Annotated earnings. falsifiable calls. FCF first. Not investment advice.
Joined September 2021
- Tweets1.7K
- Following126
- Followers192
- Likes693
Pinned Tweet
There is something you need to know.
I still expect new index highs by year end.
Friday was sold as the AI trade cracking.
Read the segments: every AI business grew. Each company failed on its own non-AI business.
• $GOOGL Alphabet: Cloud +103%. Problem: Search.
• $AMZN Amazon: AWS +39%. Problem: product sales against a Prime Day comp.
• $INTC Intel: data center +78%. Problem: PCs.
That's alpha (company-specific), not beta (the AI trade). Index investors still feel it, because these names sit in the index. The fix: underweight the weak segments, overweight the AI exposure that's working.
We’re not done yet.
Not investment advice.
$SPY $QQQ
Man discovers a one-liner and skips the homework. 2027 memory pricing is already contracted at Micron, SK hynix and Samsung, and Micron's new deals are take-or-pay, so customers pay for the volume whether they take delivery or not.
A 3x forward P/E is the market pricing in a bust that no contract has on the calendar, but sure, keep calling AI demand a PC cycle and let the people who read filings buy what you're laughing at.
$MU $SNDK $SKHY
@TradexWhisperer
Negus are quite offended nowadays.
Can’t take a joke, but can load up $NBIS above $200 and $IREN above $35. And at the same time telling everyone to do the same.
Right, @growthrapidly?
Retail buying $NBIS above $200 just realized that they were an Exit Liquidity.
Be proud of that and go "discount-cost average" into "generational" opportunity a bit more.
$IREN did save the markets though (jk, it was expected).
$NBIS: Why do people love to gamble?
$NBIS at $206.95: AVOID. My number is ~$126. Nebius passes the test $IREN failed (funding). It fails the one IREN passed (price).
The business is real. Q2 revenue $582M, +454% y/y. Annual run-rate $3.0B at Jun 30. Signed backlog $37.5B. Microsoft's deal (up to $17.4B over five years) has every tranche delivered. Meta signed $12B of dedicated capacity starting early 2027, plus up to $15B more if Nebius can't sell that capacity to others. Q2 deals priced above $20M of annual revenue per MW.
The cash is customers' money. Q2 free cash flow: −$3.4B ($2.25B operating cash minus $5.66B capex). Take out $1.2B of customer prepayments and it's −$4.6B. Last 12 months: −$5.9B. The headline $236M adjusted EBITDA (41% margin) is a $176M GAAP operating loss (−30%) before $260M of depreciation, $103M of stock pay and $50M of deal costs. Stretching server lives from 4 to 5 years also cut Q2 depreciation by $43M.
Funding, the IREN test. 2026 capex guide: $20–25B, $8.1B spent in H1. That leaves $12–17B, against ~$19B lined up: $8.0B cash (Jun 30), $5.75B of August convertibles, a $775M GPU-backed loan, and at least $4.45B more prepayments. 2026 is funded, and no debt matures before 2029. 2027 isn't funded yet: they plan to add more than 1GW a year and haven't given a capex guide. My rough math says $25–35B.
The bill is dilution. 238.7M shares in Jun 2025. ~308.7M today, counting NVIDIA's $2B of warrants and the August note swap: +29%. Another ~50M sit in convertibles that are in the money at this price. Three customers were 59% of Q2 revenue.
The price. At ~$63.9B, the market cap is 9.5x one year of signed revenue ($6.7B, from the backlog note). IREN was at 3.5x. On enterprise value, Nebius trades at 5.7x 2027 consensus revenue ($12.3B). CoreWeave trades at 3.4x. To earn CoreWeave's multiple at $206.95, Nebius needs $20.8B of 2027 revenue. That's above the highest of 22 analyst estimates ($18.0B).
My number, ~$126:
→ 2027 revenue $12.3B (consensus; ~55% already signed)
→ × 3.37, CoreWeave's enterprise value ÷ its own 2027 consensus today
→ = $41.5B, minus ~$1.1B net debt (at $126 most of the convertibles stay debt) = $40.4B
→ ÷ 321M shares = $125.86
No discount, because it's today's peer price on the same year. Treat it as an end-2027 value at 15.2% (10-year 5.22% + 10 points) and it's ~$106.
