Today is a genuinely broad red session, not a single-stock story: the S&P 500 is down a little over 1%, the Nasdaq-100 down closer to 2%, and small caps are tracking right alongside the Nasdaq rather than holding up better. The proximate cause is on the rate side — see the Econ Briefing tab for the full read, but in short, the 10-year Treasury yield just printed its highest level in this log's history, and that's exactly the kind of move that hits long-duration growth names hardest. AI-infrastructure and power names are leading the decliners today, with a couple of the more speculative growth names down high single digits; financials and energy are holding up noticeably better, consistent with that group's lower rate-sensitivity.

Catching up on the headline event from the gap since the last post: on Oct 5, every account in this book added a sector-ETF position the same day — a coordinated rotation into financials, industrials, energy, and technology-sector funds, plus a healthcare-sector top-up, closing out several allocation gaps that had been sitting open for a while. See the new Trade Log card for the specifics. It's a diversification move, not a timing call — it happened to land a little over a week before today's broader pullback, which is coincidence, not foresight.

What to watch: whether the 10-year's climb above 5.2% holds or fades. The new sector-ETF sleeve is a partial hedge against more of this kind of move (financials in particular tend to benefit from a steeper curve), but it doesn't fully offset the book's existing tech concentration.

Monday looked flat at the index level — the S&P 500 slipped about 0.3%, the Nasdaq-100 was essentially unchanged, small caps were the softest — but underneath, the market moved along one clear line. Crude oil rallied about 3%, and the energy names went with it: the two integrated oil majors in this book's coverage were the standout gainers, each up roughly 2-3%. On the other side, the parts of the market that trade like bonds sold off — the utilities sector ETF and the broad real-estate ETF were the day's laggards, both down around 1%.

Energy up while utilities and REITs fall, on a day when Treasury yields held their recent climb, is the market pricing a firmer-for-longer rate path rather than a growth scare. It's a rotation, not a de-risking: the volatility index actually eased slightly, and gold sat still. The one real casualty on the growth side was the mega-cap complex — a large e-commerce name and a big search name were both off around 2%, which is what dragged the tech-heavy part of the tape lower even as the AI-chip leader steadied after its Friday drop.

The day's biggest single winner in this coverage was a large electric-vehicle maker, up more than 5% on genuine news: a state regulator approved its robotaxi service as an autonomous-vehicle network operator, clearing the way for a commercial fleet, with a product event scheduled for early September on top of that.

What to watch: a cybersecurity leader and a PC/server hardware name both report Tuesday Sep 1 after the close, with a large semiconductor-infrastructure name on Sep 2 — the read-through test for the rest of the AI supply chain. Whether the energy bid holds is the other thread; a multi-day move would say the rotation has legs.

Quick note: this tab skipped its usual cadence for about two weeks — catching up here on the one session that actually matters for this book.

Nvidia reported after Wednesday's close and it was a clear beat on every line that matters: revenue up 106% year-over-year, data-center revenue making up the overwhelming majority of the total, gross margin holding around 75%, and forward guidance for next year running well ahead of expectations — management pointed to roughly 70% revenue growth for the coming fiscal year. The stock rose about 9% the following session. As the single largest position in this book, that move drove most of a +1.7% portfolio day on its own.

The bigger point is the read-through. A beat this size from the company at the center of the AI buildout is a demand-signal reassurance for the entire adjacent complex — the networking, server, power, and cooling names that only work if hyperscaler capex keeps climbing. Several of those held in this book moved up in sympathy. It doesn't retire the risk (a capex pause would still show up first in a hyperscaler CFO's comments, not Nvidia's), but it pushes that risk further out.

Palo Alto Networks was the day's other big mover, up around 14%, on reporting that it is shifting into acquisition mode — circling two mid-size cybersecurity/data names after being rebuffed on larger targets earlier — amplified by a strong quarter from a cybersecurity peer that lifted the whole group. That's sector sentiment plus a company-specific catalyst stacking, ahead of Palo Alto's own earnings on Sep 1. Palantir added about 5% as well.

Underneath the winners, the defensive corners were quiet-to-soft — healthcare, staples, and real estate all lagged on a risk-on day, and volatility eased to the mid-14s. Gold held roughly flat.

What to watch: Palo Alto and Dell both report Sep 1 after the close, and a large semiconductor-infrastructure name follows Sep 2 — the next real test of whether the Nvidia read-through holds across the rest of the supply chain. On the macro side, the September Fed meeting is the bigger event (see the Econ Briefing tab).

Today's tape was a genuine broad-based rally, not just a handful of names dragging the average up: the S&P 500 rose roughly 0.7% and the Nasdaq-100 outpaced it at close to 1.2%, with real stock-specific stories layered on top rather than just a macro tailwind doing all the work.

Palantir led the gainers, up nearly 5% on continued momentum from its recent quarter — 93% revenue growth, over $1B in operating cash flow, and US commercial revenue up triple digits year-over-year, with its government-AI positioning staying in the spotlight. Tesla jumped almost 4% on real news: a proposed multi-billion-dollar solar-production facility in Texas, internally dubbed "Project Crystal Sun." Worth noting the stock still carries an unresolved valuation debate from its last earnings print — a real EPS miss against a stretched multiple — so today's pop is a sentiment win on a single headline, not a reversal of that concern.

On the other side, a couple of AI-infrastructure names gave back ground with no negative catalyst attached to either move — one of them actually got a price-target raise from a major sell-side desk on the same day it traded down, a reminder that not every red name on a green day means something broke. A large healthcare name was a rare decliner with an actual identifiable catalyst: a new lawsuit alleging cybersecurity and governance failings, layered on top of ongoing regulatory scrutiny of its Medicare Advantage business — worth tracking, not yet a thesis-changing event on its own.

