TenK Shifu

$CRWV, $NBIS and $IREN have each lost about a third of their value since Oct 6. CoreWeave −33%, Nebius −33%, IREN −34%. CoreWeave's 15.5% drop on Oct 14 came after a selloff in its bonds: its 9.625% notes due 2032 were reported trading at a yield near 15%. Here's what didn't change: the businesses. Estimates for 2027 revenue barely moved: CoreWeave +0.2%, Nebius flat, IREN +1.1%. What changed is the yardstick and the cost of money. My $NBIS and $IREN numbers both use $CRWV's price per dollar of 2027 revenue. That went from 3.37x on Oct 13 to 3.05x today. Money stays expensive. $CRWV, the funding story: → Q2 revenue $2.58B, +112%. Backlog $103.7B, and 41% of it is due within 24 months. That's ~$21.3B a year, about 81% of 2027 consensus. Demand isn't the problem. → The debt is. At Jun 30 it had $35.1B of debt plus $16.5B of leases against $5.5B of cash. That's 1.5x its market value today. → Q2 interest expense was $640M, up 19% from Q1's $536M. That's 25 cents of every revenue dollar. → The $1.51B of adjusted EBITDA (59% margin) is a $49M GAAP operating loss after $1.39B of depreciation, and a $626M net loss after interest. H1 free cash flow: −$10.5B ($3.7B operating cash minus $14.1B capex). → $10.6B of debt principal comes due by the end of 2027. Every $10B it borrows at ~15% costs $1.5B a year. Q2's interest bill annualizes to $2.6B. → The cheap money is stock-linked. In September it sold $4.2B of 2.875% convertibles that convert at $97.85, and the stock is 38% below that. It also set up a program to sell up to 35M new shares, starting no earlier than Oct 18. That's $2.1B at today's price, 6.3% more shares. If the cost doesn't show up as interest, it shows up as dilution. $NBIS at $168.00: AVOID stands. My number drops to $114 from $126. All of that move is $CRWV's multiple. $NBIS's 2027 consensus ($12.3B) didn't change. → $12.29B × 3.05 = $37.5B, minus ~$1.1B net debt = $36.4B, ÷ 320.3M shares = $113.79 → Buy $85. Trim $142. The price is 48% above my number and 18% above the trim line. → NBIS trades at 4.54x 2027 revenue, 1.5x CoreWeave. To earn CoreWeave's multiple at $168, it needs $18.3B of 2027 revenue. That's above the highest of 22 analyst estimates ($18.0B). → It's down 19% since my AVOID at $206.95. It still hasn't reached my trim line. $IREN at $27.26: HOLD, and it has a level again. My number is $31 (was $36.39). Buy $23 (was ~$28). That doesn't make it a buy, because the level moved down: → CoreWeave's multiple went from 3.5x (Oct 8) to 3.05x: −$5.21 a share → More shares. The Oct 13 proxy shows 404.6M on Sep 28, up from 394.1M on Aug 14. That's +10.6M in six weeks: −$0.75 → 2027 consensus up 1.1% to $5.11B: +$0.38 My 25% margin-of-safety bar on $30.80 gives $23.10, the same place. At $27.26 the margin of safety is 11.5%. Range: $18.17 (contracted only at 2.5x) to $40.38 ($6.5B of 2027 revenue at 3.05x). Is a stressed CoreWeave still the right yardstick? Yes. The stress it prices is IREN's stress too. IREN plans $25–30B of capex this fiscal year against a ~$11.0B market value, and $3–8B of it isn't lined up yet. Here the stressed comp is the honest one. The market already prices IREN at 2.54x, 17% below CoreWeave. What I'm watching: CoreWeave, Q3 on Nov 11: ① Interest expense ≤ 25% of revenue (Q2: 24.9%) ② Backlog ≥ $120B (the >$25B it added in early Q3 should put it near $125B) ③ Any new unsecured debt priced at ≤ 12%. Closer to 15% means building just got more expensive. Nebius, Q3 around Nov 10 (date not confirmed by Nebius): ① Revenue ≥ $900M (consensus $905M) ② Sep 30 share count ≤ 290M ③ Year-end run-rate ≥ $8.0B (Feb) IREN: ① Q1 revenue ≥ $230M (~Nov 5, unconfirmed) ② 10-Q cover share count ≤ 410M. My Oct 8 line (≤ 400M) is already broken: 404.6M on Sep 28. ③ Microsoft accepts Horizons 2–4 by Dec 31 We’re not done yet. Not investment advice. $MU $SNDK $TSLA