Margin of safety: none. My bar is 25%, so I'd buy at $94 or below. If you own it, my trim line is $157, and the stock is above it. Only 1 of 9 cells in the grid clears today's price: the top estimate ($18.0B) at 5.0x, a 50% premium to CoreWeave.
What would flip me: at this price my number would have to reach ~$276. Even the top estimate at 5.0x gets $260. So the price has to come to the value, not the other way round.
What I'm watching:
① Q3 revenue ≥ $900M (mid-Nov, date unconfirmed). Consensus is $905M. Coin-flip.
② Sep 30 share count ≤ 290M. Above that, the stock-sale program is back on.
③ Year-end run-rate ≥ $8.0B, the guide midpoint (Feb). That's 2.7x June's in six months.
We’re not done yet.
Not investment advice.
The market was completely irrational with $AMZN. I am still bullish and my 12-month target is >$300.
Chart still looks great and fundamentals support the story.
We're not done yet.
Not investment advice
Who else is bullish $GOOGL?
* This week's candle swept liquidity as expected
* Price held 100 EMA
* $GOOGL, $META, $AMZN have been relatively strong given yesterday's sell-off
Same applies to $GOOGL this week. Added to my existing position. So far this red circle acts as a blackhole for the price (purely because I said it will).
If this TA analysis works, I am becoming a full-time TA degen.
Jk, I use TA sometimes when I want to find a good entry on a price ONLY if my fundamental analysis agrees for long-term hold.
$AMZN $SNDK
DONT YOU KNOW, PUMP IT UP
YOU’VE GOT TO PUMP IT UP
DONT YOU KNOW, PUMP IT UP
YOU’VE GOT TO PUMP IT UP
DONT YOU KNOW, PUMP IT UP
YOU’VE GOT TO PUMP IT UP
REPEAT AFTER ME…
Canada's Budget 2026: real investment ideas, no discipline. Ottawa got a C$158 billion forecast windfall and spent C$151.5 billion of it.
THE GOOD
• Productivity Mega-Deduction: instant write-off on new investment, tax rate on it cut from 13% to 6.4%, cost C$36 billion over 5 years - best productivity lever in years
• Power: over C$50 billion to double the grid by 2050, plus C$2.3 billion for critical minerals
• Immigration held at 380,000 permanent residents a year, temporary residents capped toward 5% of population
• Tax filing simplified: one May 15 deadline, instalment threshold C$3,000 to C$7,500
THE BAD
• Deficit C$62.3 billion in 2026-27, 1.8% of GDP, C$62.5 billion next year - no balance in any year, federal debt C$1.37 trillion rising to C$1.59 trillion by 2030-31
• Debt charges C$60.2 billion this year, rising to C$85.5 billion - fastest-growing line
• Before new measures, the 2027-28 deficit was C$24.8 billion. Ottawa spent the windfall instead of banking it
• Operating budget "balanced" in 2027-28 only by adding back C$16.8 billion of forgone tax revenue absent from the financial statements
• Housing: C$44.7 billion over 11 years, mostly non-market - too slow to move prices
• Small sprawl: C$300 million for Canada 160, C$1.5 billion for rinks, C$300 million more for CBC
Bottom line: buy the investment agenda, not the fiscal story - debt-to-GDP stuck near 40%.
$IREN Q1 scorecard: both my lines missed narrowly; the buy case stands.
• Revenue $228M vs my $230M line (consensus $221M). AI cloud revenue $200M, up 182%.
• 10-Q cover shares 424M vs my 410M line. Sep 30 was 407M, inside; October sales at ~$36 a share put it over.
• Backlog $20.3B, up 22%. New post-quarter deals: one 5-year contract has 90% of GPU capex prepaid. Horizon 1 is billing; Horizons 2-4 slipped to January on Microsoft's own design change.
• Lines hold: contracted book unchanged, new shares sold above my number. BUY stands: fair value ~$31, buy line $23.21. Close $18.03, 22% under the line; signed contracts cover 129% of it.
