Welcome to the iPrompt Newsletter
Friday Edition
Sponsor:
LLM traffic converts 3× better than Google search
58% of buyers now start their research in ChatGPT or Gemini, not Google. Most startups aren't showing up there yet.
The ones that are get cited by the AI tools their buyers, investors, and future hires already use. And they convert at 3×.
Download the free AEO Playbook for Startups from HubSpot and get the exact steps to start showing up. Five minutes to read.
iPrompt Signals
AI & ROBOTICS INVESTING — EXPLAINED SO YOU CAN ACTUALLY ACT ON IT
Issue 16 · Friday 26 June 2026 · R. Lauritsen
Micron just posted the best profit margin in tech — and the Nasdaq fell anyway. The memory chip every AI system needs is sold out into 2027, which made Micron the most profitable name in the complex this week. The same shortage made Apple and Microsoft raise prices and fall 6% and 3%. That’s the issue in one line: the AI trade has split into companies that charge for the shortage and companies that eat it. Sorting your holdings into the two is the most useful thing you can do this weekend. (It also settles last week’s Micron question: structure, not cycle.)
Weekly Scoreboard
Ticker | Price | Week % | What happened |
NVDA | $195 | -7% | Dividend paid Thursday as promised. But the stock slid with Big Tech, off the week despite Micron’s read-through. The quiet self-funder, out of favour for now. |
AVGO | $378 | -7% | Co-designed OpenAI’s first custom inference chip, “Jalapeño.” Gave back the week’s gains in the Big Tech slide. The bookings story is intact; the tape isn’t. |
SPCX | $168 | -4% | Floated $25bn of debt to a “skeptical audience,” per Bloomberg — priced at a premium. The thin-float IPO is now a live test of exactly the funding window we keep flagging. |
MU | $1,166 | +12% | The print of the quarter, and the answer to last week’s tripwire. $41.5bn revenue, $25.11 EPS, Q4 guided to $50bn at ~86% margin, $22bn pre-paid commitments. Structure, not cycle — emphatically. |
ORCL | $199 | +3% | Quiet week, steadied again. The $40bn raise still hasn’t cleared — the worked example of the maturity wall, still unresolved. |
S&P 500 | 7,357 | flat | Essentially flat on the week — but that masks a four-day Nasdaq slide as money rotated out of mega-cap tech into the Dow, which hit a record. |
VIX | 19 | — | The “fear gauge” (30-day S&P swings; >25 = real anxiety). Crept higher on the tech wobble, but nowhere near panic. The slide is rotation, not fear. |
Bottom line: Memory just posted the best quarter in the buildout and the index that owns it fell anyway. The money isn’t leaving AI — it’s moving to whoever gets paid for the shortage, and away from whoever pays for it.
Prices are Thursday 25 June closes, rounded; weekly moves vs the prior Friday close.
Top Headlines
🌱 New to investing? Here’s the one idea to hold onto — “chargers vs eaters”: when memory costs jump, a company is either a charger (it makes the scarce chip, or has a brand strong enough to raise its own prices and make them stick) or an eater (it buys the chip and can’t pass the cost on without losing the sale). Micron is the charger of the week. Apple tried to charge and still got sold; a budget laptop-maker just eats it. Sort any AI hardware name you own into one of those two buckets and you’ve done the work this issue is asking for. |
1. Micron didn’t beat. It detonated.
Revenue of $41.5bn against a $36bn consensus — a $5.6bn beat. EPS of $25.11 versus $20.78 expected. Q4 guided to $49–51bn at a near-86% gross margin, and management disclosed 16 strategic supply agreements worth $22bn in committed customer cash, with floor pricing locked. They called supply constrained “beyond 2027.” The stock popped 17%. What changed: the bears needed a price concession to call it a cycle. They got the opposite — pre-paid, floor-priced contracts stretching past 2027. This is a structure, not a cycle.
2. And then Apple raised prices because of it
Here is the same fact from the other side. With memory prices surging, Apple raised MacBook and iPad prices and Microsoft lifted Xbox pricing; Apple fell 6% and Microsoft 3% on the day. The Nasdaq dropped for a fourth straight session even as Micron soared. Memory’s windfall is the device-makers’ cost line — and the market repriced both halves at once.
3. Qualcomm crashed the data-centre party
At its 24 June investor day, Qualcomm unveiled a full data-centre line — accelerators plus a Dragonfly CPU — and named Microsoft and Meta as customers, with Meta signing a multi-generation deal. A smartphone company is now a credentialed Nvidia challenger with hyperscaler logos attached. The read: the buildout’s pricing power assumes a narrow set of sellers. Every credible new entrant — Qualcomm this week, OpenAI’s own silicon below — chips at that assumption, even if revenue is years out.
