. On the same day, Anthropic’s Claude Fable 5 ended its included-access window within subscriptions and shifted to usage credits (API $10/M input, $50/M output), having become globally available again from 7/1 after an export-control exemption. More notable is a separate Reuters item on 7/7: China’s Ministry of Commerce is leading discussions with Alibaba, ByteDance, Z.ai and others about restricting overseas access to the most advanced Chinese AI models and raising penalties for model theft. The regime-level implications are twofold. First, if AI’s excess returns are migrating from “who rents the compute” to “who uses tokens most efficiently,” then the AI-capex leg propping up incremental U.S. GDP faces not just a demand-peak question but a “good enough and several times cheaper” open-source cost competitor. Second, with the U.S. using chip controls to choke the frontier and China now weighing restrictions on the export of its own models, the AI tech stack is being weaponized from both ends — a structural question late-July earnings season must answer.

Rates Read

Treasuries were unusually quiet today: 30Y at 5.03%, 10Y at 4.52%, 2Y at 4.14% — near-zero moves across the curve intraday, with RSIs all sitting in a neutral 53-57 band. That “quiet” is itself information: the 30Y sitting firmly above 5.0% with no drama is precisely “new normal confirmation” — the long end is hedged between energy disinflation and services/tariff price stickiness, stuck along the 4.8-5.0% line. But there’s an undercurrent this week worth watching: Treasury is issuing $119B (today’s $58B 3-year, a $39B 10-year reopening on 7/8, and a $22B 30-year reopening on 7/9). With the 30Y normalized above 5%, these three auctions’ bid-to-covers and tails are a real-time fiscal-dominance stress test, especially the 30Y long bond on 7/9. The long end that’s actually moving is still in Tokyo: JGB 30Y at 3.976% (RSI 77.46, +19bp over 5 days) is approaching the 4% threshold, and 10Y at 2.823% (RSI 70.19, +18bp over 5 days) — the hawkish BOJ path is lifting Japan’s long end one-directionally. Stitch the two ends together and you get two faces of the same fiscal-dominance coin: the U.S. long end pinned at 5.0%, Japan’s long end feeling toward 4%. FX was relatively calm today: USD/JPY at 161.92 (-0.1% intraday), with a soft weekly ADP employment read (four-week average of about 21K, below the prior) giving the yen a bit of support — but it remains in its 40-year-low range. Japan’s soft May wage data (nominal +3.2% YoY, below the prior 3.6%; real +1.4%, a fifth straight positive month; household spending -0.4%, a sixth straight decline) is a mixed BOJ signal, but MUFG’s July report still targets USD/JPY toward 165, on the logic that “markets don’t believe the BOJ will keep tightening + a rebound in Japanese imports after Hormuz reopens weighs on the trade balance.” Carry-unwind’s Path A (spread compression) is thus deferred once again; the tail risk of Japan’s MoF “ambush-style” intervention persists, but no actual intervention was confirmed within the window.

Sector Spotlight

AI / Semis: reversal to the downside, the compute/infrastructure leg bleeding again. SMH -3.3%, AMD -5.07%, TSM -3.14%, Intel -8.57%, Micron -5.70% (-19.54% over 5 days), Vertiv -5.09% all weakened, triggered by Samsung’s “record profit but -6%.” The relative resilience sits in the mega-caps themselves — NVDA +1.07% (outperforming the sector on a down day), AVGO relatively firm at -0.78%; the platform/efficiency leg kept its footing — MSFT +1.80%, META +0.98%, Palantir +2.73% (+16.70% over 5 days). Signal value: a week-scale round-trip of yesterday’s surge and today’s plunge shows this is leadership rotation within AI, not a complex-wide retreat — and Samsung’s print put the memory-cycle-peak worry (“record earnings != share-price support”) squarely on the table.

Traditional Energy: extreme oversold meets a geopolitical catalyst, finally bouncing. WTI RSI 26.56, Occidental RSI 27.5, Chevron RSI 32.8 — the whole sector had been buried in extreme oversold, and today it bounced collectively on the catalyst of three Hormuz vessel strikes: WTI +2.8%, ExxonMobil +2.16%, Chevron +1.54%, Occidental +2.49%. This is exactly the setup we kept stressing last week — “extreme oversold + a catalyst that can break the linear-normalization extrapolation” — but with RSI still deep in oversold territory, the market is reading this wave as an “event pulse,” not yet an escalation into a structural repricing of the Day-60 fee.

