wangvc
Home
10-3110-3010-2910-2810-2710-2610-2510-2410-2310-2210-2110-2010-1910-1810-13All
2026-10-13

wangvc Inflection Daily · 2026-10-13

3m 18s · English narration

Today in three sentences

  1. AI leaders' share prices have already recovered from the 10-08 selloff: NVDA closed at $228.52 (+0.97%), ORCL closed at $143.20 (+4.23%) (10-13 close, Yahoo Finance); but the same day, the MOVE Index (which measures Treasury price volatility — think of it as the "instability of borrowing costs") stood at 101.09, above the prior session's close of 98.47 — equities have bounced back first, bonds haven't yet.
  2. All 3 of 3 lights are on for private credit: BDC median discount -15.2% (10-13), CCC-BB tiering spread at 1,065bps (10-12), and 5 of the past 30 days' 9 redemption-related news items hit pro-rata gate limits; meanwhile the public high-yield composite OAS (spread) is only 3.11% (10-12) — the repricing of the weakest borrowers is happening on the side without daily pricing.
  3. Spot BTC ETFs saw a cumulative net outflow of $503M over the 12-day window, with only 6 of the 12 days positive (10-13), versus 10-02 when 10 of 12 days were positive with $3bn in net inflows; over the same period BTC at $83,418 is only 2.8% below the 10-05 high — buying has already pulled back, but price hasn't caught up to that shift yet.

Main thread: AI buildout → cost of capital (Day 17, since 09-27) · Today's change: share prices recovered but capital didn't follow — NVDA +0.97%, ORCL +4.23% (10-13), while MOVE rose from 98.47 to 101.09 the same day

The one key difference between today and yesterday: share prices have largely recovered from the 10-08 selloff (NVDA +0.97%, ORCL +4.23%, 10-13 close), but two financing-cost indicators haven't eased in tandem — MOVE rose from the prior session's 98.47 to 101.09, and CCC-rated (the weakest public-market credit tier) yields at 12.53%, up a cumulative 1.77pp over 20 days (10-12), are still climbing.

Two earlier signals in this chain: 1 Hyperscaler debt issuance — over the past 90 days, AMZN filed 3 SEC 424B prospectus supplements and GOOGL filed 3 (the two most recent on 9-11 and 8-07), while ORCL, META, and MSFT each filed 0 — the pace of new borrowing for buildout is not synchronized among the leaders; 2 Electricity prices — CPI electricity +3.98% YoY, PPI commercial electricity +1.11% (both through August, monthly data, for background). Both of these move ahead of rates; right now one is rising and the other is roughly flat.

Where this goes: issuing debt to build data centers → competes with Treasuries for the same pool of money → long-end rates rise; buildout pushes up electricity prices → inflation won't come down → real rates stay elevated (10-year TIPS real yield 2.91%, 10-09, 21-day +36bps, for background) → borrowing the same money gets more expensive → the buildout ends up stepping on its own brake — this chain bites its own tail. The counterargument deserves equal weight: AI spending is largely counted as capital investment and doesn't flow directly into CPI; electricity price increases are fairly regional (concentrated in a few grids with dense data-center clusters); and over the long run, AI-driven productivity gains should actually weigh on inflation. But these three points describe the long run and the average, while rate pricing is about the here-and-now and the margin. The conditions for a thread change haven't appeared: neither real rates nor weak-credit spreads have reversed.

Private credit's three lights

All 3 of 3 lights on, unchanged from yesterday. 1 BDC(Exchange-listed private credit vehicles) median discount -15.2% (10-13, all 7 have latest NAV data), FSK -41.5%, PSEC -70.8%; 2 The semi-liquid fund exit is narrowing: Over the past 30 days there were 9 redemption-related news items, 5 of which hit pro-rata gate limits (the worst month was September 2026 with 6); Ares ASIF, Apollo ADS, and Blue Owl OCIC/OTIC have been oversubscribed for several consecutive quarters, with secondary tender offers priced at a 13%-18% discount; 3 CCC-BB tiering spread 1,065bps (10-12), CCC 20-day +1.77pp still widening, BB 5-day -0.05 has flattened out. Three independent sources point to the same thing. Yet the public high-yield composite OAS is only 3.11% (10-12, 20-day +0.46) — such a low composite spread with all three lights on shows risk has moved to where there's no daily pricing.

Twelve domains

AI — NVDA at $228.52 (+0.97%), ORCL at $143.20 (+4.23%) (10-13 close), the 10-08 selloff has fully recovered; industry score 87.5, market score 65.0 — industry enthusiasm still runs ahead of the price the market is paying. Two risks previously flagged as "not yet observed" were confirmed today: NVDA is providing up to a $105bn guarantee for OpenAI's Ohio data center (a supplier funding its own customer, with the circular flow estimated at $745bn, inkl.com); after the SEC eased ABS (asset-backed securitization) rules, annual issuance is projected to rise to $30-40bn — buildout funding is shifting from equity to structured debt.

