AI capex hits $520bn. The rotation just got real.

Transcript

Tom Buddy, if you thought AI capex was big before, wait till you see South Korea dropping five hundred and twenty billion on chip plants. The rotation just got real.

Marie It's the London Edition, Tuesday June thirtieth. I'm Marie, with Tom and Gerald. And Tom is already bouncing off the walls over semis.

Gerald Tom, wasn't it yesterday you were telling us to buy the semiconductor ETF? You're going to break your hand patting yourself on the back.

Tom For real! That call is up—what—another three percent? But here's the thing: SMH is up sixty-nine percent year-to-date, and it's still six percent below its fifty-two-week high. There's room to run, buddy.

Gerald Yeah, look, sixty-nine percent year-to-date and trading near highs is not usually where I start buying. But I get it, the capex numbers are nuts. Five hundred twenty billion from Samsung and SK Hynix—that's not loose change.

Marie And this is what I mean about the rotation being structural, not just a blip. The Magnificent Seven shed two point three trillion. Money is literally moving from software and platforms into hardware and chips. This is a capex super-cycle, not a tech rally.

Tom Exactly.

Gerald One hundred percent.

Marie That's the whole story.

Tom Nvidia at fifteen point three times forward earnings? That's cheaper than a utility! And they're the picks and shovels of this buildout.

Gerald Hold on, Tom. Nvidia at fifteen times earnings is still a chip stock, not a bond. If this capex boom hits a financing wall—and rates are still high—those multiples can compress. I'm not saying it's a short, but 'cheap' is doing a lot of work there.

Marie Gerald, I'm going to push back. The financing question is exactly what's missing from the coverage today. No one's asking if five hundred twenty billion of chip plants can earn a decent return when the cost of capital is where it is. That silence is loud.

Tom Look at Marie, finding the bear case in a capex boom. But fair point. Still, the market is voting with its feet: semis over software. And analysts revising price targets for semis after a sixty-nine percent rally? That's basically a retirement plan.

Gerald Oh, that's brutal.

Marie I mean… he's not wrong.

Gerald Right. And the market's voted Rocket Lab up fifteen point nine percent last session on its eight billion dollar Iridium deal. That's a lot of enthusiasm for a 'shortcut' to competing with Starlink.

Tom No way, Gerald—Rocket Lab is now a legit satellite operator! They've got spectrum, a constellation, and a launch business. They're the anti-SpaceX play. And Iridium jumped twenty-five percent! The deal premium is priced in, but the strategic value? That's the real story.

Marie Tom, you sound like you're ready to buy shares from the launchpad. But hold on—integration risk on an eight billion dollar deal is huge. And Rocket Lab already surged. I'd watch it, not chase it.

Gerald To be fair, Tom, at least with semis you've got a clear capex driver. Rocket Lab is more of a narrative stock right now. But hey, space is exciting. I'll give you that.

Tom Ha — fair enough! But now, talk about a value unlock: Comcast spinning off NBCUniversal and Sky? The stock surged twenty percent on the news and added four point five percent last session. It's still down twelve percent year-to-date and trading at point nine eight times book. That's a sum-of-parts re-rating waiting to happen.

Gerald Finally, a value story I can get behind. Comcast is basically giving away the media assets for free at that price-to-book. Even after the pop, there's meat on the bone. I like it.

Marie And it's a bet on connectivity over content. That structural separation makes sense—the market is rewarding focus. But execution risk on a split this size? Not zero.

Tom Execution risk, shmexecution risk. At point nine eight times book, you're paid to wait. And speaking of cheap, Gerald, British American Tobacco is cutting nine thousand jobs—that's a fifth of the workforce. The stock is flat but up over three percent this week. Eleven point nine times forward earnings for a company getting leaner?

Gerald BAT is my kind of ugly. Big job cuts usually mean margin expansion is coming. And it's a UK company, so I've been watching it. The market's starting to price that, but at under twelve times earnings, it's not fully valued for the turnaround. Even with all the regulatory smoke.

Marie Regulatory smoke—pun intended? But seriously, tobacco is always under a cloud. Still, nine thousand jobs is a massive cost-cutting signal. I can see the bull case.

