1An escalating war is bidding oil, not gold
US strikes on Iran ran to a sixth consecutive night, with Iranian retaliation, falling transit counts through the Strait of Hormuz and a seized tanker off Yemen alongside. The deepest contract on the question agrees that this is escalating: Polymarket's market on a US invasion of Iran before 2027, with $43.9m of lifetime volume and $495k traded in 24 hours, rose two points to 25.5%. Its companion — whether Hormuz traffic returns to normal by 31 December — fell four points to 51.5% on $100k of daily volume, which is barely above a coin flip and moving away from a temporary-disruption read.
Crude booked a weekly gain on it. Gold did the opposite, heading into the weekly close on course for its largest weekly fall in six, and the reason sits in a third contract. Polymarket prices a Fed rate hike at some point in 2026 at 53.5%, up two points on the day and better than even. Daily volume there was thin at $24k against $4.2m lifetime, so the tick is weak evidence and the level is the signal. Higher policy-rate expectations lift real rates, and real rates are the thing gold has to compete with — this week the rate-hike channel beat the safe-haven bid, so an escalating war showed up in crude rather than in the metal.
July is not where that gets settled. The 29 July decision prices no change at 95.05%, on $424k–$639k of 24-hour volume per leg, with the hike leg at 4.65% — up from 3.85% a day earlier — and a cut at 0.35%. Near-term the Fed is a settled hold; the hike is a year-end tail. The day's hard data did nothing to argue that tail away: housing starts printed 1,427k for June against 1,199k in May, a 19.0% jump in a single month and the standout print of the day, while industrial production was effectively flat at +0.08% and capacity utilisation unchanged at 76.1%. The White House framing pointed the same direction on energy, with a presidential post claiming oil "is flowing like never before, thanks to the awesome Power of the United States Military."
An economy that is not slowing, with an oil shock building on top of it, is the configuration in which a central bank's next move is up rather than down. That is the pricing, and it is why the safe-haven trade did not work this week.
2A 77% profit jump was no protection in the semiconductor wash-out
Taiwan Semiconductor reported second-quarter profit up 77% and announced further US expansion, and was sold with its sector anyway. The rest of the tape shows what it was sold into: the memory rout carried over from the previous session and deepened, down 6% and down 4% pre-market with Asian weakness leading, alongside them, and the chip-design software names went too — off 9.3%, off 8.4%.
The earnings reactions elsewhere were just as sharp. fell 11–12% pre-market on mixed results and guidance concerns, dropped 13% on its print, and extended a post-IPO decline to a sixth straight losing session and new lows. The pattern across those moves is a tape marking down forward expectations rather than trailing ones: the strongest reported number of the day belonged to a stock that fell with everything around it, which means a good print is currently not a defence.
3The AI bubble argument moved from the multiple to the margin
Jonathan Thomas, chief executive of American Century Investments, speaking on The Compound and Friends, made the case that cannot be a bubble in the ordinary sense. Its multiple has compressed to roughly 18 times, below the S&P's own, and the stock has stopped responding to beats and upgrades — his phrase is that it trades "as though already disrupted." A bubble expands a multiple; this one contracted. The overhang he names is competitive rather than valuation-driven: custom ASICs and hyperscaler in-house silicon chipping away at a roughly 90% share.
What he says the bears have moved to instead is an earnings bubble — a worry about the denominator's opposite number, the E rather than the P. The concern is that companies are over-earning against a capex figure that has not been normalised, and the mechanism he points at is circular financing: Nvidia backstopping neocloud purchases of its own GPUs, so funding commitments could "vanish like vapor" when the cycle turns. He frames the industry as five layers — power and cooling, chips, data-centre infrastructure, models, then applications — with the capital-intensive layers at the bottom and value migrating upward, so which layer a position sits in determines whether it is defensible.
The earnings picture underneath is unusually broad. S&P 500 earnings are expected up 23.2% year on year this quarter and, stripping out the Mag 7, still up 20.9%, with 10 of 11 sectors growing and technology up 65%. On year-to-date point contribution the S&P 493 is, in his words, "absolutely trouncing" the Mag 7, and more than half the index is up over 10%. His rotation follows from that: the adopters rather than the creators, healthcare and financials on underwriting and drug discovery, with cited as a surprise data-centre-buildout beneficiary and , and used as illustrations along the way.
Two things temper it. Thomas runs an asset manager and spends most of the episode on his own industry — the active-ETF boom, Avantis's growth, 's 20% year-on-year gain in assets — so the market view arrives inside a book-talk. And no bear scenario was put to him; the episode is constructive throughout, and none of the names in it were presented as recommendations. If the bubble really is in earnings rather than in price, the thing that dates it is capex normalisation and the financing behind it, not the multiple.
What the sources recommended
Only two single-name calls came out of the day's videos, both buys, each from one channel with no corroboration anywhere else in the material, and both inside videos that are selling something alongside the pitch.
(Norfolk Southern), from Felix on Felix & Friends, who says he bought it himself the day before. The thesis is the railroad as a toll booth — a second coast-to-coast network cannot realistically be built, and the operating ratio is grinding lower — but the load-bearing part is a proposed $85bn merger with Union Pacific to create the first true US transcontinental railroad, in which holders would receive stock in the combined company plus cash. That catalyst is not done: it needs Surface Transportation Board approval, targets a 2027 close, and faces opposition from large shippers, unions and politicians. He discloses the 2023 East Palestine derailment as a live overhang on the safety franchise, and his "institutions are accumulating" evidence is one day of volume — about 1.4m shares against a normal 700k — not confirmed holdings data. His own instructions are to build slowly and size it as a cyclical rather than a lottery ticket. The video also funnels to a free training site, though he takes no sponsors and says his purchase was a few hundred shares.
(Churchill Capital Corp 11), from Ross Givens, is the SPAC merging with Agility Robotics, whose two-legged warehouse robot Digit is already doing paid work for Amazon — also an investor. The deal values Agility at about $2.5bn pre-money against Figure AI at roughly $39bn, brings around $620m of cash ($420m trust plus a $200m PIPE), and the PIPE was taken by Nvidia, Amazon, SoftBank, Schaeffler and Foxconn at $10 a share. His instruction is to buy as close to that $10 trust price as possible; it trades near $15, having spiked to $19 and back within four days. The caveat comes from him unprompted: more than 90% of SPAC-merger companies end up below the $10 trust price, the company is unprofitable, and he frames it as a fun trade rather than a core holding. The stated edge is scarcity — the only listed pure-play in humanoids, with 's Optimus buried inside a $1trn company and Figure private. Two paid-membership upsell segments sit in the video, and he does not say whether he owns it.
Neither buy call gets tested this year; the rate tail gets tested every day
The 29 July decision is priced as a hold at 95.05%, so the meeting itself settles nothing. The numbers to watch are the ones inside and beyond it: the July hike leg, which moved from 3.85% to 4.65% in a day, and the 53.5% contract on a hike landing somewhere in 2026. What actually moves those is the oil path and the inflation path, and the Hormuz normalisation contract at 51.5% is the cleanest daily read on the first. Gold's weekly close is the second test — if the metal keeps falling through an escalating war, the rate channel is still the dominant one.
The two buy calls run on a different clock entirely. The railroad merger needs regulatory approval and targets 2027; the SPAC deal is not expected to complete until the fourth quarter, when the ticker becomes . Both are marked to market on approvals rather than on prints, which means neither gets confirmed or broken by anything in the next few weeks.