What to watch: whether today's rally has legs into the rest of the week, or whether it was a one-day bounce riding a temporary dip in long-term interest rates (see the Econ Briefing tab for the rate-side read). No large-cap names in this book's coverage report earnings in the next week, so the next real catalyst window is still a couple weeks out.

Quick note: this log skipped a post Jul 29 — catching up here rather than reconstructing a day's live color after the fact. Jul 29's numbers below are real, official closes; the color is what's confirmed, not invented.

Jul 29's headline event was the FOMC decision, and it landed almost exactly as priced: rates held at 3.5–3.75%, the fifth straight hold, though three dissenting officials wanted a 25bp hike — a real signal that inflation concerns (tied to both the Iran-conflict energy premium and AI-driven infrastructure bottlenecks) are gaining a voice on the committee, even if it didn't change the decision itself. The same evening, Bloom Energy's Q2 print resolved the trim-vs-hold question flagged on this log a few days earlier — a genuine beat, not the feared miss, with no new post-earnings low.

Jul 30 is where the real dispersion showed up. Microsoft and Meta both reported the night before, and the two prints could not have landed more differently. Microsoft beat cleanly: Azure grew 43% year-over-year and crossed $100B in full-year revenue for the first time, Copilot passed 30 million paid seats, and commercial remaining performance obligations — contracted future revenue — jumped 84% to $678B. The stock jumped double digits. Meta went the other way: EPS missed ($6.18 vs. $7.22 expected) and forward revenue guidance came in light despite 28% year-over-year revenue growth — margin bled even as the ad business kept growing. Two names with a similar AI-capex story, one clean beat and one real miss, on the same earnings night.

Bloom Energy kept running well past its own earnings reaction — up sharply again today on fresh news: a new AI-power financing deal in the billions plus a sell-side note flagging Bloom as the likely fuel-cell supplier for a pair of large AI data centers in Panama. Broad chip and hardware names rallied alongside Microsoft's Azure-capex-confirming beat, a reminder that a single hyperscaler's guidance can move an entire adjacent sector. The VIX, meanwhile, fell sharply — confirming this reads as a genuine risk-on session, not a nervous one, even with a real earnings miss in the mix.

What to watch: Apple and Amazon both report tonight (Jul 30, after close) — the last pair of the busiest earnings stretch this log has tracked all summer. Early-circulating reports point to Apple beating on both lines but guiding weak enough to move shares down after hours, and Amazon beating on AWS growth — neither is confirmed or reflected in tonight's official closes yet.

Today's tape split cleanly into two stories, and neither one was about rates. Coca-Cola reported before the open and beat on both lines — EPS and revenue both ahead of Street estimates, volume up 5% (its biggest quarterly jump in 17 years outside the pandemic), and full-year EPS growth guidance raised from 8–9% to 9–10%. Shares jumped roughly 7% to a fresh 52-week high, a clean example of a beat-and-raise getting rewarded rather than "sold on the news" the way some recent earnings reactions have gone.

On the other side of the tape, chip-adjacent names kept sliding for a second straight session, though today's driver looks different from Monday's China-lithography headline. AMD fell about 3.4% on fresh concerns over AI-accelerator demand guidance, compounded by reports of tighter potential export restrictions on high-performance compute headed to Middle East data centers — a direct hit to the international-expansion story for AMD's newer accelerator lines. The pressure wasn't limited to the US: Japanese and South Korean chip stocks fell hard overnight, with Samsung down more than 7% and SK Hynix down over 9%, dragging the Korea Composite index down more than 7% on the session. Dell fell roughly 9% despite not being due to report earnings until September — a reminder that a stock can move sharply on capex/margin and valuation worries alone, without any company-specific news that day. Palantir fell nearly 6% on a mix of open-source competitive pressure (a free alternative to one of its platforms surfaced on GitHub this month) and regulatory scrutiny of a UK healthcare-data contract, continuing a rougher stretch heading into its own Aug 3 earnings.

On the rate side, still nothing new: the Fed funds rate sits at 3.63%, the 10-Year Treasury yield at 4.69% (up about a quarter point over the trailing month), and the 2-Year at 4.33%. Worth noting, though, that despite the visible chip-sector stress, the VIX — the market's fear gauge — actually fell today rather than rose, a sign this reads as a sector-specific rotation rather than a broad flight from risk.

What to watch: tomorrow, Jul 29, is shaping up as the single busiest day of the summer for markets — the FOMC rate decision lands at 2pm ET (consensus expects a hold at 3.5–3.75%), the same day Microsoft and Meta both report earnings after the close. Bloom Energy also reports tonight, continuing the busiest earnings stretch of the year.

Semiconductor stocks sold off sharply Monday afternoon — figures below are intraday, as of roughly 2:40pm ET, with the market still open — after a report that a Shanghai state-backed firm has begun mass-producing homegrown immersion DUV lithography machines, the class of equipment used to manufacture advanced chips. The news raised fresh fears about how quickly China can build domestic chipmaking capacity independent of Western suppliers like ASML. AMD led the decliners, down 7.5% intraday, with Teradyne down 6.1%, Nvidia down 5.1%, the semiconductor-sector ETF SMH down 3.7%, and Micron down 4.5%. Micron's drop carried a second catalyst on top of the equipment headline: Chinese memory maker ChangXin Memory Technologies (CXMT) went public in Shanghai and saw its shares soar more than 500% on debut, instantly making it the world's fourth-largest DRAM maker at roughly 8% global market share, with reports that Apple has begun testing its chips — reviving the same China-competition anxiety on the memory side. The selling also picked up spillover momentum from a negative post-earnings reaction in Intel shares.

CNBC's Jim Cramer, watching the reversal unfold, described the sellers driving it as "monstrous, motivated and often margined" — a reference to leveraged positions unwinding in names like AMD and Intel that had been up sharply for the year. Selling of that kind tends to be harder to arrest than an ordinary pullback driven by buyers simply stepping back, because forced or margin-driven selling doesn't wait for a better price.

