Sponsored by
Welcome to the iPrompt Newsletter
Friday Edition
What you get in the Friday Edition
Weekly Scoreboard
Top Headlines of the Week
Our Investing Angle
Three Ideas to Research This Weekend
AI Investment Framework
Your Move
…all in a FREE Weekly newsletter.
Sponsor:
Stop Paying for 6 Tools. One AI Does It All
Most e-commerce sellers are running their store across 6 to 8 separate tools — and paying hundreds of dollars a month for the privilege. StoreClaw replaces your entire stack with one autonomous AI engine that monitors competitors, optimizes listings, automates marketing, and tracks real profit across Shopify, Amazon, and beyond.
It doesn't wait for you to ask. It runs 24/7 in the background, so you wake up to a full dashboard instead of a list of things you forgot to check.
Connect your store, and StoreClaw gets to work — no prompts, no complex setup, no six-app stack.
Free to start. No credit card required.
iPrompt Signals
AI & robotics investing — explained so you can actually act on it.
ISSUE 11 // Friday, 22 May 2026 // 6 min read
The Hook
On Wednesday night Nvidia printed the cleanest quarter of its life — record $81.6 billion in revenue and a Q2 guide that sailed past every estimate on the Street. By Thursday’s close the stock was down 1.8%. Last Friday I told you the market had spent its surprise budget. It just did — and the receipt is a perfect print that moved nothing. The catalyst isn’t the story anymore. What’s repricing AI now is sitting in the bond market.
🎯 YOUR ONE DECISION THIS WEEKEND The earnings season is over. The macro season just started. Pick one of three lenses for a rate-driven market: 1. TSM — the cash-rich compounder trading at a discount to the chip names that need to borrow. 2. CRWV / debt-funded compute — the short-side lens. Who refinances into a 5%+ long bond? 3. Japan robotics — the rate-insensitive watchlist add. Cash flows now, not in 2028. Everything below builds the case. The deliverable: one ticker, one tripwire, before Monday’s open. |
Weekly Scoreboard
Ticker | Price (21 May) | Week % | What Happened |
NVDA | $219.51 | −7.0% | Perfect print. Stock fell anyway. |
AVGO | $412.11 | −7.9% | Custom-silicon trade gave back the rally. |
TSM | $407.15 | −0.2% | Flat. ~20% discount to the SOX index. |
MU | $762.10 | +1.4% | Memory still the rally’s last leg. |
AMD | $449.59 | +1.7% | Held up. ARM up 16% on the week. |
BOTZ | $40.10 | −0.9% | Kawasaki–Nvidia physical-AI tie-up. |
S&P 500 | 7,432.97 | −1.1% | Off last week’s 7,512 record high. |
US 30Y | 5.19% | +24 bp | Highest yield in 19 years. |
Bottom line: the best earnings report in the sector landed in the same week as a 19-year high in long-term yields — and the yield won. That is the whole issue in one line. |
The Three Stories That Matter
1. Best quarter ever. Stock down anyway.
(Reported: Nvidia IR, CNBC, 20–21 May.) Revenue $81.6bn, up 85%. A Q2 guide of $91bn against an $87bn consensus. A 25-fold dividend hike and a fresh $80bn buyback on top. You cannot write the quarter better. The stock fell 1.8% Thursday, 7% on the week. What it means: the demand question is settled. The market didn’t fall on doubt about Nvidia — it fell because the thing that prices Nvidia changed underneath it.
2. The bond market quietly became the most important AI chart.
(Reported: CNBC, CNN, 19–21 May.) The 30-year Treasury yield hit 5.19% — the highest in 19 years — after a weak 20-year auction and sticky inflation data. The 10-year is near 4.6%. My read: the AI demand story looks intact. The AI valuation story now has a rate problem it didn’t have a month ago — because growth stocks are long-duration assets, and a higher discount rate works against them (the box below unpacks how).
🌱 New to investing? Here’s what this means: Think of a stock as a promise of future cash. A government bond yield is the “risk-free” return you could get instead. When that safe yield jumps from 4% to 5%, every risky promise has to look more attractive to compete — so investors pay less for it today. The same Nvidia, the same earnings, a lower price. Nothing about the company changed; the alternative got better. |
3. Beijing is now blocking the chips Washington approved.
(Reported: Bloomberg, Tom’s Hardware, CSIS, 15–20 May.) Set the rate story aside for one paragraph — this is the week’s secondary risk, and it closes a loop. The Trump–Xi summit produced warm words and no signed tech document. The twist: roughly ten Chinese firms hold approved US H200 licences — and not one chip has shipped, because Beijing is steering its own companies toward Huawei and domestic GPUs instead. Last Friday’s bear case said “the summit produces nothing concrete.” It did one better: it produced a green light nobody will use. Whether that’s a delay or a permanent redirection is the open question — and it reprices the China-AI complex either way.
