XX

Intel up 11% on Google's 3M-TPU foundry order. The credibility win for 18A is real, but read the fine print: revenue lands in 2028, leading-edge fab economics only work at scale, and Intel still needs two or three more anchor clients to make foundry profitable. I treat this as a re-rating catalyst, not a fundamentals fix yet.

Memory supercycle, take two. The setup now is different from past cycles. HBM is sold on multi-year contracts to a handful of AI buyers, 2026 capacity is locked, and the shortage may run to 2028. SK Hynix at +186% ytd and a $1T cap is the upstream tell. The risk is the same as always, memory is still cyclical, but the visibility this time is unusually strong.

SpaceX is set to price Thursday and debut Friday under SPCX at USD 135, a deal that could be the year's largest IPO. The real question for investors is not whether day one pops, but what the roughly USD 1.75 trillion valuation demands. With Goldman estimating capex near USD 360 billion through 2028, this is an infrastructure bet on launch plus Starlink, not a quick trade. The downside risk is real: that valuation sits well above some independent estimates, and a rising-rate, risk-off tape is an unforgiving backdrop for a richly priced new listing.

SpaceX IPO Nears: Is the Capex Story Bigger Than the Listing?

With SpaceX set to debut on the Nasdaq on June 12 under SPCX at USD 135 per share, the market's focus is shifting from the listing itself to the scale of the business behind it. According to Goldman Sachs estimates, SpaceX could spend as much as USD 360 billion in capital expenditure through 2028, a figure that reframes the company as a multi year infrastructure build rather than a one off IPO event.

The real question is not whether the debut pops. It is who captures that spending. Capex of that magnitude flows through a wide supply chain of aerospace, components, and satellite connectivity names, and appetite is already visible: the NASA ETF drew USD 2 billion of inflows in May alone.

Risks remain in play. The reported valuation near USD 1.75 trillion sits well above some independent estimates, and a market that just saw crypto and chips pull back may not reward richly priced new listings. Both the upside of a generational franchise and the downside of a stretched entry price are on the table for investors. 📊

two things stand out, BCRED gating redemptions and AI driven layoffs rising into payrolls. watch whether tonight's number breaks the soft landing story 🧠

Optical is quietly the tightest link in the whole AI buildout. The 2026 AI optical transceiver market is heading for around US$26 billion, but 800G output is running 40 to 60 percent below demand, and the choke point is the EML laser. Nvidia already moved to secure supply, committing US$4 billion to lock priority access at Lumentum and Coherent and pushing everyone else past 2027. For investors the read is simple: in a shortage, the names that actually own the laser capacity hold the pricing power, and right now that is a very short list.

Nine green sessions, then one hawkish Logan comment and the Dow sheds 600. The setup matters more than the headline: with the 10Y rising and the Beige Book flagging Middle East inflation, this is the market repricing rate risk it had been ignoring. Watch if dip buyers show up before Friday's NFP 🧠

HPE's Q2 Blowout: The After-Hours Pop Is About AI Servers, Not Just a Beat

Hewlett Packard Enterprise (HPE) reported fiscal Q2 results ahead of expectations and raised full-year guidance. The stock spiked as much as 40% in after-hours trading, one of the sharpest single-session moves for a large-cap hardware name this year.

Why this matters now

The real question is not whether HPE beat. It is what the guidance raise says about enterprise AI infrastructure demand. A lift of this size points to order momentum in AI servers and networking that is still building, rather than cooling into a digestion phase.

Where the upside sits

HPE occupies the part of the AI buildout that gets less attention than the chip names: the servers, storage, and networking that data centres need once the GPUs are bought. If hyperscaler and enterprise capex keeps flowing, the systems vendors capture a second wave of that spend.

The risk

A 40% reaction prices in a great deal. If the raised guidance leans on a small number of large AI orders rather than broad demand, the re-rating could prove fragile. Margins on AI servers also run thinner than on legacy gear, so revenue growth may not convert into proportional profit growth. Both the upside catalyst and these downside risks remain in play.