Crude oil moved in the opposite direction, falling hard after the US and Iran reportedly paused strikes against each other over the weekend, unwinding much of the war-risk premium that had built into prices on fears of a Strait of Hormuz supply disruption. Brent crude fell about 5.2% to $91.73 a barrel and WTI fell about 5.4% to $84.45, per a Monday-morning read from Rigzone; other outlets had the intraday move running anywhere from roughly 4% to more than 8%, so the decline is best read as landing somewhere in the 5%-to-6% range rather than as a single precise number.

With chip names selling off and oil falling, money rotated into defense, megacap tech, and defensive consumer names. RTX rose 3.4% on a new multi-billion-dollar Department of Defense contract for next-generation integrated air and missile defense systems, adding to a recent beat-and-raise quarter and record backlog. Microsoft gained 2.9% heading into Wednesday's fiscal Q4 report, with Guggenheim reiterating a Buy rating and raising its price target to $586 the same day. Coca-Cola rose 2.2% in a classic rotation into defensive, non-cyclical names ahead of its own Tuesday earnings. Alphabet added 1.9% on broad-based megacap tech buying tied to continued investor enthusiasm for its AI trajectory, including Gemini usage growth. Separately, Palantir jumped 6.7% after Citi analyst Tyler Radke maintained his Buy rating — trimming his price target to $200 from $225 on a lower software-sector valuation multiple, but raising his 2027 revenue-growth forecast to 53%, well above the Street's 45%, and calling for a commercial rebound ahead of the company's own earnings.

The practical read: the net effect on the broad market was muted relative to the size of the individual moves. The S&P 500 slipped roughly a quarter to a third of a percent to 7391.17, the Nasdaq-100 fell about 0.9% to 27873.28, the small-cap Russell 2000 actually held slightly positive at 2937.92, and the VIX rose about 4.7% to 19.46 — a sign of genuine risk-off tone underneath the surface even with oil falling. Today isn't one story, it's three running at once: a China-driven derating of chip-equipment and memory names, a geopolitical de-escalation trade in oil, and a rotation into names with their own earnings- or contract-driven catalysts. None of them cancel each other out; they just happened to land on the same tape.

Bloom Energy's slide got worse into Friday's close than it looked at midday — and a weekend news check found the reason is more serious than the "valuation call" framing used earlier this week. The overhang has escalated from a sell-side desk calling the stock fully priced to a short-seller report now under a securities class-action investigation over supply-chain and production-capacity claims. That's a materially different risk category: a valuation debate resolves with an earnings print: a fraud allegation resolves with an investigation timeline that can run for months regardless of what Tuesday's numbers show. Earnings land Jul 28, but this one might not get fully resolved by the report alone.

Zooming out, the week ahead is the heaviest earnings stretch this book has tracked all year: five separate reporting days from Tuesday through Friday, including two pairs of same-day mega-cap reports. That's arriving on top of a Treasury yield that's now four straight sessions into a climb — a rising discount rate stacked against still-elevated AI-capex guidance is exactly the kind of setup that punishes a soft print harder than it would in a calmer rate environment.

The practical read: two genuinely different stories are converging in the same week — a company-specific credibility question on one name, and a systematic rate/earnings-season risk sitting on the broader growth complex. Worth tracking them separately rather than lumping every red print into one narrative, since they resolve on different timelines and for different reasons.

Tesla's slide is the clearest case of the three. Wednesday's report was a genuine fundamentals miss, not a sentiment wobble: auto gross margins landed at 16.3% against an 18.4% consensus, EPS of $0.33 missed a $0.55 estimate, operating margin fell to 1.4% from 4.1% a year earlier, and free cash flow turned negative as AI/robotics capital spending ramped. Today's further decline is a continuation of that reaction, not a new catalyst — worth treating as a real, still-unresolved story rather than noise.

Bloom Energy is a different kind of story entirely. The stock extended a roughly month-long slide today despite a genuinely strong quarter and a large new AI-cloud financing deal that, taken alone, would read as bullish. The actual driver looks to be valuation: at least one sell-side desk flagged infrastructure-buildout delays at two major cloud/utility players as a risk to the company's 2027–28 estimates and called the stock fully priced at its current multiple. Earnings land Jul 28 — four days out — and will be the real test of which read is right.

Apple sits at the opposite end of the spectrum: solidly higher today with no single identifiable catalyst, continuing a run of outperformance into its own Jul 30 earnings. Sometimes a green day really is just momentum, not a story.

The practical read: three names, three different reasons to move, all of which would look identical on a one-day chart. A confirmed earnings miss, a valuation debate ahead of results, and plain momentum only reveal themselves once you ask what's actually driving each one — the direction alone tells you almost nothing.

Markets were roughly flat on the surface today, but that calm masked real dispersion underneath. A leading server maker jumped nearly 10% after a competitor's after-hours update the night before reinforced that enterprise AI-server demand remains strong — a GPU designer and a chip designer with a diversified compute business both extended gains too, marking a third straight session where AI-infrastructure-demand news has outweighed the rate backdrop for that corner of the market. Rising Treasury yields would normally argue against exactly this kind of long-duration, high-multiple rally, but stock-specific confirmation is winning out again, for now.

On the other side, two mega-cap software names extended multi-month drawdowns ahead of their own upcoming earnings — one is down roughly 18% year-to-date on investor skepticism over its AI-capex spending plans. A prominent AI-software name fell hardest of all with no specific news attached, reading as broad pre-earnings de-risking in richly-valued names rather than a new negative catalyst. Oil prices also climbed on the day, adding a modest inflation-watch wrinkle to an otherwise quiet macro backdrop.

The bigger story is what happens after today's close: a search-and-cloud giant and an EV/energy-storage name both report earnings tonight — the first live test of this reporting season for names actually held in this book. Two defense names follow with earnings before tomorrow's open.