4. OpenAI built its own chip — with Broadcom’s hands
OpenAI and Broadcom unveiled “Jalapeño,” OpenAI’s first custom inference processor, designed for large-language-model workloads. The headline is OpenAI going vertical; the investable story is Broadcom, which did the silicon. It is the clearest sign yet that the biggest model labs want to design around Nvidia for inference — and that custom silicon, not just GPUs, is where the next margin pool forms.
5. The hot inflation print nobody talked about
Lost in the Micron noise: Friday’s May PCE — the Fed’s preferred gauge — showed prices rising again, and durable-goods orders fell 4.5%. The hawkish dot plot from Issue 15 just got a supporting data point. Higher-for-longer is no longer only a projection on a chart; it is showing up in the prints. Why it matters: the funding clock from last week is still ticking under everything. Memory just proved a self-funder can shrug it off. The names that can’t — the ones facing the window — get no such reprieve.
Our Investing Angle
Everyone’s celebrating who got paid for the shortage. The smarter question is who has to pay for it next.
The thesis: You already have the lens: chargers at the top of the shortage, eaters at the bottom. The mistake this week is treating that as a tidy two-bucket split. It isn’t — the most interesting names sit in the middle, where the answer is “it depends.” Nvidia and Apple both pay more for memory, but one bundles it into a system the world can’t do without and the other leans on a brand. That’s a different kind of protection from Micron’s, and it’s thinner than it looks. Three things this week drew the real line:
1. Memory is now tech’s margin king — here’s the mechanism the headline number hides. Those 16 pre-paid contracts with floor pricing mean Micron books demand before it spends a cent on capacity — so a memory shortage doesn’t just lift its price, it transfers its capital risk onto the customer. That’s what a true top-of-chain position looks like: you charge for scarcity and someone else funds it.
2. The same shortage is a cost shock downstream — and the same transfer runs the other way downstream. The device-maker can’t pre-pay out of a shortage or make its own DRAM, so the cost lands whole — and its only defence is its own pricing power. Apple has it and raised prices this week; the stock fell anyway because even Apple’s buyers have a limit. A commodity box-maker has no such lever. The shortage doesn’t hit a sector evenly; it sorts the chain by who can charge their way out of it.
So name the losers, not a category. Memory-heavy hardware on thin margins wears it first: think the PC box-makers (Dell, HP), the consoles (Microsoft’s Xbox, already repricing), and any device brand without Apple’s pricing power to pass it on. The custom-silicon entrants — Qualcomm, and OpenAI via Broadcom — are the slower-burn threat to Nvidia’s inference margins. The winners are the three at the top of the shortage: Micron pre-sold, SK Hynix and Samsung rallying with it, and the equipment names that sell them the tools. What this issue can’t tell you is how much of each holding’s margin is exposed to a memory bill it can’t pass on. The deep dive builds the tool: a pass-through map that sorts the AI hardware chain into who charges for the shortage and who eats it: Memory’s windfall is someone else’s bill.
⚠️ What could go wrong? (The bear case) 1. The cost passes through cleanly. If end demand is strong enough, device-makers raise prices and buyers pay — as Apple is betting. Then the “downstream squeeze” never bites, margins hold, and shorting the memory buyers is just shorting a strong consumer. 2. The memory cycle turns, as it always has. Micron, SK Hynix and Samsung are all adding capacity. Memory has never escaped its cycle for long; coordinated supply additions could normalise pricing into 2027 and turn today’s “structure” back into a textbook glut. The margin-king crown is the most cyclical thing in tech. 3. The new chips are vapour for years. Qualcomm’s and OpenAI’s silicon ship in volume in 2026–27 at best, against an Nvidia software moat a decade deep. If they slip — as first-gen accelerators usually do — the “competition for Nvidia” angle is a 2028 story dressed up as a this-week one. Size your position for the possibility that the downstream squeeze is shallower than the math suggests — and that memory’s margin crown is, as ever, rented not owned. |
Three Ideas to Research This Weekend
Not recommendations — starting points for your own research. One lead idea, two quick follows.
Lead idea — The pass-through trade: who eats the memory bill
Why now: Apple and Microsoft raised hardware prices this week because memory got dearer, and both stocks fell on the news. The cost shock from Micron’s shortage is now visibly hitting the companies that buy memory to ship product.
The case: the cleanest version of the trade is the spread, not a single name: long the makers of the shortage (memory), wary of the buyers who can’t pass it on (commodity PC and console hardware). The shortage that mints Micron’s margin is a tax on everyone downstream of it.
