S&P 500 close (Oct 6, 2026)
7,819
Record close; first time above 7,800; first record since mid-August
S&P 500 daily move
+0.58%
Second consecutive record close after Monday's print
10-year U.S. Treasury yield
5.27%
Down ~3 bps from Monday; touched 5.35% intraday (24-year high)
30-year U.S. Treasury yield
>5.10%
Multi-year high; levels last seen in the late 2000s
Top-10 weight in S&P 500
~40%
More than double the ~16% share a decade ago
Nvidia weight in S&P 500
>8%
Largest single-stock weight in the index's history
Hyperscaler AI capex 2026
~$725B
Combined Amazon, Google, Microsoft, Meta, Oracle; ~75% AI-attributed
Bank of America estimate: capex/OCF
~90%
Hyperscaler capex as share of operating cash flow in 2026, vs. 65% in 2025
The headline tape on October 6 looked schizophrenic. The S&P 500 closed at 7,819 — a fresh all-time high and the first close above 7,800 — while the 10-year Treasury yield briefly traded at 5.35%, the highest reading since April 2002, before easing to 5.27% by the bell. The 30-year Treasury touched above 5.10%, and the selloff was global: UK 30-year gilts cleared 6% for the first time since 1998, and Japan's 10-year JGB crossed 3% for the first time in roughly three decades. Markets that should move opposite to one another moved in the same bearish direction — bonds down everywhere — while U.S. equities rose. That contradiction is not noise; it is the story.
The signal
AI stocks are doing all the lifting — the index is narrower than the headline
Strip the AI enablers out of the S&P 500 and the picture changes. Per FT analysis cited alongside the record, the index excluding AI names is up only ~16% over the last two years and remains below its prior record. Tech is the only S&P 500 sector positive over the past month; real estate, financials, and materials have each dropped more than 4.5% as borrowing costs rose. The rally is being carried by a handful of stocks and the conviction that AI capex keeps compounding.
- NVIDIA closed near a $5.8T market cap (approaching $6T), set its third straight intraday record on Oct 6, and now accounts for more than 8% of the S&P 500 by weight — the largest single-stock concentration in the index's history.
- Microsoft and Apple sit beside NVIDIA in the top three; together the trio represents ~21% of the index. Meta Platforms and Alphabet round out a top-10 that now carries ~40% of the S&P 500, more than double the share a decade ago.
- Marvell Technology jumped ~5.8% on Oct 6 after raising its fiscal 2028 revenue outlook to ~$20B and laying out a path to $70B–$90B in annual revenue by fiscal 2031 at its Investor Day — same session as the record close.
- Advanced Micro Devices added 2.8% after CEO Lisa Su said the company plans to ramp chip supply next year, reinforcing the second-tier beneficiary leg of the AI trade.
The bond side
The bond rout has its own drivers — and they're not going away
The 10-year Treasury briefly cleared 5.35% on October 6 before pulling back to 5.27% — still the highest level since April 2002 — as the global bond complex rolled over. The move is not a U.S. story in isolation. UK 30-year gilts crossed 6% for the first time since 1998, Japan's 10-year JGB traded above 3% for the first time in three decades, and the 30-year JGB hit an all-time high near 3.28%. Global government-bond yields rose 17 basis points over the 20 trading days through late August — an unusually synchronized move.
- Term premium is widening as investors demand more compensation for holding long-dated government debt, with the heaviest fiscal supply pressure in the U.S., UK, and Japan — the three markets now at multi-decade yield highs.
- Inflation persistence, not recession risk, is the dominant driver: 10-year breakevens have stayed elevated even as the curve has steepened, signaling that markets price a higher-for-longer policy path rather than a 2027 Fed cutting cycle.
- Bank of America estimates the four largest U.S. hyperscalers will spend roughly 90% of their operating cash flow on capex in 2026, up from ~65% in 2025 — a financing dynamic that crowds out other investment and pressures long-dated yields higher.