Supporting signals (for portfolio adjustments, not main trades)
● Kawasaki × Nvidia (reported, 21 May): Kawasaki Heavy will build physical-AI robotics on Nvidia’s stack and open a US robot centre. The infrastructure layer is reaching into the Physical AI layer — a structural read-through for the robotics names, not a one-day trade.
● Memory holds the tape (reported, CNBC 17–21 May): Micron +1.4% on the week even as the chip leaders sold off; Seagate had its worst week since March on capacity-ramp worries. Within a falling sector, memory is doing the heavy lifting — a narrowing of leadership worth watching.
Our Investing Angle
Everyone’s asking whether Nvidia’s quarter was good enough. Wrong question. The quarter was extraordinary and the stock still fell — so the question is what’s pricing AI now that earnings can’t.
The thesis (my call — calibrate against your own read): the AI trade has quietly switched engines. For two years it ran on demand surprise — every quarter beat, every guide rose, every stock followed. That engine just stalled, not because demand fell but because the surprise is fully priced. The new engine is the discount rate — and it sorts the sector into who earns cash now versus who promises it later.
The split: on one side, the self-funders — TSM, Nvidia, the cash-generative core that doesn’t need the bond market. On the other, the duration-stretched. The risk-adjusted move is to shorten your duration — favour the side of the sector whose cash arrives sooner.
Who gets hurt — and how. These are three different mechanisms, not one:
● Refinancing risk. CoreWeave — roughly $25bn of debt that has to be rolled into a rising long bond. The hit lands through interest expense, not the multiple.
● Valuation-multiple risk. Palantir at ~46x forward sales — the business is fine; the multiple is the rate-sensitive part. A higher discount rate compresses the multiple directly.
● Backlog-funding risk. Oracle’s debt-funded compute backlog — a hybrid of the first two, and it reports 11 June. Plus the single-name longs who bought Nvidia for the print: the catalyst landed and paid nothing.
→ Deep dive: the AI duration map — every layer scored by how far out its cash flows sit, and what a 5.5% long bond does to each. Read it →
⚠️ WHAT COULD GO WRONG? (the bear case) 1. Yields reverse fast. One soft inflation print and the 30-year drops back through 5%. The duration trade unwinds and the high-multiple names you avoided lead the bounce. 2. AI demand is so strong it overrides rates. If hyperscaler capex guides keep rising and cloud backlogs keep growing through the June prints, earnings growth can simply outrun a higher discount rate — watch Oracle’s 11 June backlog number as the first hard checkpoint. 3. The cash-rich names aren’t cheap either. “Shorten duration” is not “buy anything safe.” If TSM re-rates with the sector, balance-sheet quality won’t save you from a multiple reset. Size for the possibility that this is a rate scare that passes — not a regime change. Conviction, not concentration. |
Your Three Research Paths
Not recommendations — starting points. Pick one. Tighten it before Monday.
Path 1 — Taiwan Semiconductor (TSM). The self-funder trading at a discount.
Why now: TSM trades at roughly a 20% discount to the SOX semiconductor index on forward P/E (reported, Bernstein via CNBC, 18 May) while the chip names it manufactures for sold off on rate fears.
Thesis: Every AI chip — Nvidia, AMD, Broadcom, the hyperscalers’ custom silicon — is fabricated by TSM. It funds its own capex from operating cash flow, so a 5% long bond barely touches it. It is the lowest-duration way to own the entire AI build-out.
Risk: Taiwan geopolitical risk is permanent and binary. And a discount can persist longer than your patience.
Tripwire: TSM’s discount to the SOX index closing back inside 10% = the market has noticed and the easy gap is gone. Discount widening past 25% = either a real warning or a deeper opportunity — check why before acting.
How to research: Ticker TSM (US ADR) or 2330.TW. Compare against SOXX/SMH for the diversified version.
Path 2 — Debt-funded compute (CoreWeave, CRWV). The lens I keep coming back to.
