The AI trade has quietly shifted from cash-rich giants to companies funding buildouts with borrowed money, and SoftBank's $11.1 billion bond issuance for OpenAI is the clearest sign of that. Borrowing gets more expensive and harder to roll over as the bon

Verdict: Half the pair is live, half is still arming — wait before committing

The thesis has real teeth: CoreWeave shows exactly the debt-funded strain the idea targets — $14.7B of long-term debt against $3.3B of equity, a current ratio of 0.46, negative free cash flow of $7.3B in fiscal 2025, and Q1 2026 margins moving the wrong way (gross margin down 2.1 points to 65.5%, net margin down 6.9 points to -35.6%) with shares up 5.9% in one quarter. The strongest point for the trade is that the CRWV technical entry is fully live today: price at $87.59 is $0.62 below its 50-day EMA of $88.21, with momentum, 21-day rate of change, and trend strength all confirming. The strongest point against is that the realized evidence rests on a single 12-month backtest window (39.8% return, 16 trades, 13.2% max drawdown) with daily-bar exit fills and no completed 24- or 60-month evaluation — one good year, not a multi-regime track record. SoftBank's own fiscal 2025 results actually cut against the short: net income swung to ¥1.15T, net margin reached 15.9%, and ¥3.7T of cash gives it options a pure leveraged player lacks. And its leg isn't triggered — price at $19.75 sits $1.01 above the $18.74 EMA with 21-day rate of change still at +20.9% and ADX at 11.7 versus the required 20. The verdict flips bearish-tactical only if SoftBank closes below roughly $18.74 with momentum and trend strength confirming, or if CoreWeave's next quarterly report shows the margin slide accelerating.

Trade now: one leg is live, one is waiting

This is a two-leg setup and only one leg is ready. On **CRWV** ($87.59), all four entry conditions are met right now: the price sits $0.62 below the 50-day EMA of $88.21, 10-day momentum is negative at -$1.40, 21-day rate of change is -0.48%, and the 14-day ADX of 20.5 clears the 20 trend-strength floor. The third entry route — a close below the nearest support level at $87.17 — needs only $0.42 of downside. On **SFTBY** ($19.75), the trade is not triggered: momentum is the only condition met, while the price sits $1.01 *above* its 50-day EMA of $18.74, the 21-day rate of change is +20.9% versus a need for it to be negative, and the ADX of 11.7 is well short of the required 20. 'Wait' on SoftBank means a close below roughly $18.74 with momentum turning negative and trend strength rebuilding above 20. Risk parameters are fixed by the rules: an 8% stop and a 16% take profit on each position, a 2:1 reward-to-risk profile, capped at 25% per position with 2.4% account risk sizing. On CRWV at $87.59, that means an effective stop zone near $80.58 and a target zone near $101.6. The exit signal — a close back above the 50-day EMA after 60 bars of holding — is already marked 'near' on CRWV because the price is only $0.62 under that line, so any entry should be expected to live or die quickly by that level. The evidence base for this configuration is the completed 12-month backtest on the CRWV pair: a 39.8% return over 16 trades with a 56.3% win rate and a 13.2% maximum drawdown. The exits were filled on daily trigger bars rather than intrabar data, so treat stop and target fill quality as approximate. No robust parameter setup was established — the parameter search could not run on incomplete data — so the frozen rules above are what you trade, not an optimized variant. The idea's bearish macro thesis (debt-funded AI infrastructure names get hit hardest as financing costs rise, per the Citigroup rates commentary cited in the idea) aligns directionally with the CRWV signal being live, but note the strategy itself trades these names long on weakness with the crossover entry — so the deepest value add today is the SFTBY leg arming, not the CRWV leg firing twice.

The Debt-Funded AI Short Has the Balance Sheets on Its Side

The thesis is straightforward: when financing costs rise, the most leveraged names fall hardest, and CoreWeave (CRWV) is exactly that name. The fiscal 2025 balance sheet shows $14.7B of long-term debt against just $3.3B of equity, with total liabilities of $46.0B against total assets of $49.3B — a current ratio of 0.46, meaning current liabilities ($16.4B) far exceed current assets ($7.5B). Free cash flow was negative $7.3B in fiscal 2025, and the deterioration continued into Q1 2026 with free cash flow of negative $4.7B for the quarter. Capex of $10.3B against operating cash flow of $3.1B in fiscal 2025 tells you this buildout is not self-funding. This is precisely the 'borrowed money AI infrastructure' profile the idea targets. The margin trend is moving the wrong way at the worst possible time in the rate cycle. Gross margin fell from 67.6% in Q4 2025 to 65.5% in Q1 2026, a 2.1-point drop quarter over quarter. Operating margin worsened from negative 5.7% to negative 6.9%, and net margin deteriorated from negative 28.7% to negative 35.6%. Return on equity fell to negative 15.5%. Meanwhile, shares outstanding grew 5.9% in a single quarter, from 502.4M to 532.0M — dilution is doing part of the financing work, which typically accelerates as debt costs rise. On the SoftBank side, the Reuters report from September 24, 2026 confirms the $11.1B bond issuance for OpenAI — the literal event the thesis is built around. SoftBank's own fiscal 2025 numbers (period ending March 2025) show negative free cash flow of about ¥650.6B against operating cash flow of only ¥203.6B, so it too is extending on credit. The Bloomberg piece quoting a Citigroup rates executive that the bond sell-off 'isn't finished' supplies the timing catalyst: refinancing gets harder, not easier, from here. The rule-based strategy evidence supports a tactical bearish read. Over the trailing 12-month window, the completed backtest on CRWV produced a 39.7% return across 16 trades…

Scores

• Conviction score breakdown: 57
• Thesis support: 74
• Trade readiness: 55
• Risk quality: 52
• Backtest evidence: 44
• Fundamentals trend: 58

Watch items

• SFTBY — Close vs 50-day EMA
• SFTBY — ROC (21)
• SFTBY — ADX (14)
• SFTBY — Close vs nearest support
• CRWV — Entry conditions (EMA/momentum/ROC/ADX)
• CRWV — Close vs nearest support
• CRWV — Close vs 50-day EMA (exit signal)
• CRWV — Gross margin, QoQ change