Oracle beat everything. The bill came anyway.

iPrompt Signals

AI & ROBOTICS INVESTING — EXPLAINED SO YOU CAN ACTUALLY ACT ON IT

Issue 14 · Friday 12 June 2026 · R. Lauritsen

Oracle did everything right on Wednesday night, and the stock fell 9% anyway. The number that did the damage wasn’t in the results. It was $40 billion: what Oracle needs to raise next year to keep building. The AI trade just got a bill, and from here the market is sorting names by who can pay it. This issue shows you the sort; the deep dive gives you the tool to run it on your own holdings.

Weekly Scoreboard

Bottom line: Oracle fell 9% on a clean beat while the index round-tripped to flat. The market has stopped grading results — it’s started grading funding plans.

Prices are Thursday 11 June closes, rounded; weekly moves vs the prior Friday close.

Top Headlines

1. Oracle’s perfect quarter arrived with a $40bn invoice

Revenue of $19.2bn, up 21%. An earnings beat. Backlog up $85bn in one quarter, to $638bn. And underneath it: $55.7bn of capex this year ($5.7bn over guidance), a plan to spend roughly $70bn (net) next year, and about $40bn of debt and equity to raise on top of the $43bn of debt already taken on. The stock dropped about 9%. What changed: the market read straight past the income statement to the funding plan, and priced the dilution before the growth.

2. $75bn just left the pool

That’s the maximum raise for SpaceX, which priced the largest IPO in history at $135 a share, a $1.75tn valuation, and lists on Nasdaq today as SPCX. The prospectus shows Starlink earning $11.4bn of revenue at a 63% EBITDA margin inside a group that still loses money overall. Those are the filings. Our read is the part that matters to stocks you already own: that $75bn comes out of the same pool that funds every other AI name, and liquidity at this scale is zero-sum. Jim Cramer spent the week warning that investors are selling CrowdStrike and Broadcom to make room.

3. Memory got receipts: the Nvidia–SK Hynix pact, and Micron’s HBM is sold out

On Sunday, Nvidia and SK Hynix signed a multi-year partnership spanning the Vera Rubin platform and the Jetson Thor robotics brain, and the Korean memory giant pledged to triple its HBM capacity. Micron — certified as an HBM4 supplier for the same platform, with its remaining FY26 HBM output already committed under long-term contracts — guides to roughly $33.5bn of revenue at its 24 June print. The funding angle: memory’s growth is customer-funded. Buyers commit before the chips exist — the strongest position on this week’s ladder, and the reason MU erased a 9% drop in a day.

4. Intel had the best week in the AI trade

Up 11% on Monday after Alphabet selected it for chip manufacturing; up 10% again on Thursday on an analyst upgrade; up well over 200% this year. Nobody on a Computex stage called Intel the next trillion-dollar company. It just signed a customer, and in a funding-conscious market, a manufacturing contract from a hyperscaler beats a compliment from a keynote. That is the lesson Marvell taught the hard way last issue.

5. The week wasn’t about AI at all

May CPI printed at a three-year high, producer prices at 2022 levels, and the ECB became the first major central bank to answer the energy shock, hiking to 2.25%. A Thursday-night de-escalation in the Gulf dropped oil 3% and lifted the S&P 1.8%, but a December Fed hike stayed fully priced. The read: strip the noise and one thing happened — money got more expensive. Dearer money punishes whoever needs to borrow or dilute to build, which is the whole issue in a sentence. The next print that matters is the Fed’s dot plot on Wednesday.

Our Investing Angle

Everyone’s watching the round trip. The smarter bet is watching who paid for it.

The thesis: the AI trade has grown a second axis. Issue 13’s sort — numbers versus noise — still stands. This week bolted a funding axis onto it: the cost of money turned positive for the first time in this buildout, and it doesn’t tax every AI name equally. Two groups are forming:

1. Self- and customer-funded — growth paid from operating cash flow or from customers’ own commitments. NVDA raised its dividend this month: a capital-returner in a capital-hungry sector. Micron’s HBM is pre-sold under long-term contracts; Broadcom sits on $30bn+ of AI bookings. For these names, rate prints are background noise.

2. Market-funded — growth that needs the equity or debt window to stay open. Oracle’s $40bn. Alphabet’s announced equity raise, and the reports that Meta is weighing one. The leveraged neocloud builders. A record IPO queue, with SpaceX at the front of it. For these names, every CPI print is now an earnings event.

The losers aren’t a sector — they’re a capital structure: Oracle until the raise clears, SoftBank as the most levered AI proxy in our Global layer, the CoreWeave-class neoclouds, and — if Meta confirms — the comfortable assumption that hyperscalers are immune. The winners are the boring self-funders and whoever gets paid either way. But here’s what this issue can’t do: tell you which group each of your holdings is actually in. Some names sit in both at once — Oracle does — and the split isn’t visible from the share price. The sorting tool is this week’s deep dive: a four-rung funding ladder, a five-minute test using only published filings, and the Oracle worked example: Who pays for the AI buildout?.

Three Ideas to Research This Weekend

Not recommendations — starting points for your own research. One lead idea, two quick follows.

Lead idea — The customer-funded print: Micron into 24 June

Why now: the stock just round-tripped to flat into the most loaded report of the month: guidance of roughly $33.5bn revenue and a near-tenfold jump in EPS, with HBM4 certification for Nvidia’s next platform in hand.

