1The Fed held rates and the long end sold off anyway
The July meeting ended with no move and three dissents wanting a hike, and the bond market's reaction ran the wrong way round. Front-end and belly yields fell, as a hold implies. The 30-year went the other way, jumping to 5.21% during the press conference itself — its highest since mid-2007. The 10-year-minus-2-year spread widened 10 basis points in that single session, from 0.35 to 0.45, and the Dow fell 1,100 points, its worst day since April 2025. A curve that steepens because the long end is selling into a decision not to tighten is not pricing a different rate path. It is pricing a different Fed.
Joseph Wang, on Fed Guy, points at the exchange he thinks did the damage. Chair Warsh confirmed he targets 2% but said he watches "a broader set of inflation data than PCE", and declined to name which series. Wang's read is that a chair with an undisclosed inflation measure raises the risk premium on the long bond directly, and from there into everything priced off it. He had expected a hike himself, on the reasoning that waiting only pushes hikes closer to the midterms.
Mike Gapen of Morgan Stanley, speaking on Prof G Markets the same day, reads the identical tape as a dovish hold. The front end fell, breakevens rose and the dollar weakened, which he takes as the market testing the Fed rather than losing faith in it. Both men are looking at the same four prices; one calls it a credibility repricing, the other an easing trade. What separates them is whether the long end keeps selling while the front end rallies — if the curve keeps steepening from the back, Wang's version is the one getting confirmed.
Contract pricing moved the way the language did not. Polymarket's contract on a Fed hike at any point in 2026 fell 11 points in a single day to 65.5%, on $6.0m of volume with $900k traded in 24 hours, and is down 5 points on the week. The September contract prices a 25bp hike at 52.5%, down 2 on the day, against no change at 44.5%, up 6 — on $8.2m of volume and $1.2m of liquidity, the deeper of the two Fed contracts checked. September is close to a coin flip, and the day's drift was toward no move. A hawkish statement that moves hike odds down is a statement the pricing is discounting.
Cameron Dawson of NewEdge and Dave Nadig, recording on Excess Returns before the decision with an implied hike probability near 36%, make the structural version of the point: near coin-flip odds going into a meeting is itself the story, and a direct consequence of removing forward guidance. Their consequence is worth holding on to. Take away both the guidance and a stated reaction function, and the volatility has to go somewhere — it lands in the front end of the curve, which is where floating-rate loans and private credit reset.
2Inflation's composition flipped — core halved while energy did the work
June core PCE rose 0.132% on the month, an annualised 1.60%, against 0.332% in May. The monthly core pace halved. The year-on-year rate is still 3.29%, so the level has not fixed itself, but the monthly impulse has: core CPI's monthly print was near zero, which Wang describes as one of the softest on record, and the Fed still called inflation a problem.
The pressure is in the price data outside core. Q2 real GDP came in at +1.50% annualised, down from +2.09% in Q1, while nominal GDP ran at +7.87%. The gap between the two implies a deflator near 6.3% annualised — a figure derived from the two published series rather than lifted from one, so treat it as indicative rather than official. Something outside core is doing that work, and the Fed's own account names it: tariffs, the AI build-out and energy, of which energy and AI are both worsening rather than fading. The labour market is not the constraint here — unemployment fell to 4.2% in June from 4.3%.
Brent back above $90 on the Iran war was the most-cited macro theme of the day, appearing on three of the seven channels that published and in the news feed as well. That makes the ceasefire contracts an inflation forecast as much as a geopolitical one. Polymarket prices an effective US–Iran ceasefire by 31 July at 22%, by 14 August at 46.5% (up 5 on the day) and by 31 August at 54.5% (up 2.5), on $8.1m of volume. A US invasion of Iran before 2027 sits at 24.5% on $50.7m — the deepest contract checked, and priced as a tail rather than a base case. So the priced lean is toward the shooting stopping inside a month, but not sooner.
Gapen's forecast rides on exactly that: inflation around 3.3% by year-end with the Fed on hold for the rest of 2026, and his named risk case is oil's second-round effects into airfares and food. If a ceasefire prints, the energy contribution decays and his number is reachable. If the strikes continue, headline inflation keeps running hot while core does not — which is the hardest configuration to read for a central bank that has stopped saying which inflation series it watches.
3The AI trade started sorting mega-caps by who can still fund the capex
Three mega-cap results landed inside the window, and the price reactions treated them as three different businesses rather than one theme.
grew revenue 18% and profits 32%, with Azure accelerating to 43% and guided to about 45%, $20bn of free cash flow in the quarter, and positive free cash flow guided for next year. Gil Luria of DA Davidson, on Prof G Markets, calls it a narrative-breaking result — and the detail he says did the breaking is a guidance word rather than a growth number. Capex was guided merely "up", not "significantly up".
grew revenue 28% and lost 13% of its profit, with costs up 55% and operating margin falling from 43% to 31%, guided below. Luria's complaint is that no monetisation plan came with the spending. The stock fell 11% after hours.
went free-cash-flow negative and signalled that it intends to stay there, which Luria called jarring. That sets up his explicit mean-reversion call: a year ago Microsoft traded at 30x and Alphabet at 18x, and today it is Microsoft at 18x and Alphabet at 30x. He expects that to revert.