The practical read: a flat index-level day with double-digit dispersion underneath is a sign the market is already sorting winners from losers ahead of results, not waiting for them. Tonight's prints, followed by two more mega-cap reports in the next two weeks, will do more to settle the "is AI capex still justified" debate than any macro data point on the calendar right now.

Four days after the Philadelphia Semiconductor Index (SOX) formally confirmed a bear market, the sector staged one of its sharpest reversals of the summer. The catalyst was two-fold: a newly announced supply deal between a leading GPU/CPU designer and a major cloud provider, expanding that designer's compute, networking, and software footprint for AI workloads on the cloud provider's platform; and a bullish call from Morgan Stanley predicting memory-chip prices will keep rising on sustained AI demand. Memory names led the move, up double digits, with the GPU designer and a hardware-integration name both posting high single-digit gains, and a chip foundry name up meaningfully as well.

The reversal is a useful reminder of how fast sentiment can swing in a sector this headline-sensitive: the same names selling off on AI-monetization doubts a few sessions ago are rallying today on evidence that AI-related demand (for memory, for compute supply) is still very real and still growing. Neither the bear case nor today's bull case has been fully settled — this is best read as the market repricing probability, not as either side "winning."

Separately, a fuel-cell name that's been caught between a bullish infrastructure deal and unresolved short-seller allegations for over a week finally got a decisive, one-sided catalyst: a major bank raised its price target meaningfully, introducing multi-year delivery estimates backed by existing customer agreements. The stock rallied double digits, the sharpest single-day move of the group.

The practical read: a reversal this sharp, this fast, doesn't resolve the underlying AI-capex-sustainability debate — it just shows which side of it the market is leaning toward today. The next real test is still the wave of Big Tech earnings landing over the next two weeks, where actual capex guidance (not analyst speculation) will do more to settle the question than any single day's price action.

The semiconductor sector has now formally entered a bear market. The Philadelphia Semiconductor Index (SOX) fell roughly 20% from its late-June record, dropping from a peak near 14,655 to about 11,674 by Jul 17. SMH, a broad chip-sector ETF, separately posted its worst single week since March 2025 (down about 11% in that one week) on the way down. Sector ETFs are stabilizing today, with SOXX up 0.41% and SMH up 0.29%, but that's a pause, not a reversal — the underlying index is still deep in bear-market territory.

The proximate cause isn't a rate shock or an earnings miss at any single company — it's a valuation and growth-expectations question. Investors are increasingly skeptical that the current pace of AI infrastructure spending across the industry can keep justifying itself, and that skepticism sharpened after a competitive new Chinese AI model, Kimi K3, chipped away at assumptions about Nvidia's dominance in AI compute. It's a confidence crack in the AI-buildout trade itself, not a breakdown in any one company's fundamentals.

Several sell-side banks are already pushing back with a contrarian, buy-the-dip case. Morgan Stanley named NVDA and AVGO its top AI-compute picks amid the selloff, calling the pullback an attractive entry point. JPMorgan told clients the drawdown is a buying opportunity, citing tight AI chip supply-demand dynamics that it expects to persist through 2028. Mizuho's Vijay Rakesh went further, saying there's "still a lot of gas left in the tank." The market seems to be listening, at least today: NVDA is up 0.25% and AMD is up 1.56%.

The practical read: bulls and bears aren't actually disagreeing about whether AI infrastructure spending continues — both sides assume it does. The disagreement is about whether the returns on that spending will show up fast enough to justify current multiples. That's not a question today's price action answers either way; it's a question the next few quarters of hyperscaler capex guidance and AI-revenue disclosures will have to settle.

Two AI-adjacent names are sitting on very different kinds of uncertainty today. Bloom Energy (BE) fell 7.62% even after Brookfield announced it is expanding its AI-infrastructure partnership with the fuel-cell maker fivefold — from $5B to $25B — to finance power for AI data centers. Normally that kind of headline would be an unambiguous positive. Instead, the stock is being pulled down by two unresolved short-seller reports: Hunterbrook alleges an undisclosed dependency on Chinese scandium supply chains that contradicts the CEO's public statements of having "no China supply chain," and a follow-on report from Crossroads Capital makes similar supply-chain claims while also questioning management credibility and predicting a further share repricing. Bloom has filed an 8-K formally rejecting both sets of allegations, but the dispute is genuinely open — a real corporate deal on one side, real unresolved claims on the other, with no independent resolution yet.

SpaceX (SPCX) is dealing with a more straightforward, if still costly, setback. The stock is down roughly 45% from the intraday high it hit on Jun 16 — four days after its Jun 12 IPO and Nasdaq debut — and now sits at $120.02, down 3.21% just today. The trigger was a Jul 16 test flight of Starship's 13th vehicle, which was automatically aborted when its engines failed to start. Musk has since rescheduled the next attempt for Thursday, Jul 23, which is the clear near-term catalyst for the stock. Notably, Cathie Wood's ARK reportedly bought into the drawdown, adding roughly $51M of SPCX shares while the stock was falling.

The practical read: BE is a "wait for more information" situation — the bull case (a real, large infrastructure commitment) and the bear case (real, specific supply-chain allegations) are both concrete and neither has been independently confirmed or debunked yet. SPCX is more mechanical: Thursday's launch attempt is a binary near-term event that should move the stock meaningfully in either direction.

A broad, sector-wide selloff hit semiconductors today, capping their worst week since the tariff-driven meltdown of spring 2025 — the group is now down roughly 20% from its recent record. This wasn't one company's earnings miss; it was a confidence crack in the core AI-infrastructure trade itself, on doubts that the scale of AI-capex spending across the industry can keep justifying itself at current valuations.