I keep coming back to this one because it’s the cleanest test of the whole thesis. Why now: CoreWeave carries roughly $25bn of debt and a capex plan that needs continuous refinancing. The 30-year just hit a 19-year high.
Thesis: This isn’t a call to short anything — it’s the cleanest live test of refinancing risk. CRWV has huge backlog, real revenue, and a balance sheet that lives or dies on the cost of capital. If the rate thesis is right, it shows up here first.
Risk: Backlog of ~$99bn is genuine. If CoreWeave terms out its debt before yields rise further, the squeeze never arrives.
Tripwire: Any new CoreWeave or Oracle debt issue priced above 7% = the rate thesis is live and visible. Oracle’s 11 June print is the calendar checkpoint — watch interest expense as a share of cloud operating income.
How to research: CRWV and ORCL filings — read the debt maturity schedule, not the revenue line. This is a lens for sizing your other positions, not necessarily a trade itself.
Path 3 — Japanese industrial robotics. The rate-insensitive watchlist add.
Look, I’ve flagged this trade three issues running and it hasn’t broken out yet. But in a duration-driven market the logic only gets stronger. Why now: the Kawasaki–Nvidia physical-AI partnership (reported, 21 May) puts a fresh spotlight on Japan’s robotics base just as rates punish anything that promises cash in 2028.
Thesis: FANUC, Keyence and Yaskawa earn real industrial cash flows today at roughly 28x forward earnings versus ~58x forward earnings for US humanoid names whose revenue is years out — same secular trend, half the multiple.
Risk: It’s been cheap for a reason — slower growth. And a stronger yen can erode the export earnings.
Tripwire: A Japanese robotics name landing a named Nvidia physical-AI design win before end-July = the re-rating catalyst arrives. No partnership news by Q2 earnings (late July) = it stays a slow-burn value hold, not a catalyst trade.
How to research: BOTZ (heavy Japan and industrials weighting). FANUC (6954.T), Keyence (6861.T) directly.
AI Investment Framework
Living portfolio framework. Not financial advice — research starting points only.
Layer | Tickers | YTD | Conviction | Risk | Sizing |
Infrastructure | NVDA, AVGO, TSM | +22.1% | HIGH ↔ | ●●●○○ | 15–20% |
Platforms | GOOGL, MSFT, AMZN | +18.4% | HIGH ↔ | ●●○○○ | 15–20% |
Applications | PLTR, CRM, NOW | −6.1% | MEDIUM ↓ | ●●●●● | 5–10% |
Physical AI | BOTZ, ISRG, TSLA | +9.2% | DEVELOPING ↑ | ●●●●○ | 5–10% |
Cybersecurity | CRWD, PANW, ZS | +6.4% | MEDIUM ↔ | ●●●○○ | 5–10% |
Global | BABA, 9984.T, SAP | +15.9% | DEVELOPING ↓ | ●●●●● | 5% |
The layers up close
Infrastructure
The picks-and-shovels layer — and now the most rate-watched.
NVDA — record quarter, falling stock. Earnings no longer set the price.
AVGO — custom-silicon trade gave back the week. Watch the next ASIC design win.
TSM — the self-funder. ~20% discount to the SOX on forward P/E; least exposed to the long bond.
Platforms
Cloud and model APIs — the layer that monetises everyone else’s compute.
MSFT — Azure AI is the anchor. Watch capex-to-cash-flow as rates bite.
GOOGL — Gemini traction, cheapest mega-cap multiple. Antitrust is the overhang.
AMZN — AWS reacceleration is the swing factor. Margins over headline growth.
Applications
Enterprise AI software — highest growth, highest multiple, longest duration.
PLTR — ~46x forward sales. The most rate-sensitive position we track; the multiple is the risk.
CRM — Agentforce adoption is the proof point. Watch net revenue retention.
NOW — the steadier compounder. Still priced for flawless execution.
Physical AI
Robots and factory automation — cash sooner than the humanoid hype suggests.
BOTZ — heavy Japan and industrials. The rate-insensitive way in.
ISRG — surgical robotics, real recurring revenue. The quiet compounder.
TSLA — Optimus is option value; timeline keeps slipping. Late-July V3 reveal.
Cybersecurity
AI cuts both ways — this layer sells the defence.
CRWD — platform consolidation intact. Watch module-attach rate.
PANW — the platformisation bet. Slower but broader than CRWD.
