wangvc Inflection Daily · 2026-10-28
Today in three sentences
- 10-year US Treasury yield at 5.028% (intraday 10-28, TradingView), while the bond market volatility index MOVE fell to 93.60 — Day 3 of declines. Volatility expectations are falling, but funding costs are not. Today's FOMC decision is the referee for this contradiction.
- The CCC-BB credit spread is 1134bps (10-27) — narrowing for 2 straight days but still elevated; the BDC discount rate median is -12.9%. Rate-level cooling has not passed through to the weakest borrowers.
- GE Vernova's order backlog is $197bn, up 96% (10-28); but Zhipu AI is still 24.93% below its 9-13 placement price — the order book is bought in, secondary-market pricing isn't yet.
Main thread: Rates Cooling ↔ Credit Stratification Unmoved (Day 2) · Today's change: MOVE at 93.60, down 4.4% on the day and lower for a third straight day, while the 10-year yield moved the other way, rising to 5.028% (+0.58% on the day); CCC-BB still at 1134bps
The two readings have diverged for the first time: the volatility side keeps easing (MOVE down from about 97 on 10-21 to 93.60), while the price side is tightening (10-year at 5.028%) — the cooling is coming from volatility expectations, not funding costs. Credit is moving the same direction but only slightly: CCC-BB narrowed for 2 straight days to 1134bps, with a BDC discount rate median of -12.9%. The 10-27 2-year auction stopped out at 4.708%, below September's 4.787%, so short-end demand held up; there's also a 7-year auction on 10-30. The evidence doesn't show both reversal conditions, so the main thread stands.
Private credit's three lights
3/3 lit, same as yesterday. 1 BDC discount rate median −12.9% (10-28); 2 Semi-liquid fund exits remain tight: redemption events 5 over 30 days, 16 over 90 days (window as of 10-28), on 10-05 Cox Capital's secondary tender offer for OCIC was priced at a 20% discount to NAV; 3 CCC−BB 1134 bps (10-27). The key counterpoint is that the public-market high-yield composite spread (OAS) is only 3.18% (as of 10-28) — the risk has moved to the side without daily pricing, not disappeared. BDC NAV is still from the 6-30 quarterly report.
Twelve domains
AI — Dashboard industry score 87.5 / market score 47.5, with VIX at 25.3 approaching the 28 warning line. The credit-side red flag remains unresolved: on 10-23 ORCL/SpaceX CDS spreads hit record levels (implied default probability of 20%/16%), and on 10-24 NVDA's $36bn guarantee faced its first stress test. The Anthropic S-1 disclosed on 10-26 showed H1 revenue of $16.3bn and a net loss of $80.7bn (mostly non-cash revaluation), with two customers accounting for 25% of revenue.
BTC — Current price $80,840 (10-28). On-chain accumulation count is 8/10, 2 days short of the 10-day threshold; the short-term holder realized price (average cost basis for holders under 155 days) is $75,825, and the current price is still above it. ETF flows were net positive on 7 of the last 12 days, with cumulative inflows of $425M, below the $1bn threshold; on 10-26 Strategy disclosed that 10-19 through 10-25 was its first full week without adding to its position. The macro paradigm is only 1/4 met (MOVE at 93.6).
Gold — Spot gold is $4038.42/oz (10-28), down 26.72% from its 250-day high (it would need to rise 36.5% to get back there). Real rates fell to 2.77% on 10-22, but the implied probability of December policy action rose to 84-85% and the dollar hit a two-week high, pressuring gold. Flows are moving the other way: gold ETFs added 67 tonnes in September to a record 4256 tonnes, and central banks added 23 tonnes in September to 2410 tonnes, marking 23 straight months of net buying.
A-share — The STAR 50 closed at 1282.93 (10-28), up 2.45% on the day but down 22.6% over the past 21 days, and down 41.89% from its 52-week high. Only 4% of constituents are above their 200-day moving average and 14% above their 50-day (weights as of 09-30), making a bottom hard to call. Domestic substitution hasn't been disproven: Changchuan Technology's net profit for the first three quarters was up 103.92% (10-26), and domestic equipment's market share rose from 16% to 21% in 2025 (10-17).
Bonds — See the main thread. Additional detail: electricity CPI +4.37% YoY (September), high-yield composite OAS at 3.18% (as of 10-28). Today's FOMC decision, and the 10-30 7-year auction.
Congress — Rep. Sessions reported selling more than 25 blue-chip stocks in a single day, including Apple, Microsoft, Nvidia and Tesla, for a total of $1.35M-$3.24M (filing date 10-27, House Clerk PTR — the House Clerk's public record of member trading disclosures), the largest concentrated sale in recent memory. The Senate's "ban lawmakers from stock trading" motion fell short 53-47 on 9-30, so the rules remain unchanged.
Biotech — XBI (the US biotech ETF) closed at $155.86 (10-28), still below its September high of 165-168; on 10-27 Artiva discontinued AlloNK and withdrew it from Phase III, leaving $319M in cash under review for a possible sale. The industry signal score is 34.8 (stale) and the market-layer score is 86.7, putting the read as "market stronger than fundamentals." Catalysts: Inovio's INO-3107 on 10-30 and Agios' mitapivat FDA decision on 11-01.