BTC — BTC at $83,418 (10-13), -2.8% from the 10-05 high; on-chain status remains EXPANSION (not a bottom signal), but the score has slipped from +0.85 on 10-11 to +0.54 — momentum is fading, but the regime is unchanged. The confirmation framework still holds 2 of 3: closing price above short-term holder cost basis for 49 straight days (a 15.2% gap, as of 10-06), long-term holder profit multiple at a 14-day median of 1.18 (10-06, versus just 0.86 in early September); the item that's failed is ETF flows. The macro regime monitor meets only 1 of 4 conditions (gold below $4,300 for 12 straight days), and the 2-year Treasury at 4.80% hasn't broken its prior high of 4.92%.

Gold — Spot at $4,160.83/oz (10-13, TickDB), -0.23% on the day, +1.28% over 5 days, down 24.5% from the YTD high of $5,510.73 — getting back to that high from the current price would require a further 32.4% gain; the decline and the round-trip aren't the same number. The current drag comes from real rates: the 10-year TIPS real yield is 2.91% (10-09, 21-day +36bps), while the 10-year breakeven inflation rate is roughly flat at 2.33%. The floor comes from central banks: China's gold reserves stood at 2,409.59 tonnes at end-September, up 23 tonnes MoM, the 23rd straight month of additions (released 10-07 by the State Administration of Foreign Exchange), but gold is only 8.5% of official reserves — this is a two-decade-scale reserve reallocation, not a reason for the current price action.

A-share — STAR 50 closed at 1,427.18 (10-13, +0.03% on the day), down 9.06% over the past 5 days and -35.36% from its 52-week high, with thematic ETFs broadly down 9%-11% (the STAR chip ETF weakest); breadth is even worse — of the 50 constituents, only 10 are above their 200-day moving average and just 4 above their 50-day. But the reverse-confirmation signal of "domestic substitution being disproven" has not triggered: on 10-12, Hua Hong signed agreements with NAURA, AMEC, ACM Research, and Huahai Qingke covering deposition/etch/clean/CMP; on 10-08, SMIC and others put out over CNY 12bn in Q4 tenders with domestic suppliers winning over 60% of bids. Global semiconductor sales in August were $15.97bn, up 144.3% YoY (10-05, SIA basis) — it's valuations that are falling, not orders.

Bonds — 10-year at 5.24%, 2-year at 4.80% (10-09), 2s10s spread +44bps, near the upper end of its 20-day range; the long-end rise is driven mainly by real rates rather than inflation expectations (TIPS 20-day +0.36pp, breakeven inflation actually down 0.07). Three auctions beat demand expectations: 3-year at 4.932% (2.62x bid-to-cover, 10-06), 10-year reopening at 5.300% (2.77x, 10-07), 30-year at 5.618% (2.54x, indirect bidders 72.3%, primary dealers took just 6.8%, 10-08), easing worries about a lack of buyers for long-dated debt (CNBC 10-07). MOVE at 101.09 (10-13), above the prior session's 98.47, still oscillating in the elevated range below the 10-05 high of 113.60.

Congress — Disclosed filings total 16,718 (as of 10-13; House 3,355, Senate 1,688), with the latest batch stalled at the 41 filings from 09-24/09-25 (captured 10-09), of which Salazar accounted for 39, covering tickers including GLW, CSCO, SNOW. Topping the Sentinel list is MSFT (score 13, 13 trades from 3 members including 4 options trades, window 08/23-10/07); McCormick's 10 trades via Goldman Sachs structured notes, and power-sector trades (DelBene, Hern), were all one-directional buys — consistent direction says more than trade count. Paper filings total 61 cumulatively, with Harshbarger's 14 the most.

Biotech — XBI closed at 147.33 (10-13), down 6.5% from 9-14, IBB down 2.0%; in A-shares, the SW Pharma & Biotech index fell 1.0% and biologics fell 1.7%, with the equal-weighted decline exceeding the cap-weighted one — small-cap drugmakers are weaker. Two events: Pfizer's Tukysa combination regimen won approval on 10-08 for HER2+ breast cancer maintenance therapy, and AbbVie's Temab-A received 2 breakthrough therapy designations the same day (BioPharmaWatch). The structure is unchanged: biosimilars have outnumbered first-in-class molecules among new FDA BLAs since 2024 (24 vs 19; 22 vs 17 in 2025), and the phase 1-to-phase 3 conversion rate for cell therapy is just 2.1% — the lowest of any modality. Generics continue to crowd out first-in-class in the data.

Cybersecurity — The sector's biggest pricing gap persists: CRWD trades at 40.7x EV/Sales (enterprise value/annual revenue) versus ZS at 8.9x, a 4.6x gap, even though the two posted almost identical YoY revenue growth in their latest results (25.8% vs 25.4%). On 10-13, ZS reaffirmed full-year revenue guidance and announced an IBM/Red Hat partnership, rising about 6%-8%; CRWD fell about 4.4% to near $263; PANW rose about 5% (Yahoo Finance). Among 12 A-share network security companies, median H1 revenue growth YoY was just +1.2%, while network equipment maker Unisplendour posted revenue of CNY 63.5bn (+33.9%) with net profit doubling to CNY 2.167bn — money is flowing into infrastructure, not security budgets.