We’re not done yet.
Not investment advice.
$NBIS
How many jobs have you created? How many products have you shipped?
Pausing AI is easy when the answer to both is zero.
Stupid.
@elonmusk
He went on to argue that even if 2031 operating profit were certain, we would still have to account for 2032 operating profit.
He went on to argue that even if 2033 operating profit were certain, we would still have to account for 2034 operating profit.
He went not to argue that even if 2042 operating profit were certain, we would still have to account for 2052 operating profit.
$MU $SNDK
Today I met with a veteran investor who argued that memory stocks are not cheap even at a P/E of 3. Below is an edited version of our conversation.
Memory LTAs are basically meaningless. The best the LTAs being signed today can do is get customers to shoulder part of the memory makers' capex. Once prices fall, customers will gladly forfeit their deposits (prepayments) and walk away from the LTAs.
Historically, whenever prices fell more than 60% from the peak, customers broke their LTAs even after signing them.
So even if a 2027 shortage is certain and the stocks trade at a P/E of 3 on 2027 operating profit, 2028 operating profit is uncertain, which makes it hard to justify a higher valuation.
He went on to argue that even if 2028 operating profit were certain, we would still have to account for 2029 operating profit.
I pushed back and asked why the market doesn't discount TSMC the same way. Why does TSMC get such a high valuation? Wouldn't TSMC also take a big hit if the AI boom fades?
His answer: the problem was how violently memory prices rose in the first place.
If memory makers had raised prices in a healthy way like TSMC, it might be a different story. But because they hiked prices like a commodity, the market has no choice but to treat memory stocks like commodity stocks.
Even if memory makers stay in shortage through 2030, he said, the market will price in the risk of oversupply in 2031.
He closed his argument this way: even at 3x 2027 earnings, if memory prices fall in 2028 by enough to break LTAs, that 3x multiple on 2027 works out to 7.8 to 10x on 2028. So current valuations are not cheap.
$MU at $690 is priced below my worst-case value for it. Still a BUY.
• Price: $690.24 (12:14 PM ET Nov 5). That's -13% today and -30% since the Oct 28 close of $983.23. The chip index is -10% today.
• My number holds at $1,360: FY27 earnings per share of $170, with stock pay counted as a cost, times 8. FY27 is the year we're in, so there's no discount. The price is 49% under it, and my bar is 25%.
• Buy below $1,020 (75% of my number): unchanged. New trim line: $1,510 (best-case FY28 EPS ~$205 x 8, in today's dollars).
• At $690 the stock trades at 4.5x Micron's own guide for this quarter, annualized ($38.15 a share). At my 8x it needs FY27 EPS of $86, 43% below that pace.
• The grid: all 9 values are above the price, the lowest at $828. Each cell is a late-2027 price, discounted 0.82 years at my 10.9% cost of equity (the 10-year at 4.90% plus 6 points).
• What $690 implies: my bust case. FY28 prices per bit halve, matching the 10-K's worst year in five, and the price is ~8x that case's FY28 EPS ($87).
What flips it, at the December print:
① FQ2 revenue guide below $66.4B.
② FQ2 gross-margin guide below FQ1's actual.
Either one, and the BUY goes back to HOLD while I rebuild.
We’re not done yet.
Not investment advice.
Is memory structural now? For 2027, Micron's contracts say yes. Past that, I can't prove it, and 2026's boom was mostly price.
• The cyclical tell: FY26 DRAM sales rose 252% on prices up ~180% and bits up ~25% (10-K). Price did ~80% of the work. The 10-K's own five-year range for yearly price changes runs from +200% to -50%.
• The structural case: the take-or-pay contracts carry minimum and maximum price bands (10-K). Over 75% of FY27 shipments are committed (call), and customers put down $12.7B of cash deposits last year.
• Keep it in proportion: the filed contract book is $134B, with ~$33.5B due in the next 12 months. That's ~12% of my FY27 revenue under contract floors. It's a floor under a slice of the business, not the whole of it.