ZS — zero-trust demand is structural; the multiple has priced a lot of it.
Global
Non-US AI — the diversifier, hostage to the US–China chip stalemate.
BABA — cheap on cloud plus Qwen, but the H200 licence it holds is one Beijing won’t let it use.
9984.T — SoftBank as a leveraged proxy for the whole AI buildout. High beta, high noise.
SAP — the steady European enterprise-AI play. Boring is the point in a volatile layer.
YTD performance
Changes this week
● Infrastructure cut HIGH ↑↑ → HIGH ↔. Demand conviction is unchanged — the signal cut is about valuation in a higher-rate market, not the businesses.
● Applications held MEDIUM ↓. PLTR at ~46x sales is the most rate-sensitive position in the framework. Risk raised to 5/5 dots.
● Global cut DEVELOPING ↑↑ → DEVELOPING ↓. The H200 framework exists on paper but Beijing is blocking it. Last week’s asymmetric catalyst is now a stalemate.
Next two dates that matter
Date | Event | What It Resolves |
Tue 27 May | US Treasury 7-year auction | Auction demand is the cleanest read on whether the yield spike has further to run. |
Thu 11 Jun | Oracle Q4 earnings | Interest expense vs cloud operating income — the debt-funded-compute thesis, tested. |
Disclaimer: This newsletter is for informational and educational purposes only and does not constitute financial advice. iPrompt Signals is not a registered investment advisor. Always conduct your own research and consult a qualified financial professional before making investment decisions.
Your Move
Pick one of the three research paths from the top of this issue. 45 minutes. Latest filing. Your own tripwire. Hit reply and tell me which lens you’re taking into a higher-rate market.
Two things to do before Monday:
● Check one position you own for refinancing exposure — does it need the debt market in the next 12 months? If yes, the long bond is now your risk, not its earnings.
● Put Oracle’s 11 June print on your calendar. Its backlog and interest expense are the first hard test of whether demand outruns rates or the reverse.
🌱 SHORT TAKE For the broad-exposure reader: the lesson this week isn’t about one stock — it’s that rates now matter as much as earnings. If you don’t want to pick sides, SOXX (iShares Semiconductors) or SMH (VanEck) hold the cash-rich names and the duration-stretched ones together — you average the split rather than betting it. Not a recommendation — a starting point. |
Stay curious — and stay qualified.
— R. Lauritsen
Editor, iPrompt Signals
Know someone building an AI position? Forward this — they’ll thank you by Friday.
P.S. — Last Friday I wrote that the market had “spent its surprise budget” into Nvidia’s print. I’ll be honest, I half-expected to be wrong — a $91bn guide usually overrides everything. It didn’t. A 10-out-of-10 report, and the stock still fell. That’s the part worth sitting with.
TERMS USED THIS WEEK
Beat-and-raise — A quarter where a company beats current estimates and also raises its forward guidance. Nvidia did both and the stock still fell.
Discount rate — The return investors demand to hold a risky asset. It rises with bond yields — and a higher discount rate lowers what any future cash flow is worth today.
Duration — How far in the future an asset’s cash flows sit. Long-duration assets (high-multiple growth stocks) lose the most value when rates rise.
H200 — Nvidia’s export-market GPU for China. Currently US-approved but Beijing-blocked — a licence nobody is using.
SOX index — The Philadelphia Semiconductor Index, the benchmark basket of chip stocks. TSM trading at a discount to it is the gap behind Path 1.
Self-funder — A company that pays for its own capital spending from operating cash flow — so a rising bond yield barely touches it.
Sponsor:
$992 Billion in Art Could Change Hands. Why Are These 71,105 Investors Paying Close Attention?
Deloitte ran the numbers. They project UHNW art and collectibles wealth -- already at $2.5 trillion -- to hit $3.47 trillion by 2030.
The institutional world has been quietly preparing for this. Back in 2011, 25% of wealth managers surveyed offered art-related services. In 2024, 51%. Family offices now average a 13.4% allocation to art and collectibles. And it’s not just because they love art. It’s because they like the math.
These positions were built over decades through private dealer relationships most investors never had. The access just wasn't there.
Masterworks is changing that:
71,000+ investors
$1.3B deployed across 525+ artworks
29 closed sales
Net annualized returns like 16.5%, 17.6%, and 17.8%, not including those unsold.
Investing involves risk. Past performance is not indicative of future returns. See important disclosures at masterworks.com/cd.