1The AI build-out is being paid for out of enterprise software budgets
fell 25% on 14 July, the worst single day in its 115-year history, and the reason attached to it is the most consequential claim of the day. Felix, on Felix & Friends, calls it "capex reprioritization": customers are not cutting their technology budgets, they are redirecting them — out of enterprise software licences and into AI compute and GPUs. If that is what happened, the print is not an execution problem at one company, it is a transfer of demand, and the vendors on the losing side of it have not reported yet.
The sympathy move is the supporting evidence. , (Salesforce) and (ServiceNow) were dragged down on the same read-through, repricing names that had not missed anything themselves. The Compound's Animal Spirits 473 came at it from the tape rather than the thesis, reacting to down 23% pre-market alongside the morning's bank earnings, and Prof G Markets logged the same 25% close. Three channels, one event, and only one of them supplies a mechanism.
Felix's wider argument is about how the build-out is funded rather than how it is bought. He describes the circular loop — a chipmaker funds a startup, the startup buys its chips, the sale books as revenue — and puts $182bn of new AI debt at roughly 20% of the US debt market, with banks now offloading that debt, citing the FT. is where he flags that circularity; he attaches no price to it.
The chip tape split the same day. rose 3% pre-market on a 30% capacity expansion, while memory sold off hard — down 7.9%, SK Hynix down 6 to 8%. Equipment up and memory down is not the same story as software budgets moving to GPUs, and nothing in the day's material joins the two; it is worth noting only because the halves of the semiconductor complex stopped moving together.
What the tape is not doing is confirming a broad unwind. The Compound notes the S&P 500 and its equal-weighted version sitting at all-time highs together, so the damage is name-specific rather than index-wide, and Polymarket's contract on a US recession by the end of 2026 prices 10% — on roughly $3,000 of 24-hour volume, which is too thin to lean on in either direction. The consequence, if the budget-shift read is right, is that the AI trade and the enterprise software trade stop being one trade: the first is being funded out of the second.
2A record fee quarter and a 500-times-sales buy rating came out of the same deal
The -SpaceX listing paid out about $500m in fees, and Prof G Markets sets that inside a broader banking boom — investment-banking fees up 30% to 55% year on year, equities trading up 45% to 90%. rose 9% on profits up 80% year on year, its strongest quarter on record. printed record revenues and rose 2%, beat and rose 2%, beat and still fell 2% on deposit costs, and fell 5%.
Felix's episode, titled "Wall St Is Pumping AI Stocks — So Why Are They Falling", points at the other end of the same transaction. -SpaceX is about 38% off its high despite a fast-track inclusion in the Nasdaq-100, and 18 of the 19 analysts covering it rate it a buy, with targets running to $800 — around 500 times sales. His objection is not the multiple, it is who wrote it: the banks collecting the listing and financing fees are the ones publishing the targets, and the SEC terminated the Global Research Analyst Settlement about seven months ago, which was the rule separating research from banking.
Ross Givens arrives at the same conclusion on his own numbers, bearish -SpaceX at roughly $135 on about 85 times sales, with a 20% unlock in August and rolling tranches behind it, and a target of $80 to $100. The two accounts describe arithmetically the same stock — $800 at 500 times sales and $135 at 85 times sales imply the same multiple — so the distance between them is not a disagreement about valuation method. It is the entire gap between what 18 covering analysts and these two channels think one revenue line is worth.
The consequence is that the best banking quarter on the tape and the loudest bear case of the day run on the same cash flow, read from either end. Nothing in the day's material settles which end is right. The unlock schedule is the part of it that carries a date.
3June's disinflation came from energy prices that have since reversed
June CPI cooled to 3.5% annual from 4.2%, with consumer prices down about 0.4% month on month — the largest monthly drop since April 2020 — and Treasury yields fell on the print. Prof G Markets is explicit about where the drop came from: lower energy prices following the US–Iran ceasefire. That ceasefire has since been declared over, the US launched a fourth consecutive day of strikes, and Brent is back to about $85. The input that produced the good number has been withdrawn, and the July print is the first that has to manage without it.