Energy — Grid equipment is up a median 51.85% YTD (GEV/PWR/AMSC) versus solar manufacturing down 29.81% to 41.63% (10-13, TradingView) — opposite ends of the same power chain moving in opposite directions. The biggest anomaly is in nuclear: share prices down a median 50.62% (OKLO -57%, SMR -55%), even as uranium sits near a 19-year high — the UxC/TradeTech indicator at $89.70/lb (10-13) corroborates Cameco's spot price of $89.63/lb (end of September) (the +17.45% YoY figure is from July data, for background); fuel is tight yet the stocks have been repriced — what's missing is financing cost, not demand. Independent energy storage demand (excluding purchases between related parties under common control) was 10.93GWh over the past 6 weeks and 21.54GWh over the past 90 days (10-12), with related parties accounting for 33.6% of the 90-day figure in dollar terms — discount the total order figures accordingly.

Emerging — Thematic ETFs are diverging and broadly weaker versus 10-11: QTUM is up 32.1% YTD, CIBR up 49.3%, while BLOK is up just 1.3% YTD and down 7.8% over the past month (10-13, TradingView). Total stablecoin market cap is $312.8bn, and RWA (real-world asset) tokenization TVL is $4.09bn (10-13, DefiLlama). In humanoid robotics, Morgan Stanley raised its long-term market size estimate to $5 trillion on 10-12, versus Goldman Sachs' earlier $138bn — a 36x gap, which signals the narrative is being priced, not the demand (startup funding YTD of $8.6bn is an 8-09 reading, for background). Two regulatory items the same day (10-13): the CFTC's new crypto asset rules were published in the Federal Register (comment period through 12-12), and the Fed and other central banks are asking banks whether their AI models can be "switched off instantly."

Venture — Most proxy assets weakened: SpaceX at $156.89 (-4.00% on the day, +4.59% YTD), Figma at $22.17 (-41.66% YTD), Circle at $84.30 (+4.27%), Fermi at $3.85 (+18.31% on the day but still -52.47% YTD, backed by an AMD credit guarantee, Benzinga); the drag came from a chip selloff (Samsung -7.4%, SK Hynix -10.1%, 10-13). Billion-dollar-plus rounds totaled 64 deals YTD worth $409.55bn, with AI accounting for 89.7%, but Q4 has seen just 2 deals worth $13.78bn (DeepSeek roughly $11.93bn, Ramp $1.85bn, Bloomberg 10-09) — large primary funding has clearly slowed. Anthropic's S-1 still returns 0 results on EDGAR as of 10-11, making the 11-09 roadshow timeline tight.

Winners — Two Hong Kong-listed AI stocks weakened in tandem: Zhipu (2513.HK) at HKD 601, down 2.91% from HKD 619 on 10-12 (market cap approx. HKD 301.8bn), MiniMax (0100.HK) at HKD 212 (-0.93%, with an intraday drop of over 4% at its worst); 21st Century Business Herald attributed this to Intel and Broadcom CEOs' warnings about AI financing risk, not company-specific bad news. The anomaly is in Concentrix (CNXC): closed at $26.7-27.0, up about 5%-6% from the prior $25.47, but volume was only 4% of the daily average (MarketBeat) — a low-volume rally, and the Q4 guidance of "another 3%-5% decline from AI automation" was not withdrawn. The primary-market pipeline held steady at 33 deals worth about $64.164bn with no new rounds; the Sentinel's 16 chokepoints were unchanged (7 tightening, 4 newly identified, 4 loosening, 1 on watch).

Cross-dashboard links

Signals worth watching next

  1. 10-year TIPS real yield: Currently 2.91% (10-09). Breaking above the 120-day high of 2.95% would move the macro regime monitor from 1/4 to 2/4 and keep the main thread intact; only a pullback below 2.70% would count as the first real easing in this cost-of-capital chain.
  2. CCC-BB tiering spread: Currently 1,065bps (10-12). A further 100bp widening, or the high-yield composite OAS breaking above 3.60% from 3.11%, would mean private-side pressure is spilling into public markets; narrowing below 950bps would be the first piece of reverse evidence. An earlier signal in the same direction: ORCL, META, and MSFT have each filed 0 424Bs over the past 90 days — any one of them filing even 1 would signal renewed leverage on the buildout side.
  3. Spot BTC ETF 12-day window net flow: Currently -$503M, with 6 of 12 days positive (10-13). Returning to 10 of 12 days positive would restore the confirmation framework's second condition; if net outflows widen to -$1bn, the framework would drop from 2/3 to 1/3.
This page compiles and analyzes publicly available information,Not investment advice; see methodology, sources, and disclaimer atTerms of Use。