The catalyst sharpening those doubts: a Chinese AI startup released a competitive new open-weight model, chipping away at the assumption that US AI leadership requires the current pace of chip and data-center spending. Layered on top: a major streaming name dropped double digits on disappointing earnings, and oil prices climbed on renewed Middle East tension — three separate pressures landing in the same session rather than one dominant story.

The largest AI-chip designer, a leading chip designer with a diversified compute business, and a newly-public space-launch name (still working through its own valuation story from recent weeks) were among the names hit hardest — all central to the AI-buildout trade this selloff is questioning directly. On the other side, energy majors, cybersecurity, and gold all held up or gained — a fairly clean rotation away from the richest AI-capex-exposed names into more defensive or commodity-linked ground, not a broad flight from equities generally (major indices were down more modestly, in the 1-1.5% range).

The practical read: the real test isn't today's price action, it's whether AI-capex guidance actually gets trimmed at the wave of Big Tech earnings landing over the next two weeks. Until then, this reads as a valuation reset built on a real (if unresolved) question, not noise to be ignored.

No new Fed or Treasury data landed today, so this session's moves are coming from company and sector news, not a fresh rate signal — and it produced a genuine split within the tech trade rather than a broad move in either direction.

A major PC/server hardware name fell double digits on reports that a large social-media company is exploring leasing out its own surplus AI compute capacity to enterprise customers — a signal that hyperscalers may have over-built data-center capacity, which would mean fewer future server orders for integrators. That was compounded by rising memory-cost margin pressure and a fresh analyst downgrade to Hold after the stock's roughly 200% year-to-date run put its valuation on shakier ground. A leading memory-chip maker (on this site's watchlist, not currently held) dropped nearly 8% on reports of Chinese DRAM competitors gaining pricing power plus possible tighter U.S. export controls on high-bandwidth memory — dragging the broader chip/memory complex, including a networking-hardware name and a chip-designer name held here, down alongside it.

On the other side of the same trade, mega-cap software did the opposite: a search/cloud giant, a consumer-electronics maker, the largest enterprise-software name, and the largest e-commerce/cloud name all traded up together, along with a major bank. The largest individual AI-chip designer held here sat out both stories entirely, essentially flat — it designs chips rather than integrating servers or manufacturing memory, so neither narrative applies to it directly.

The practical read: this reads as a rotation within the AI/tech trade — hardware and memory names down, mega-cap software up — not a broad risk-off day or a reason to rethink tech exposure as a whole. Worth watching whether the "AI-server overbuild" narrative spreads beyond the one hardware name into the broader server-supply-chain over the next few sessions, which would be a more structural read than a single-name valuation correction.

Almost a mirror image of Monday. June CPI fell -0.4% month-over-month — the largest single-month drop since April 2020 — on a sharp -5.7% decline in the energy index as gasoline prices eased with cooling Mideast tensions. Core CPI held closer to flat, near +2.85% year-over-year. Markets read it as exactly the scenario that had been flagged as the week's key swing factor: a soft print unwinding some of the hawkish-Fed repricing that's been pressuring long-duration growth names since mid-June.

The reaction was broad-based: semiconductor and AI-infrastructure names rallied together in the 3-4% range, a fuel-cell name extended its bounce off last week's short-seller story, and a defense-AI software name up in the high single digits pushed its own price above the average analyst 12-month target — worth flagging as a case where a rally has now outrun where analysts think the stock should trade, not a reason to panic but a genuine data point. A hardware name jumped hardest of all on reports of an expanding AI-server backlog plus a same-morning divergence trade: a large enterprise-software peer issued a revenue warning, and money rotated straight from software into server/compute hardware on the news.

On the other side, yesterday's winners cooled: energy majors and a couple of steadier defensive names gave back some ground, the year's mirror-image of Monday's rotation into those same names.

The practical read: one data point doesn't reverse a policy stance — the real test is whether this shows up in how the Fed talks at the July 28-29 meeting, not in one day's price action.

Monday looked like a mirror image of the Jul 2 rotation, playing out inside the same rate backdrop rather than on any new data — no FRED release has landed since last Friday. AI-infrastructure and high-multiple growth names sold off together: semiconductor and AI-power names each fell in the 3-5% range, and a fuel-cell name still working through last week's short-seller allegations gave up similar ground, with no new company-specific news behind the move. A newly-public space-launch name continued its own separate slide, still working through the bond-offering/valuation story covered here last weekend.

On the other side, energy majors caught a real bid — both up 3-4% as the cracked Iran ceasefire keeps a floor under oil — and mega-cap software (the Azure/cloud name that's been the market's "safer" AI bet lately) was one of the few green names in tech, up modestly even as riskier AI-infrastructure names sold off around it.

The practical read: correlated moves across a whole sector, on a day with zero new data, are a rotation, not a verdict on any single name. The real test is tomorrow.

Quick follow-up on last week's short-seller story: the fuel-cell company that took a credibility hit from a Hunterbrook Capital report alleging supply-chain misstatements filed a formal SEC 8-K on Jul 9 rebutting the claims point by point — it says current and backlog demand for the raw material in question is covered without the dependency the report alleged, and categorically rejects the financial-misconduct allegations. The stock popped nearly 7.5% premarket on the filing, then gave back part of that move into Friday's close. Worth noting the distinction: a dated SEC filing is a materially stronger response than a press-release denial, but it's still the company's word against a short-seller's — no independent confirmation or regulatory action either way yet. This one stays a "watch for follow-up reporting" situation, not a resolved one.

Looking ahead, three catalysts land before next weekend. Tuesday's June CPI print (Jul 14, 8:30am ET) is the headliner — one closely-watched nowcast puts it around 3.96% year-over-year, a touch cooler than May's 4.17%, and it's the last major inflation read before the Fed's Jul 28-29 meeting. Same morning, big-bank earnings kick off (JPM, Goldman, Citi, Bank of America, Wells Fargo), with BlackRock and Morgan Stanley following Wednesday — a real-world read on credit conditions, not just what futures markets are guessing. And the rate-hike odds priced into futures have moved further hawkish this week specifically: July FOMC hike probability now sits around 31% (up from roughly 17% a week earlier), September around 84% (up from 61%) — a sharper, more concrete repricing than the vaguer "leaning toward a hike" framing from a few weeks ago.