Tom You two are actually agreeing on a value stock? The world is ending. But wait—crypto signals are all over the place today. CoinDesk says the record thirty-four and a half billion net long dollar and near three million SOFR shorts could unwind and boost bitcoin. But the same outlet says Korea's five hundred eighteen billion AI chip plan shows crypto losing the capital race. And the UK is easing crypto rules? I'm getting whiplash.

Marie That CoinDesk take—the one about AI winning the capital war over crypto—that's the most original thing I've read today. It's not about regulation or ETFs; it's about industrial-scale investment leaving digital assets behind. The sheer size of the chip commitment screams that institutional money is fleeing to physical AI infrastructure.

Gerald Right, and bitcoin's down thirty-three percent year-to-date, fifty-three percent below its high. The crowded dollar trade could reverse, sure, but if that unwind happens because of a dovish Fed, it's going to crush the long-dollar crowd and maybe send bonds soaring. And that's what scares me about the whole 'buy semis, sell bonds' consensus.

Tom But Gerald, that's the trade! Long SMH versus short TLT captures the industrial boom and the rates-for-longer bet. Yeah, there's a record short in bonds, so a squeeze is possible. But until we get that dovish pivot, it's working.

Marie Hold on—that's exactly the ticking clock. Positioning, not fundamentals, is the day's biggest risk. If a data print triggers a bond rally, the long-duration Treasury ETF is near its fifty-two-week low. It would rip higher, and momentum in SMH and Nvidia would probably get hit.

Gerald And that's before we even talk about mortgage bonds. The Wall Street Journal is warning that less Fed guidance means higher volatility premiums, which could hit mortgage rates. The MBS ETF is flat year-to-date, but if spreads widen, it's a sell.

Tom Okay, okay, I hear the bond doom-loop returning. But back to the bright side: Airbus just got a three point four billion dollar loan from the European Investment Bank for tech sovereignty. Defense and commercial aviation tailwinds! Stock is down near eight percent year-to-date, around twenty-three times forward earnings. That's a recovery play.

Gerald Airbus is an interesting one. That EIB loan is a stamp of approval for their tech push. And defense spending isn't going away. At twenty-three times, it's not dirt cheap, but the growth outlook is solid. I could see it as a buy.

Marie And here's a deep value one: Prosus plans a five billion dollar buyback to reduce reliance on Tencent. The stock is down thirty percent year-to-date, at one point five times forward earnings! That's absurdly cheap if the discount to Tencent narrows.

Gerald One point five times!

Tom Yeah, that's – that's crazy.

Marie Nailed it. Sometimes the ugliest stocks are the best.

Tom But Tencent itself is down thirty-three percent year-to-date. If Prosus is reducing reliance, it could pressure Tencent further. I'd hold off on Chinese tech until the regulatory fog clears.

Gerald Tom, that's the THIRD time this month you've been scared of a value play because of macro. At one point five times earnings, you're not paying for any recovery. I'll take that bet.

Marie Same here. Sometimes the trade is just too cheap to ignore.

Gerald Alright, shifting to a messy one: Crédit Agricole and other car finance lenders are challenging the UK compensation scheme on human rights grounds. That's an unusual legal headache. The stock trades at point six seven times book, but with this overhang, it's a sell for me.

Marie It's a structural risk. If they win, it sets a precedent that could upend consumer protection. If they lose, the compensation bill stands. Either way, uncertainty kills bank stocks. So sell.

Tom Agreed. Even at point six seven times book, this is one to avoid until the legal fog clears.

Gerald And as always, none of this is investment advice. Just three friends talking markets.

Marie To wrap: the day screams AI capex as the gravitational center, sucking flows from everything else. The cleanest expression is long SMH versus short TLT. But the counterargument is real: the Magnificent Seven are funding this chip demand, and if they keep selling off, the circularity snaps back. Plus that record short bond position is a powder keg.

Tom Fair points, but the price action is telling us to ride the wave. SMH is still below its highs, and the money is rotating. That's a trade until it isn't.

Marie Tom, you said that about semis in Q1, and you were right. But I remember you also said it about crypto three times last year.

Tom Ha! Fair. But this time it's different—it's real steel and silicon, not tokens.

Gerald Right, because semiconductor fabs have never been cyclical...

Tom If you're just finding us, hit follow on Spotify—or check investmentflash.com for the full digest with charts and sources.

Marie And we're back later today at nine a.m. New York time for the New York Edition. Don't miss it.