Cybersecurity — A-share network security stocks broadly rebounded on 10-28 (NSFOCUS +6.85%), an oversold bounce after the 10-27 Q3 earnings selloff, not a trend reversal. CISA has added PHP2Shell (CVSS severity score 9.2 out of 10) to its known-exploited list, and federal systems must remediate by 10-29. Valuations remain rich (as of 10-28, revenue is YoY from the latest quarterly report): CRWD revenue +25.8%, EV/Sales (enterprise value to revenue multiple) at 50.1x, higher than similarly-growing FTNT (+25.6%, 17.6x).
Energy — Grid equipment is the strongest leg: GE Vernova posted Q3 EPS of $4.81 (vs. $1.64 a year earlier), with a $197bn order backlog up 96% (10-28); Vertiv's net sales rose 61%, and both companies raised full-year guidance. Grid-equipment stocks are up a median 33.62% YTD, while solar manufacturing is down a median 40.78%. Natural gas is decoupling by region: Henry Hub is at $3.042 (as of 10-28), while European TTF broke above €100/MWh intraday for the first time since 2022. Energy storage saw 14.93GWh of independent demand over the past 6 weeks, with related parties accounting for 16.0% — demand still holds even excluding them.
Emerging — Thematic ETFs are diverging: CIBR is up 49.9% YTD, while BLOK is the weakest, down 14.0% over the past month. The AI chip chokepoint has shifted from wafers to TSMC's CoWoS advanced packaging (the process of packaging multiple chips with high-bandwidth memory onto one substrate) — Nvidia has locked up about 60% of its 2026 capacity, with relief unlikely before 2027. NIST released the final IR 8547 on 10-27: federal systems must migrate cryptographic algorithms by 2030. Note: NVD added 3718 new CVEs in the past 7 days, a 15.7% single-day jump from 3213 on 10-27 — the count is unstable, so patch prioritization should rely on in-house review.
Venture — Secondary proxies broadly declined (Fermi and Figma led losses). YTD billion-dollar-plus rounds total 67 deals / $427bn, with AI at 89.8%; but the pace is slowing: Q3 had 36 deals / $89.1bn versus Q4's 5 deals / $31.2bn, with none added in the past 7 days. Form D filings (private fundraising filings with the SEC) over the past 30 days total 545 / $11.72bn (as of 10-28), down from 602 as of 10-24; the drop in dollar amount is because large September filings rolled out of the window, not because new filings slowed.
Winners — Zhipu AI (HKEX: 2513) closed at HK$571.5 (10-28, +376.25% YTD), still needing to rise 24.93% to reach its 9-13 placement price of HK$714; MiniMax is at HK$180.2 (down 34.09% over the past month), needing to rise 48.72% to reach its 7-10 placement price of HK$268 — about half of Zhipu AI's recovery progress. DeepSeek's roughly CNY 100bn funding round gained further confirmation: Bloomberg reported on 10-21/22 that CATL and Tencent are the largest investors, at a pre-money valuation of about CNY 500bn; as of 10-28 no corporate registry change has been found. Of 16 chokepoints, 7 tightened and 4 loosened (as of 10-28); hyperscaler capex remains in "tightening" mode.
Cross-dashboard links
- Real rates ↔ gold:Real rates at 2.77% should be bullish for gold, but an 84-85% implied probability of December policy action plus a two-week dollar high are pinning gold at $4038.42. Meanwhile ETF holdings hit a record 4256 tonnes and central banks have bought for 23 straight months — price is driven by rate pricing while positioning is driven by reserve reallocation. The divergence itself is not a contradiction.
- AI Capex ↔ Credit Pricing:Capex YoY median is +100.1% (as of 10-28), with CoWoS capacity 60% locked up — primary demand is real; the cost shows up on the credit side, with ORCL/SpaceX CDS at record levels and NVDA's guarantees under pressure for the first time. Electricity CPI at +4.37% (September) has not turned, and this leading signal is still building.
- MOVE ↔ BTC Macro Paradigm:BTC macro is only 1/4 met, and the only improving metric is MOVE (93.6) — the same reading as on the Treasury dashboard. Bond market volatility is a leading variable on the BTC macro side, so on-chain accumulation leading and macro confirmation lagging is structural.
- Demand Concentration ↔ Supply Lock-in:In the AI dashboard's Anthropic S-1 (10-26), about $49.8bn (76%) of the $65.7bn in compute contracts is non-cancellable, and two customers account for 25% of revenue — supply-side lock-in and demand-side concentration are two sides of the same coin.
Signals worth watching next
- Can BTC clear both thresholds at once?:The accumulation count would need to go from 8/10 to 10/10, while cumulative ETF inflows would need to rise from $425M past the $1bn threshold. Meeting only one still counts as one-sided on-chain accumulation.
- Is credit stratification finally turning, for the first time?:CCC-BB needs to fall from 1134bps below 1100, and the BDC discount rate median needs to narrow from -12.9% to within -10% — both must hold at once for the three-light signal to flip; if OAS also widens from 3.18%, that would mean risk is moving back from the dark side into the open.
- Can the 10-year hold above 5%?:10-29 Q3 GDP advance estimate and September core PCE, the 10-30 7-year auction stop-out yield (vs. the 2-year's 4.708%), and 11-06 October nonfarm payrolls. If the stop-out yield comes in higher than last time and the 10-year holds above 5%, the "rates cooling" half of the thesis is invalidated, and the main thread must be rewritten to reflect both reversals.