Separately, the Iran ceasefire cracked again over the weekend, pushing Brent back near $77/bbl and WTI near $72 — a continuing tailwind for energy names and a live upside risk for Tuesday's inflation number, two threads that are connected but often discussed separately.

The practical read: a quiet weekend with genuinely new information isn't a contradiction — the price didn't move, but the picture going into next week did, on both the company-specific (BE's rebuttal) and macro (Fed odds, CPI setup) fronts.

Every laggard in a portfolio looks the same on a screen — a red percentage next to a ticker. They are not the same underneath, and a weekend is a good time to actually check rather than assume. Five names that have been underwater for a while got the same question this week: is the decline a broken thesis, or just an unlucky entry point relative to where the market has since moved?

Gold's ETF proxy is down meaningfully from where it was bought — but that entry came within days of an all-time-high spike in late January, right before a sharp one-day crash tied to a Fed-chair nomination headline. Spot gold has been grinding back since, sitting near multi-week highs this week as the Iran/Strait-of-Hormuz tension flared again, and both HSBC's and JP Morgan's medium-term targets still imply real upside from current levels. That's a "bad entry timing, not a bad thesis" story.

The two energy majors and the defense name in this bucket tell a similar story from a different angle: all three were bought during a mid-2026 trough, and all three are now catching a real tailwind — energy from the Hormuz-driven oil spike, defense from a fresh multi-billion-dollar acquisition and a NATO allied-spending commitment. Both were actually green on the most recent session, even while sitting underwater on a cost-basis view. A mega-cap tech name in the same bucket is down nearly 30% from its own 52-week high despite genuinely strong underlying numbers (cloud growth in the 40% range, a multi-billion AI revenue run-rate) — the market's real argument there is about the timing of AI-capex payoff, not whether the business is winning, and consensus price targets still imply substantial upside.

The one name that came out of this review looking different: a newly-public space-launch company down over 20% from its post-IPO entry. Unlike the others, this one has real, specific news behind the move — a large bond offering that reignited capital-intensity concerns, a valuation still around 100x sales even after the pullback, and a slow-motion float expansion from upcoming share-lockup expirations. That's not "the market being emotional about a fine business" — it's a genuine, still-unresolved risk that deserves an actual watch level rather than being waved off as noise.

The takeaway: "down a lot" and "thesis broken" are different claims, and conflating them either way is a mistake — holding something out of inertia because "it'll come back" is just as sloppy as panic-selling something that's actually fine. The discipline is checking each one on its own terms, on a schedule, not just when it's already scary.

Three AI-adjacent names moved sharply today, and lumping them into one "AI trade" narrative would miss what's actually going on. Meta jumped roughly 6% — its best week since early 2024 — on a genuinely new piece of news: the launch of "Meta Compute," a cloud business that will sell AI computing power and models directly, putting the company in the same conversation as AWS, Azure, and Google Cloud for the first time. That landed alongside news that in-house "Iris" AI chip production starts in September, targeting real gigawatt-scale capacity next year, plus a fresh AI model release. At least one desk raised its price target meaningfully on the back of it. This is a company-specific re-rating, not a sector-wide move — a business getting structurally bigger, not just a stock catching a bid.

Bloom Energy fell about 6% on the opposite kind of story: a short-seller report from Hunterbrook Capital alleging misleading statements about BE's supply chain for scandium, a raw material used in its fuel cells, plus questions about production capacity and disclosed financials. Management has denied the allegations and says it stands behind its audited numbers. This is a credibility dispute, not a demand-side or macro question — the AI-data-center power thesis that's driven fuel-cell names higher all year hasn't changed on its own; what's newly in question is whether the company's own disclosures can be trusted at face value, which is a different and arguably more serious kind of risk until it's resolved one way or the other.

Palo Alto Networks fell nearly 4% for a third kind of reason entirely: valuation compression. High-multiple software names are the most exposed part of the market to the hawkish rate repricing covered in yesterday's note on this log — when the market prices in more rate-cut delay (or a hike), the present value of a distant growth story shrinks mechanically, independent of anything the company itself did today.

The practical read: "AI stocks were mixed today" is a true but low-information sentence. A demand-story re-rating (Meta), a credibility/fraud-adjacent allegation (Bloom Energy), and a rate-driven multiple compression (Palo Alto Networks) are three different risks that happen to be showing up in adjacent tickers on the same day. Worth tracking each on its own terms rather than reading them as one signal about "the AI trade."

The Fed's June 17 meeting held the funds rate steady at 3.50-3.75% — unsurprising, and mostly priced in already. The real news was buried in the dot plot: under new Chair Kevin Warsh, 9 of 18 committee members now project at least one hike by year-end, a reversal from the cut-leaning projections markets had been pricing through most of H1. Futures markets have since started pricing a real possibility of a hike as early as October.

This matters more than a single data point because it changes the direction of the discount-rate story that's been the backdrop for every rate-sensitive name on this log. A market pricing eventual cuts and a market pricing a possible hike arrive at very different valuations for the same long-duration growth cash flows — the 2s10s dynamic covered here back in early July was already flagging the curve moving in an unfriendly direction; this is the policy-side confirmation of why.

The practical read: sectors with pricing power, steadier demand, or a direct AI-infrastructure angle that isn't pure mega-cap software (utilities, energy, industrials) have had a stronger run in 2026's rotation than the names most exposed to a "cuts are still coming" thesis. That's not a reason to abandon the AI-buildout story — the demand signal hasn't changed — but it is a reason to be more selective about paying up for further length in names that already re-rated sharply higher since June, rather than adding indiscriminately across the whole growth complex.