• Peers agree on 2027: SK Hynix says its DRAM and HBM are effectively sold out through 2029 (call). Samsung says 2027 demand "will far exceed the industry supply." 2027 HBM is priced up at both Hynix and Micron.
• Where my model was too kind: I had 2027 industry DRAM supply growing 17.8%. Hynix's call puts 2027 industry growth in the mid-20s. In my model that version ends the shortage in 2027, not 2031.
• New fabs: Micron's first Idaho wafers come mid-2027; its second Idaho fab and Hiroshima, late 2028. All three makers raise capex in 2027.
My read: 2027 pricing is contracted. 2028 is open. It's a damped cycle, not no cycle.
What breaks $MU, and what a break costs if $MU is most of your portfolio.
• Customers: the top ten are over half of revenue. No single one is 10%+ (10-K).
• Price caps: Micron's largest contracts cap prices near spring-2026 levels. Floors cut the downside, and ceilings cut the upside.
• China: China and Hong Kong customers were 11% of FY26 revenue. Management sees China in single digits in FY27. The 2023 Chinese purchase ban still stands.
• Execution: most DRAM output comes from Taiwan. The Taichung union settled Oct 22, but the Taoyuan union's profit-share demand is still open.
• The cushion: no bond comes due before 2032, and after debt and refundable customer deposits there's ~$49 a share of cash left.
• The arithmetic at $690: the stock is already down 30% since Oct 28. If the market prices my bust case (FY28 prices per bit halve) at 6x, that's $481, another -30%. For a holder with everything in it, that's -30% of the portfolio. At a 5% weight it's -1.5%. The same falsifiers carry far more weight in a concentrated book.
• Checks: the December guides; in late January, Hynix and Samsung 2027 supply plans and hyperscalers' 2027 capex; and in January, whether the 10-Q's contract book holds $134B.
Vistra $VST closed at $120.29 Wednesday, at my $120.40 buy line. The price moved, not my number: BUY.
• My number: ~$161 = 2027 free cash flow to common, $13.38 a share, x 12. 2027 is today's value, no discount.
• Margin of safety: 25.1% vs my 25% bar. At the line, not past it.
• Buy under $120.40. Trim above ~$179.
• Changed: Cogentrix closed Oct 30, with $0.1B more cash and debt than the deal terms. My number: -$0.20.
• The 13.8% drop came with a KeyBanc downgrade on Texas data-center risk.
• Method holds: cash flow x 12. Talen's new 2027 guide puts it near 9x. At 9x my base is $120.40, today's price.
We’re not done yet.
Not investment advice.
The election doesn't reach Vistra's 2027 cash. Texas data-center policy can reach 2028.
• Nuclear tax credit: no channel. Repeal needs a law and a signature. The 2025 tax law left it largely intact.
• Texas: bites, but from 2028. New data-center hookups are paused since August, maybe into January's session. 94% of 2027 output was hedged in August.
• Rates: supply, not politics. A 10-year above 5.13% by end-March costs my number ~$3 a share.
• FERC, EPA, tariffs: executive calls. Headline only.
• AI demand: watch hyperscaler 2027 capex guides, late Jan-early Feb.
Friday's tests, from Oct 23: "Q3 ongoing EBITDA ≥ $1.87B keeps the 2026 midpoint; a 2027 organic guide under $7.4B cuts my number (~$3.60 per $100M); Cogentrix closed or a firm date; $Meta deliveries started."
• Cogentrix: met, closed Oct 30.
• New: a 2027 guide even $0.1B under $7.4B drops my buy line to ~$118, under today's price. HOLD.
$VST
Alphabet $GOOGL doesn't have an AI demand problem. Its own usage numbers say so.
• Its AI models via its developer API: nearly 32 billion tokens (chunks of text) a minute, +40% vs last quarter, 4x+ vs a year ago.
• The Gemini app reached 1 billion users.
• New models: Gemini 4 Argon (October) and Carbon, its strongest, released first to vetted cyber defenders.
• YouTube Music and Premium: 150M+ subscribers, trials included.
• The gap is money: Search revenue +12% vs last year, and the CFO sees Q4 Search growth a couple of points lower on a tough comparison.
We’re not done yet.
Not investment advice.