The risk: Apple has the brand to make price rises stick, so “memory victim” may not hold for the strongest names. The squeeze is real; the list of who actually suffers is shorter than the math implies.
Tripwire: the next round of PC and smartphone vendor guidance (mid-July onward). Any maker guiding gross margin down and citing memory or component costs = the pass-through is biting, and the downstream-squeeze thesis is live. Margins held with prices raised = the buyers have pricing power and the trade is wrong.
How to research: the memory side via MU or a broad semi ETF like SMH; the squeezed side by reading device-maker margin guides directly. The single question: for any hardware name you own, can it pass a memory price rise on, or does it eat it?
Quick follow — The custom-silicon pick-and-shovel: Broadcom
I keep coming back to this one. Every story about a lab building its own chip to escape Nvidia — OpenAI’s Jalapeño this week — is, underneath, a Broadcom story, because Broadcom does the silicon. You can stay neutral on which lab wins and still own the company that gets paid to design every custom accelerator. Tripwire: Broadcom’s next AI-bookings update versus the $30bn-plus it has flagged. A higher custom-silicon backlog with new named labs = the design-around-Nvidia thesis is compounding. A flat number, or Google in-sourcing more of its own chips = the moat is narrower than the bulls think. How to research: AVGO direct, or a custom-silicon-weighted semis ETF; either way, watch the bookings line, not the share price.
Quick follow — The non-US tell: SoftBank’s March-2027 wall
Look, I’m wary of single-stock macro proxies — they’re a clever story until they’re just a stock. But SoftBank earns it: a $40bn unsecured bridge loan maturing March 2027 — a concrete date on the maturity wall — and a $6bn margin loan against its OpenAI stake that just failed to close. With money confirmed dearer, that 2027 wall is the cleanest deadline-risk in the Global layer. Tripwire: how SoftBank refinances that $40bn bridge as March 2027 nears. A roll at a higher coupon = the wall is real and biting the most levered first. An OpenAI IPO that lets it repay from asset sales = the wall clears and the leverage flips to a tailwind. How to research: 9984.T (Tokyo-listed) direct; the broader non-US AI complex via an ex-US tech ETF.
AI Investment Framework
6 layers · Updated weekly · Not financial advice

Layer | This week’s signal |
Infrastructure | Signal up to ↑. Micron’s blowout validated the self-funder thesis and lifted the whole memory complex (SK Hynix, Samsung, Kospi). The layer led on fundamentals even as the tape sold mega-cap tech. MU is now tech’s margin king. |
Platforms | Signal stays ↓. Apple and Microsoft fell on hardware price rises driven by memory costs; the platforms now wear the shortage as a margin line. OpenAI’s custom chip (via Broadcom) is a slow crack in the GPU-only moat. Cash flows still vast; the cost base just grew. |
Applications | Still the YTD laggard at -13%. A confirmed higher-for-longer path (hot May PCE) keeps pressure on long-duration software multiples. PLTR/CRM/NOW remain the proof-or-not watch, now with the rate overhang reinforced. |
Physical AI | Signal back to ↑. Nvidia’s Halos safety stack for physical AI extends the platform into robotics, but the layer cooled after last week’s shipping surge and remains the most window-dependent, capital-hungry bet. The cleaner cycle read remains FANUC order intake. |
Cybersecurity | Signal ↑. CRWD firmed as defensive, cash-generative software caught the rotation out of mega-cap tech. The 26 Aug print is still the permission slip to add. Roughly flat on the year. |
Global | Signal up to ↓. Micron’s beat lifted SK Hynix, Samsung and the Kospi, easing some stress — but SoftBank’s March-2027 bridge loan keeps the layer’s deadline risk live. BABA steady; SAP the boring ballast. |

Infrastructure’s late-June tick up is the Micron print in one line — the self-funder at the top of the shortage, pulling away from the device-makers who have to pay for it.
WHAT WE’RE WATCHING
Date | Event | Question to track |
Mon 30 Jun | Alphabet joins the Dow | Alphabet replaces Verizon in the Dow while raising $85bn for its buildout. Does index inclusion steady a stock on a four-week losing streak — or does the funding need keep weighing? |
Mid-Jul | PC / phone vendor guides | Do device-makers guide margins down and blame memory costs? The first vendor to flag a memory squeeze confirms the pass-through thesis is biting. |
Late Jul | Hyperscaler Q2 capex guides | Do MSFT/GOOG/AMZN hold or raise spend while absorbing higher memory costs? The first guide that cuts reprices the entire memory-demand thesis underneath Micron. |
CHANGES THIS WEEK
Infrastructure: signal ↑ → ↑↑? No — held at ↑. Micron’s blowout is a fundamental confirmation, not a fresh catalyst to chase; the stock already ran into it. Conviction stays HIGH. The thesis was proven, not upgraded.