What to watch: whether the "at least one hike" contingent on the committee grows or shrinks heading into the next meeting, and whether the October hike odds currently being priced actually firm up or fade as new data arrives. Either direction is a bigger swing factor for this kind of book right now than any single earnings print.

Two sessions after the Samsung-earnings/DeepSeek-headline selloff hit semiconductor and AI-adjacent names, most of that move is reversing. The same names that led the decline are leading the bounce back: strong gains across the group as the initial scare fades without any new negative catalyst showing up behind it.

This is worth reading the same way the selloff itself was read two days ago — as a headline-driven, sentiment-based swing rather than a fundamentals story in either direction. Nothing about Nvidia's supply-chain position or the underlying AI-buildout demand actually changed between Tuesday and today; what changed is how much weight the market was putting on a single competitive headline. A round-trip like this, on its own, doesn't confirm the thesis is fine any more than the selloff confirmed it was broken.

The practical read: two-day whipsaws like this are the cost of holding a sector-concentrated, headline-sensitive book. The useful signal isn't today's bounce — it's whether these names hold above their pre-selloff levels through the next earnings cycle, or whether this becomes a pattern of round-trips on every AI-competition headline.

Today's move was narrower than it looked at first glance. Samsung's quarterly results landed soft and triggered a slide in South Korea's KOSPI, and a Reuters report that China's DeepSeek is developing its own AI chip — reducing potential reliance on Nvidia — added a second, more direct hit to the semiconductor trade specifically. The Nasdaq fell about 0.6% and the S&P 500 about 0.2%, while a popular semiconductor ETF was down closer to 4.5% at the open — a much sharper move than the headline indexes suggest.

Inside the chip complex the damage was uneven: some of the most levered semiconductor and semi-equipment names fell hard, while Nvidia itself was little changed to higher on the day — a sign the market is treating this as a supply-chain and competitive story more than a blanket "AI trade is over" verdict.

The practical read: this is worth tracking as a distinct risk from the rate-driven rotations covered in earlier notes on this log. A single "here's a competing chip" headline doesn't confirm a multi-year thesis change on its own — but it's exactly the kind of story that deserves following up on with subsequent earnings calls and analyst checks, rather than dismissing as pure noise on day one.

Markets reopened today after the long holiday weekend, and the move was close to a mirror image of the Jul 2 rotation day: the names that got sold off into the break led the rally back, while the defensive names that caught a bid on Jul 2 mostly gave a little back. Bloom Energy led with a near-9% gain, Arista climbed over 8%, Tesla and AMD each rose more than 6%, and Dell added over 4%. Palo Alto Networks, Meta, Palantir, and Google all posted solid gains as well.

On the other side, the rotation-day winners cooled off modestly: Lockheed Martin and Coca-Cola each slipped around 1.4%, Healthcare (XLV) eased about 1%, and Microsoft dipped under 1%. Nothing dramatic — the pullback in the defensive names was a fraction of the size of Thursday's rally in them, so this reads as a genuine risk-on session rather than a full unwind.

The practical read: single-session rotations in either direction, on their own, aren't a reason to change conviction on anything. The Jul 2 rotation into defense/staples looked notable at the time; two trading sessions later, most of that move has already round-tripped. It's a good reminder that a one- or two-day sector swing is often just noise until it's confirmed by something more structural — an earnings print, a rate move, or a real change in the underlying business.

The last session before the July 4th break produced a genuinely odd pairing: the Dow closed at a fresh record — its fourth straight — while the Nasdaq fell for a second consecutive day. The trigger was the June jobs report, which came in soft: the economy added just 57,000 jobs against expectations for roughly 113,000, and it was the weakest print since February. Normally that would read as bad news. Instead, the Dow rallied on it.

The logic: a cooling labor market lowers the odds the Fed hikes rates later this month. Futures pricing for a July hike dropped to roughly 20%, down from about 30% the week before. Slower jobs growth, in other words, got read as "the Fed has less reason to tighten" rather than "the economy is weakening" — a distinction that matters a lot for how different parts of the market reacted. Rate-sensitive, defensive, and industrial names (the kind that make up a big chunk of the Dow) benefited; richly-valued growth names did not.

That growth weakness showed up most sharply in semiconductors, which fell for a second straight session and are now down roughly 12% over two days by one popular chip-sector index. This follows a very strong first half of the year for the group — this looks like profit-taking after a big run, not new evidence anything is broken with the underlying AI-buildout demand story. Tesla added its own "sell the news" example: shares dropped about 7% despite comfortably beating Q2 delivery estimates, a reminder that when expectations are already high, even a genuine beat isn't always enough.

Zooming out past the US: the pan-European STOXX 600 gained just under 2% on the week, while Japan was mixed — the Nikkei 225 slipped slightly, but the broader TOPIX index gained. Markets are closed Friday July 3rd (observed) and Saturday July 4th itself, reopening Monday July 6th. Two things worth watching into that reopen: whether the chip selloff stabilizes or keeps extending, and a tariff-related court case (a May ruling against a specific import surcharge, now under appeal) that's still an unresolved input into corporate cost assumptions. The new Fed Chair has also moved toward a more purely data-dependent approach and away from traditional forward guidance, which likely means more of these single-data-point swings rather than fewer.

Today was a textbook rotation session, not a broad selloff. Growth and AI-adjacent names gave back ground hard — Tesla and Dell both fell over 7%, Bloom Energy dropped more than 6%, Meta and AMD each lost 4-5%, and Arista and Nvidia slipped 2-4%. At the same time, a distinct group rallied together: Apple, Lockheed Martin, RTX, Coca-Cola, and Healthcare (XLV) all posted gains of roughly 2.5-4.7%, and gold (GLD) added another 2% on top of an already strong run.