Hold one thing against his own bullishness. Luria is constructive on the AI build-out in general while separately naming circular financing and off-balance-sheet vehicles as what worries him — two positions that are awkward to hold at once, and he does not reconcile them in the episode. Felix, on Felix & Friends, presses the same point harder, describing -style vendor financing as a circular money trap: the chipmaker funds the customers who then buy its chips. Three of the seven channels with material raised AI-capex funding as a live question, which is a different worry from whether the demand is real.
The sorting variable, then, is no longer how much a company spends on AI but whether free cash flow survives the spending. That is a question earnings season can actually settle, one print at a time — and it matters more than usual because of how narrow the index has become: Felix notes the top 10 names are roughly 40% of the S&P 500, against about 25% in 2000.
4The humanoid-robot supply chain has re-rated before the robots have revenue
Roughly two-thirds of a humanoid robot's cost sits in movement — joints, hands, actuators — and only about 6% in the AI brain. That is the arithmetic behind Business with Brian's thesis: own the bill of materials rather than guess the winning robot brand. The chokepoint is magnets, where China controls about 90% of processing, and building a unit without Chinese parts takes cost from roughly $46k to roughly $131k, close to a tripling.
The trouble is that the parts have already been priced. has re-rated to an all-time high on less than 1% of revenue coming from robots. , whose content per unit scales about 30x — from around $5 in a robot vacuum to around $150 in a humanoid — trades at the richest valuation in its history, and its first robot design wins are with Chinese OEMs. , the only US-listed maker of both strain-wave and cycloidal reducers, sits near the top of its 10-year range, which is why he keeps it on the watch list rather than in the basket. makes world-class actuators for jets and spacecraft with essentially no robot revenue at all. He also flags roller screws and linear guides as the thinnest gap in the US supply chain, where the strongest names are OTC-traded or foreign-listed.
is the one name he calls still reasonably priced, and the arithmetic is specific: the 49x trailing multiple is an artefact of the Altra write-down, next year's profit puts it near 15x, PEG is close to 1, and free cash flow is $893m. is the other structural case — pre-earnings and burning cash, but with the Department of Defense as its largest shareholder on a 10-year floor price and full offtake, and Apple signed for about $500m of US magnets.
Two caveats travel with all of this. He makes the first himself: the demand is entirely forward-looking, and several of these names have re-rated on humanoid hope while booking near-zero humanoid revenue today. The second is scope — this is one channel's thesis, argued at length, and nothing else in the day's material speaks to it either way. He also discloses a paid tool sponsorship mid-video, which is a product deal rather than a stock promotion. The consequence is that the risk in the robot supply chain has migrated into the multiple: the thesis can be right and the shares still expensive.
What the sources recommended
Five buys and two sells came out of seven videos, and three of the five buys are a single cybersecurity trade from one channel, expressed through three funds — with a disclosure attached to it.
The buys. , from Luria on Prof G Markets, on the quarter, the Azure acceleration and the free-cash-flow guide, at high conviction over months. , from Business with Brian, the only robotics name in his basket he does not describe as expensive. Then , and from Equity Mates, framed as a decade-long thematic rather than a dated catalyst: they size the cybersecurity market at $340bn in 2026 growing to about $495bn by 2035, roughly 11.5% a year, and prefer funds to single names on the argument that cyber spend is diffuse — unlike semiconductors, where captured it. They flag at 138x forward earnings while unprofitable, and at 237x, as too expensive to own individually. Two things attach to that call and both belong with it. Equity Mates Media disclosed, in the closing disclaimer only, that it is an entity within the BetaShares Group — is BetaShares' own fund and was named first of the three. And one of the hosts said he is not currently invested in it. Their own named risk is the hyperscalers bundling security as a free feature and hollowing out the point-solution vendors, Microsoft first among them.
The sells. , from Ross Givens, is a timing call rather than a short — he says he is bullish the business long term and is naming an entry, not a position. The mechanism is supply. 6 August unlocks 911m insider shares against a 639m tradable float, 1.4 times the float in one morning, with a staircase releasing about 7% every two to three weeks into late October, 28% more after the November Q3 report, and everything free on 8 December. He leans to the high end of the 15–40% early-selling range that lockup research suggests, because the venture holders are past stated fund life and owe their investors. 196m shares are already short, 31% of the float. His Monte Carlo median low is $60, he says he will not buy above it, and he is targeting early November. Two things to weigh against the precision: the specific entry levels sit behind a paid subscription he promotes in the video, and the stock has about three months of trading history to model from. , from Luria, rests on the negative free cash flow and the multiple reversion described above.
The day's most urgent-sounding video withholds its pick. Felix names no ticker on camera and holds the actual buy for a free-to-attend, no-replay seminar, so his and references are context rather than calls.
Amazon's capex line is the next test of the sorting
Amazon is the next mega-cap checkpoint after Microsoft's beat and Meta's miss, and the line that matters for the third section is capex guidance measured against free cash flow, not the revenue headline. The Bank of England decides Bank Rate at 07:00 ET, with a hold at 3.75% expected and hike pricing already trimmed after the bank flagged limited spillover from Iran. German July inflation rose while Q2 growth beat, which supports the same on-hold read for the ECB.
The energy leg gets tested continuously rather than on a schedule. Drone strikes hit two gas vessels at Egypt's Damietta port and near the Suez Canal, and the Houthis are reported to be attacking Saudi Arabia from Iraq — each such headline feeds directly into the deflator problem in the second section, while reported rare progress in Gaza ceasefire talks runs the other way. The lockup call has its own clock: first-ever earnings on 4 August, then the 6 August unlock, which is the first hard test of whether supply mechanics beat a retail bid.