Mark Zandi of Moody's, on the same episode, does not read 3.5% as progress. He calls inflation "uncomfortably high" and sticky, the highest in the G7, with AI build-out demand and immigration restriction both adding to it, against a labour market where hiring is confined to healthcare and where unemployment would be 5% if participation had held. That pairing is stagflation, and his conclusion from it is that the Fed holds.
The oil read inside the day's own material does not reconcile. Prof G has Brent back at $85 on renewed strikes, while the news feed and the President's Truth Social posts describe crude sinking and oil "flowing like never before" — supply winning in the near term even as the fighting runs into a fourth day. One of those descriptions is lagging the other. The July energy component settles it, not the argument. Polymarket prices a US invasion of Iran before 2027 at 18% on about $385,000 of volume, which caps the escalation tail without saying anything about the price of a barrel this month.
4The rate argument is now hold-versus-hike, and the deepest contract on the board prices the hold at 95%
Kevin Warsh, the new Fed chair, answered the CPI drop with "CPI drop does not mean mission accomplished", and Prof G Markets reports the bond market has moved to pricing two quarter-point hikes, with the 10-year at 4.62% and the curve steepening. That is a change in the question being asked — not how quickly the Fed eases, but whether it tightens again.
Polymarket has not followed. The July meeting prices no change at 95%, a 25bp hike at 4% and any cut near zero, across roughly $4.6m of liquidity — the deepest board checked that day, and about as settled as these contracts get. Further out, zero cuts in 2026 prices at 81% and one cut at 14%, but that contract carries only about $10,000 a leg, so it corroborates higher-for-longer at a size one determined trader could move. Zero cuts is not two hikes, and the gap between what the bond market is pricing and what the contract is pricing is where the surprise sits.
How far the frame has travelled shows up in the day's most speculative segment. The Compound's Animal Spirits 473 spent time on "Will the Fed buy stocks?" — a Fed equity-purchase programme in the next downturn, tied to the Warsh era, treated as a plausible hypothetical rather than a call. Alongside it they note the absence of a recession in 17 years and Deutsche Bank's stretched CAPE reading. The July decision itself is priced to carry almost no information; the language around it is the whole event.
What the sources recommended
Five names carried a direction, three up and two down, and only one of the four videos produced an explicit single-name buy — from the channel that sells a trading course alongside it.
The buys. (Destiny Tech 100) and are one paired trade from Ross Givens rather than two ideas. is a listed fund holding private companies — 18% Anthropic, 14% SpaceX — trading near $25 against a claimed net asset value of about $38, a 32% discount he expects to close on demand ahead of an Anthropic listing he places in late 2026 or 2027. Because 14% of the fund is -SpaceX, which he is separately bearish on, he pairs it with , a 2× inverse SpaceX ETF, at $1 for every $10 of , to strip that exposure back out. The caveats travel with the call: a single source, a 2.5% management fee on the fund, and a video that runs a paid-course upsell throughout. is not a channel call at all — it is a news relay of Goldman Sachs "doubling down" on the name.
The sells. and -SpaceX are the two down-calls, and the entry is an aggregation rather than three separate calls: Felix's caution was implied rather than stated, The Compound's segment was a discussion format that issued no ticker call, and Prof G Markets reported the print. -SpaceX is the more genuinely doubled-up one — Felix on the unlock overhang and the conflicted targets, Ross Givens on 85 times sales and the August tranche, reached from different starting points.
The August unlock is a dated test; the July meeting is not
The July Fed decision is priced at 95% no change across roughly $4.6m, so the decision itself carries almost nothing — what carries information is whether Warsh's language matches the two hikes the bond market has moved to. The -SpaceX unlock is the opposite kind of event: 20% of the float in August with rolling tranches behind it, which dates the overhang argument two channels made today. The July CPI energy component tests the third claim — with the ceasefire over and Brent back near $85, the largest monthly price drop since April 2020 would have to repeat without the thing that caused it. And one policy event carries a date without an obvious instrument behind it: the President posted that "fat drug prices" are coming down "next week", a pricing action aimed at an industry rather than at a named company.
Two moves sit outside all of it. rose 16% on acquisition rumours, and and both caught a bid on China approving Apple's AI features. Neither attaches to any of the arguments above.