That pairing — growth down, defense/staples/healthcare/gold up, on the same day — is the classic signature of a risk-off rotation rather than a fundamentals-driven repricing of any single name. Nothing in today's move looks like a thesis break for the names that sold off; it reads much more like profit-taking in the winners of the last few weeks (several of the down names had just posted double-digit gains recently) funding a rotation into names that hadn't moved as much.

The practical takeaway for anyone watching a similar book: a one-day rotation like this is exactly the kind of noise that's easy to overreact to. The question worth asking isn't "did my growth names go down today" — they did, across the market — it's whether anything about the underlying business actually changed. On a day like today, for most names, the honest answer is no.

Tech (XLK) led the market by a wide margin in H1 2026, up roughly 25% — but it wasn't a straight line. The sector round-tripped through a sharp February–March drawdown (AI-capex jitters, a broader growth-stock de-rating) before the recovery accelerated hard into Q2 on strong hyperscaler earnings and reaccelerating AI infrastructure spend.

The more interesting story is who else showed up. Energy (XLE, +16.6%) and Industrials (XLI, +16.4%) both nearly matched tech's pace — a reminder that "AI trade" and "old economy" aren't mutually exclusive right now. Industrials in particular has quietly been a second-order AI-infrastructure play: data centers need power, cooling, and physical buildout, and that shows up in the industrial base before it shows up in the chip numbers.

Consumer Discretionary (XLY) was the lone red sector for the half, down about 1%. Everything else — Staples, Utilities, Healthcare, Financials — posted positive but unremarkable single-digit gains, the kind of steady grind that doesn't make headlines but does the job in a diversified book.

The AI buildout narrative has quietly shifted. Through 2023–2025 the gating constraint was compute — GPU supply, fab capacity, chip yields. Heading into H2 2026, the constraint is increasingly electricity: data center power demand is projected to keep climbing sharply through 2030, and US data center capacity is set to nearly triple over the same window. Nvidia itself has started explicitly linking future AI growth projections to power availability, not chip supply.

That reframing matters for how the AI trade gets played. It's no longer just semiconductors and hyperscalers — it's the power producers and grid-adjacent names sitting behind them. Unregulated power producers with direct long-term supply agreements to hyperscalers are a cleaner, more contracted way to get AI-adjacent exposure than chasing valuation on the chip names directly. Same logic applies to grid buildout, cooling infrastructure, and anything that turns electrons into usable data-center capacity.

The risk to watch: this is now a crowded, well-understood thesis. The "power is the bottleneck" trade has been running for a while, and the names most levered to it can move sharply on single-day news — a lost contract, a competing technology announcement, insider selling — even when the multi-year thesis hasn't changed at all.

The 2-Year Treasury yield has climbed faster than the 10-Year through H1 2026 — a flattening move in the yield curve. That's a market signal about the *path* of rates, not the level: the 2-year reflects where traders expect the Fed to be over the next couple of years, while the effective Fed funds rate itself has barely moved. When the 2-year rises while the policy rate sits still, it means the market is pricing in less near-term easing than it previously expected — not that policy has tightened.

See the chart in the Econ Briefing tab for the actual data. The practical read for equities: long-duration, high-multiple growth names are the most exposed to this kind of repricing, since their valuations lean more heavily on cash flows far in the future getting discounted at whatever rate the market settles on. It's a slower, quieter risk than a Fed rate hike — but it shows up in exactly the kind of pullback tech saw earlier this year before the H1 rally resumed.

A few ground rules that shape every move on this log, independent of any single ticker's story:

Conviction sets position size, not price. A brand-new idea starts as a small starter position — enough to track it seriously, not enough to hurt if the thesis is wrong. It only grows into a core holding after the thesis has actually played out for a while, not because the stock went up fast.

A thesis break is a sell signal; a price drop alone isn't. The BE trim and the PANW/AMD/GLD profit-taking in the log are both examples: sizing changed because something about the setup changed (a lost contract, a valuation that ran ahead of itself), not because the position was simply down or simply up. Noise gets held through; a broken thesis gets acted on.

Diversify across what actually drives returns, not just ticker count. Ten AI-adjacent tickers isn't diversification if they all move on the same headline. The sector-gap fills on this log (Healthcare, Financials, Energy, Defense, Utilities) exist specifically to reduce correlation to the AI trade, not to check a box.

Earnings are a hold decision, not a trade decision, unless something's actually broken. Positions built on a multi-quarter thesis get held through the print that thesis depends on — MSFT and META both get held through their July earnings for exactly this reason. Exiting right before the event that's supposed to prove the thesis defeats the point of having one.

The 10-year Treasury yield printed 5.24% this week (Oct 9) — the highest level tracked on this page, up 29 basis points over the past month and 62 over the past quarter. The 2-year followed it higher, now 4.80% (+24bp over the month). The Fed funds rate itself hasn't moved: it's still at 3.88%, where it's sat since the September hike. This move is the market repricing the long end on its own, not a new Fed action.

Unemployment ticked up to 4.20% (from 4.10% a month ago), and the CPI trend is running close to 4% year-over-year. Neither print is dramatic alone, but together they read as a labor market cooling slowly rather than cracking, with inflation still above target — a combination that argues for the Fed holding rather than cutting soon, which keeps pressure on the long end of the curve.

The practical read for equities: long-duration, high-multiple growth names are the most exposed to a move like this, since more of their value sits in cash flows far in the future that get discounted harder as yields rise. That's showing up directly today in a broad pullback led by growth/tech names, while more rate-tolerant sectors (financials, energy) are holding up better.

What to watch: the next CPI print, and whether the 10-year's climb above 5.2% is a one-off adjustment or the start of a sustained move. A break above roughly 5.25–5.30% would mark fresh territory for this cycle.