The structural break
The stock-bond hedge is broken — and that's why today's tape matters
For two decades the standard portfolio playbook assumed a negative correlation between stocks and bonds: when equities sold off, duration rallied, smoothing drawdowns. That relationship has broken down repeatedly since 2022, and the Oct 6 tape is the most acute example yet. Stocks at a record high and bonds at a multi-decade low on the same day is the textbook signal that the 60/40 hedge is not working — and it is not working because the two assets are now responding to the same shock: AI capex is making equities richer (through earnings) and bonds poorer (through term premium and supply) at the same time.
- Pension funds running 60/40 or risk-parity mandates are entering Q4 with stocks overweight and bonds underweight their policy targets — meaning rebalancing flows would require selling AI winners and buying duration, the opposite of what the tape is rewarding.
- Risk-parity funds, which lever bonds to equalize volatility with equities, have been forced to delever as both legs lost simultaneously in early October — a margin dynamic that amplifies the moves in either direction when it snaps back.
- Per the IMF and BlackRock analyses published in 2026, stock-bond diversification has offered materially less protection from drawdowns than it did in the 2010s — a structural change driven by correlated inflation and fiscal regimes, not a cyclical anomaly.
Spreading investments across asset classes can reduce risk and smooth returns. The classic diversification between stocks and bonds worked [for decades]. Today, amid the hedging breakdown, higher volatility coincides with higher expected bond returns, with prices declining steeply in the current period.
Upstream
Who's getting paid — the picks-and-shovels names behind the AI capex
If the rally's foundation is real AI capex rather than narrative, the upstream names feeding it should be the most reliable expression of the trade. The Q3 prints and forward guidance delivered so far in 2026 say they are.
- Taiwan Semiconductor Manufacturing posted August 2026 revenue of NT$514.81B, up 53.3% year over year — a monthly record. Through August, cumulative 2026 revenue ran NT$3.39 trillion, up 39.3% YoY, and management has guided full-year 2026 capex to $52B–$56B.
- ASML raised its full-year 2026 net-sales forecast to €43B–€45B (up 16%) for the second time this year after Q4 2025 bookings surged to a record €13.2B — direct evidence that AI accelerator demand is pulling advanced lithography orders forward.
- Broadcom is tracking toward $58B in AI revenue for fiscal 2026, up 186% year over year, on the strength of custom-silicon and connectivity content per AI rack.
- Oracle delivered fiscal 2026 revenue of $67.4B (up 17%) and disclosed large AI contracts now totaling $75B against a remaining-performance obligation of $638B — the largest contracted AI backlog in the U.S. cloud peer group.
Downstream
Who's buying the power — the second-order winners from the same capex cycle
On the same day the S&P 500 printed a record, Constellation Energy and Alphabet announced a 20-year power purchase agreement to bring 890 megawatts of new nuclear capacity onto the PJM grid. Less than a week earlier, Constellation Energy had signed a similar agreement with Amazon. The pattern is unmistakable: AI capex has migrated from silicon to electrons, and utilities with dispatchable, 24/7 clean generation are the downstream beneficiaries.
- Constellation Energy signed a 20-year, 890 MW nuclear PPA with Alphabet on Oct 6 — adding to its Amazon deal announced days earlier and anchoring multi-decade cash flows against PJM power-price volatility.
- Vistra has locked in 20-year nuclear and gas PPAs with Meta Platforms for 2,600 MW and with Amazon for 1,200 MW; Q2 2026 adjusted EBITDA surged 30% to $1.77B as a result.
- Microsoft is the largest buyer in the cohort through its 20-year deal with Constellation Energy to restart Three Mile Island Unit 1 — the deal that defined the AI-power trade in 2024 and remains a template for the others.
What's next
The catalysts that can break or extend the divergence
Two setups will decide whether this tape extends or unwinds. The first is the cluster of mega-cap earnings reports in mid-November — NVIDIA prints fiscal Q3 2027 on November 17 with sell-side modeling $108B in revenue (~12% sequential growth) and management commentary on Vera Rubin uptake. The second is the fiscal and policy calendar: Q4 is when pensions and endowments rebalance, and any further leg higher in 30-year yields would force mechanical selling of equities to restore duration targets.