The case: the strongest funding position in the stack. Customers commit under long-term contracts before capacity exists, so the buildout costs Micron’s shareholders the least — no dilution, no debt binge, just pre-sold output.

The risk: SK Hynix tripling HBM capacity is the seed of the next glut, and at around a $1tn valuation the market is pricing a memory cycle as a permanent structure.

Tripwire: the FY27 HBM pricing commentary on the 24 June call. Contract prices confirmed at or above 2026 levels = the structural story holds. Any concession on price = the cycle is still a cycle, and the multiple is wrong.

How to research: MU direct; or SMH if you’d rather own the whole memory-to-GPU chain than one print. The one question to answer before the 24th: are you paying for a cycle, or a structure? The pricing line settles it.

Quick follow — The dividend tell: Nvidia

I keep coming back to one small number from the past fortnight: Nvidia raised its dividend. In a sector where everyone else is borrowing or diluting to build, the biggest spender of all is handing cash back — that’s rung one of the funding ladder, advertised. Tripwire: the July round of hyperscaler capex guides. MSFT, GOOG and AMZN all holding or raising while funding from cash flow = the self-funders’ premium is earned; the first cut reprices everything, self-funders included. How to research: NVDA direct; the platform layer via QQQ.

Quick follow — The toll collectors: ASML and the equipment names

Look, “picks and shovels” is the most consensus line in investing, and I’m wary of it. But the logic survives this week intact: Oracle alone plans $70bn of net capex next year, and every one of those dollars, however it’s raised, flows through the same fabs and the same Dutch lithography machines. Equipment gets paid whoever wins, and whoever funds it. The risk is the mirror image: equipment is the first order cancelled if funding stress turns into capex cuts. Tripwire: ASML’s Q2 report in mid-July. Net bookings up sequentially = the buildout is funded and rolling; a sequential decline while capex guides still stand = financing strain reaching the supply chain early. How to research: ASML (Amsterdam-listed) direct; SMH includes the equipment complex.

AI Investment Framework

6 layers · Updated weekly · Not financial advice

A $1tn wipeout and a full recovery inside one shaded column. Flat isn’t calm — it’s two forces cancelling.

WHAT WE’RE WATCHING

CHANGES THIS WEEK

Infrastructure: signal ↓ → ↔. The selloff round-tripped, and the demand evidence (the SK Hynix pact, “insatiable” bookings) outweighs the price noise. Conviction unchanged at HIGH.

Cybersecurity: signal ↑ (recovering). The de-rating found a floor; we’d still rather add on the 26 Aug print than ahead of it.

Global: signal ↓ (cooling). The funding toll lands on the levered names first. SoftBank is the watch item.

Conviction Check — Platforms (HIGH, unchanged): why hold HIGH while Alphabet and possibly Meta sell stock? Because raising equity at record prices to fund contracted demand is opportunism, not distress. The conviction breaks on a capex cut, not on a raise — watch the July guides, not the filings.

Everything else holds.

Your Move

1. This week’s one task: pick your largest AI holding and find a single ratio — capex divided by operating cash flow, latest fiscal year. Above 1.0 means the market is funding the growth, and you’re the market. Ten minutes, one filing. Everything else below exists to give that number context.

2. Wednesday tells you what the number costs: the Fed’s dot plot is now an AI earnings event, and a validated December hike taxes the above-1.0 names first. Oracle just showed you the repricing.

3. SPCX tells you how urgent it is: comfortably above the $135 issue price and the funding window is open, so the toll argument waits. A broken IPO means tighten the above-1.0 exposure before the next raise is announced.

Now research one. The ratio, one holding, tonight. Write the number down, put the next earnings date in your calendar, and decide before the print, not after it. Then make it a Friday ritual — one holding per issue, and your whole AI book is graded inside a month.

Stay curious — and stay qualified.

— R. Lauritsen

Editor, iPrompt Signals

P.S. Issue 13 asked whether the selloff was a RESET or a TOP. The week’s answer was “both, within four sessions” — a $1tn wipeout fully recovered by Thursday, with the proven-cash names recovering first. That’s the reset pattern. The formal verdict is still Micron’s 24 June print.

P.P.S. One-word reply, and it genuinely shapes next week’s issue: would you have taken SpaceX at $135 — IN or OUT? Your answers are a better funding-window survey than anything the sell side publishes.

Quick Glossary

Capex (capital expenditure) — Money spent on long-lived kit — land, buildings, chips, data centres. It leaves the bank account now and earns its keep, or doesn’t, over years.

Dilution — What happens to your stake when a company sells new shares to raise money — more slices, same pie. Oracle’s $20bn programme is this week’s case study.

Dot plot — The Fed’s quarterly chart of where each official expects rates to go. Markets read it as the central bank’s voting intentions.

EBITDA margin — Earnings before interest, tax, depreciation and amortisation, as a share of revenue — a rough gauge of underlying profitability. Starlink’s 63% is software-grade.

HBM (High-Bandwidth Memory) — The stacked memory bolted onto AI chips, and the bottleneck of the buildout. Sold out years ahead at Micron and SK Hynix.

IPO (initial public offering) — A private company selling shares to the public for the first time. SpaceX’s $75bn ask is the largest ever.

RPO (Remaining Performance Obligations) — Contracted revenue not yet delivered — the backlog. Oracle’s grew $85bn in one quarter, to $638bn.