Agriculture / Fertilizer: decoupled from the oversold energy complex, quietly strengthening. CF printed $113.68 (+0.42% intraday, +5.01% over 5 days, RSI 67.9 nearing overbought), Nutrien +0.50%, Mosaic -1.53%. The nitrogen leader’s RSI pushing toward 70 while energy sits deep in oversold reflects an independent line — “the cheapest natural-gas feedstock + the world’s scarcest nitrogen supply” — decoupled from the front end of oil.

Digital Assets: weak-dollar beta keeps running, but this still isn’t the QE it’s waiting for. BTC printed $63.9K (RSI 55.45, +9.1% over 5 days), MicroStrategy +16.70% over 5 days, Coinbase +14.49% over 5 days — the crypto complex extends its recovery under a weak dollar and risk appetite. But be clear on the nature of it: BTC is a two-phase asset — in Phase 1 (risk appetite / weak dollar) it rides along, and only in Phase 2 (Fed money printing / QE) does it truly take off. With RSI still neutral today and the drawdown from its high still near -49%, this is beta, not the alpha of a liquidity flood.

China: the most extreme independent read on the board, with another layer added today. China large-cap FXI has an RSI of 23.61 and a -20.8% drawdown, still one of the deepest oversold readings anywhere — reflecting the independent line of U.S.-China tech decoupling. Today’s Reuters report that “China plans to restrict overseas access to its most advanced AI models” adds another layer to that decoupling line: tech containment is shifting from one-directional (the U.S. choking off China’s chips) to bidirectional (China now also weighing restrictions on the export of its own models).

What to Watch & Forecast Framework

7/7 after-hours API + 7/8 EIA weekly (delayed by the holiday): Watch whether the draw persists. If it does → the divergence between paper (RSI 26) and physical draw holds, and the extreme-oversold recovery elasticity remains; a first shift to a build → “physical normalization” gets stamped by the data, hardening the lower bound of the oil range.

7/8 FOMC minutes (6/16-17 meeting): The “rate-cut vs. stagflation” debate that ISM temporarily pushed aside yesterday gets its true textual verdict here. Watch whether Warsh’s hawkish path in the minutes is committee consensus or a market-extrapolated tail — a confirmation of “inflation first, committing to neither a hike nor a cut” → the 2Y stays high, spreads don’t compress; any signal of concern about softening employment → the window for restarting rate-cut pricing truly opens.

7/8-7/9 Treasury auctions ($39B 10Y + $22B 30Y): With the 30Y normalized above 5%, bid-to-cover and tails are a real-time fiscal-dominance stress test; a weak auction (wider tail) adds kindling to the long end’s upside.

7/11 U.S.-Iran talks restart (Islamabad): The three vessel strikes land 4 days before the restart, with the focus still stuck on the three open questions — the nuclear program, sanctions/frozen assets, and Hormuz stability. Watch whether this attack disrupts the negotiating framework itself. Watch the far-end Brent: a lift toward $80+ = the market beginning to price a Day-60 fee showdown; a push back to $65 = a bet on a full return to pre-war.

7/24 Section 122 global 10% tariff expiry (17 days out) + USTR Section 301 forced-labor hearings (7/7-7/9): The latter proposes 10%-12.5% tariffs on relevant imports from 60 economies; renewal, escalation, or a switch to Section 301 are all possible, entering the window where the market is about to start pricing it in.

7/29 FOMC + Day-60 fee showdown (~mid-August, aligned with the 8/21 sanctions-waiver expiry): The former is the true landing point of this round’s “rate-cut vs. stagflation,” so watch the direction of July CPI/PCE; the latter remains the only unresolved binary variable, and today’s three vessel strikes pushed that “zero-weight” tail one step closer to center stage.

Late-July earnings season (structural slow variable): The efficiency-vs-compute debate + open-source/Chinese models (DeepSeek, GLM 5.2) as “good enough and several times cheaper” cost competitors are the question the AI-capex leg propping up incremental U.S. GDP must carry into the season. Vistra’s Q2 results are set for 8/7.

Risk Disclaimer

This article is public market commentary and personal research notes. It does not constitute investment advice.