- Near-term (days–quarters): NVIDIA Q3 FY2027 print on November 17 is the single largest catalyst — consensus is $108B in revenue with Vera Rubin ~20% of data-center mix. A beat-and-raise keeps the AI rally rolling; a flat guide unwinds the narrow leadership that has carried the index.
- Near-term (days–quarters): Pension and endowment rebalancing in late Q4 historically redirects $50B–$100B of equity selling when stocks are over their policy bands — a meaningful headwind given current overweights.
- Long-term (1–3 years): Whether the AI capex supercycle validates the implied multiples depends on the answer to one question — can hyperscaler capex generate sufficient cash-flow conversion? Bank of America's ~90% capex-to-OCF ratio is the binding constraint, and any 2027 guide that prioritizes free cash flow over growth would mark the inflection from 'capex boom' to 'capex normalization.'
Related stocks
- Up ~5% in October and approaching $6T market cap; Q3 FY2027 print on Nov 17 is the binary catalyst for the divergence trade
- >8% S&P 500 weight means a 4% drop in NVDA alone would erase ~⅓ of the index's YTD gain — concentration is the risk as much as the opportunity
- Forward guide of $108B Q3 revenue and Vera Rubin ~20% of data-center mix sets the bar for the rest of the AI complex
- Azure cleared $100B for the first time in FY2026, up 43% — the cleanest read on enterprise AI adoption at scale
- Calendar 2026 capex guidance of ~$190B (per Q3 FY26 disclosure) puts MSFT among the highest absolute spenders; long-cycle revenue is locked
- Trailing P/E ~29x and forward P/E ~25x leave room for re-rating if Azure growth holds above 35%
- 2026 capex raised to $130B–$145B but Q3 revenue guide of $61B–$64B came in light — the trade-off between AI spend and near-term FCF is unresolved
- Forward P/E ~22x already discounts a meaningful capex drag; Q3 print on Oct 29 will clarify whether monetization is catching up to spend
- Core ad business still growing mid-20s — provides ballast if Reality Labs losses continue to widen
- Fiscal 2026 AI revenue tracking toward $58B, up 186% YoY — the steepest growth rate in the AI-exposed cohort
- Custom-silicon and connectivity content per AI rack makes AVGO the purest 'picks-and-shovels' name in the U.S. large-cap universe
- Hyperscaler relationships with Google and Meta provide multi-year visibility on the AI capex cycle
- August 2026 monthly revenue +53% YoY (NT$514.81B) — the strongest signal that AI accelerator demand is still capacity-constrained
- 2026 capex guided to $52B–$56B, well above prior expectations, confirms advanced-node supply remains the binding constraint
- Trade customers include NVDA, AVGO, AMD, MRVL — direct exposure to every leg of the AI accelerator stack
- Full-year 2026 net-sales guide raised to €43B–€45B (up 16%) for the second time in 2026 — the cleanest leading indicator of wafer-fab equipment demand
- Q4 2025 bookings of €13.2B set a quarterly record, with AI accelerators driving EUV adoption
- Sole supplier of EUV lithography means ASML captures the bottleneck economics of the entire AI build-out
- Oct 6, 2026: signed 20-year 890 MW nuclear PPA with Alphabet — adding to Amazon deal days earlier; same session as the S&P 500 record
- Nuclear fleet provides dispatchable 24/7 power, which data-center customers are explicitly preferring over intermittent renewables
- Long-dated PPAs anchor multi-decade cash flows against PJM price volatility — the AI-power trade's purest expression
- Locked in 2,600 MW PPA with Meta and 1,200 MW with AWS; Q2 2026 adjusted EBITDA up 30% YoY to $1.77B
- Diversified generation mix (nuclear, gas, renewables) gives flexibility across PJM, ERCOT, and MISO pricing environments
- Valuation has re-rated on the AI-power thesis but the contracted revenue backlog remains under-priced relative to peer utilities