Platforms: signal ↓, held. Memory costs are now a visible drag on Apple and Microsoft margins, and custom silicon (OpenAI via Broadcom) chips at the edges. Conviction unchanged at HIGH; the cash flows are intact, the cost base isn’t.
Physical AI / Global: Physical AI eased ↑↑ → ↑ (consolidation after the shipping surge). Global improved ↓↓ → ↓ as Micron’s beat lifted SK Hynix and Samsung; SoftBank’s 2027 wall keeps it from going higher.
Conviction Check — Infrastructure (HIGH, unchanged): why not push to a higher conviction after a print this good? Because the framework rewards demand durability, not single beats. Micron confirmed the structure; the conviction already reflected it. It breaks on a demand crack — a hyperscaler capex cut — not on one blowout quarter. Watch the late-July guides.
Everything else holds.
Your Move
1. This week’s one task: for every hardware name you own, ask one question — can it pass a memory price rise on to its customers, or does it eat it? Apple can. A commodity PC or console maker can’t. Sort your holdings into chargers and eaters. An eater isn’t an automatic sell — it’s a name to look at twice, because that’s where the shortage quietly becomes your margin problem.
2. Micron settled the self-funder question: $41.5bn revenue, $50bn guided, 86% margins, supply sold past 2027. The structure-versus-cycle debate is over for now — the top of the chain is a structure.
3. But the same shortage is a cost shock downstream — Apple and Microsoft fell on price rises driven by it. The buildout is now a chain with a winner up top and a bill flowing down, and your job is to know which end each holding sits on.
Now research one. One holding, one question, tonight: charger or eater? Write “can pass memory costs on?” next to each AI hardware name you own and answer yes or no. Make it the Friday ritual — one holding per issue — and inside a month your whole book is sorted by who profits from the shortage and who pays for it.
🌱 Short Take One idea: if you only act on one thing this week, it’s the chargers, not the eaters. The makers of the shortage have pricing power the buyers can only dream of. One action: if you want exposure to the makers of the shortage without picking one name, a broad semiconductor ETF like SMH owns the memory-to-GPU chain that gets paid for the bottleneck — Micron included. Not a recommendation — a starting point. |
Stay curious — and stay qualified.
— R. Lauritsen
Editor, iPrompt Signals
P.S. Issue 15 asked who pays the funding toll first. Micron answered a different question this week — who gets paid for the shortage — and the answer was emphatic. But Oracle’s $40bn still hasn’t cleared, and SoftBank’s $40bn bridge still matures in March 2027. The maturity wall didn’t go away. It just got out-shouted by a blowout.
P.P.S. One-word reply, and it shapes next week: after Micron’s print and the Big Tech slide, are you net buying or net selling your AI exposure here? IN or OUT? Your replies beat any sentiment survey the sell side runs.
Quick Glossary
PCE (Personal Consumption Expenditures) — The Fed’s preferred inflation gauge. May’s reading showed prices rising again — a data point that backs up last week’s hawkish dot plot.
Gross margin — The share of revenue left after the direct cost of making a product. Micron guided to ~86% — higher than Nvidia’s — which is why it’s suddenly called tech’s margin king.
Cost pass-through — Whether a company can raise its own prices to cover a cost increase. Apple passing higher memory costs to buyers is pass-through; a maker that must absorb it has none.
Custom silicon (ASIC) — A chip designed for one specific job rather than a general-purpose GPU. OpenAI’s Jalapeño, built with Broadcom, is a custom inference ASIC aimed at cutting reliance on Nvidia.
HBM (High-Bandwidth Memory) — The stacked memory bolted onto AI chips, and the bottleneck of the buildout. Micron says supply is constrained beyond 2027 — the heart of its pricing power.
Maturity wall — The schedule of when a company’s debt comes due and must be repaid or refinanced. SoftBank’s $40bn bridge maturing March 2027 is this issue’s live example.
Strategic customer agreement (SCA) — A long-term, often pre-paid supply contract with floor pricing. Micron disclosed 16 of them worth $22bn — the mechanism that turns a memory cycle into a fundable structure.
Sponsor:
HubSpot AEO
Picture this. A buyer opens ChatGPT and asks for a recommendation in your category. Your competitor's name comes up. Yours doesn't. And that buyer never makes it to your website.
That's happening right now in markets everywhere. And most teams don't know it's happening because it never shows up in their analytics.
HubSpot AEO shows you exactly where your brand stands in AI search, where competitors are getting recommended instead of you, and tells you specifically what to fix. No expertise needed.
Try it free for 28 days. Just $50 a month after.


