Cycle on Oct 1, 2026

I'll start by reviewing the topic's open threads and existing notes, then pull market data.

I have the full picture of open threads and prior findings. Now let me pull the market data (NVDA + macro backdrop) and scan for the latest news, since this cycle covers Sept 30 – Oct 1, 2026.

Market data pulled — NVDA fundamentals remain strong ($5.49T cap, 74.7% GM, beat 4/4 quarters), 10-yr at 5.26% (new high). Now let me check the specific watch-triggers that could have fired around Oct 1: the Nebius price hike, the Oracle AMD Helios deployment (fire date Sept 30), the China RTX Pro 5500 approval, and the DeepSeek-Huawei CUDA story.

Good progress. Key threads: the Nebius hike landed today (Oct 1) with a JPMorgan upgrade behind it; the China approval is still pending; and there's a new DeepSeek-Huawei CUDA challenge story. Let me fetch the CUDA story and check the Oracle AMD Helios watch (fire date was Sept 30) plus any fresh hyperscaler capex news.

The DeepSeek-Huawei CUDA story is significant — a direct challenge to Nvidia's software moat. Let me fetch the FT breakeven piece and the Nebius/JPMorgan article for solid citations, and check the RTX Spark retail watch (fire date is today).

I have enough confirmed evidence to write the findings now. This cycle: (1) a new DeepSeek-Huawei software-stack challenge to CUDA, (2) the 10-year at 5.26% matching the 2007 peak, (3) the Nebius hike landing today with JPMorgan confirming pricing power, and (4) NVDA's market snapshot showing the fundamentals intact. Writing all four notes now.

Two notes written. Now the demand-gauge update (Nebius hike landing + JPMorgan pricing-power evidence) and the NVDA moat update (Oct 1 market snapshot).

All four findings are persisted. Watch check before closing: none of the tripwires fired this cycle — the neocloud watch is for a price *cut* (the opposite happened), the Oracle AMD Helios deployment has no confirmed commercial availability yet, RTX Spark retail lands Oct 8, and the China RTX Pro 5500 approval and GPU-insurance talks remain pending. The Huawei Ascend watch already covers the third-market follow-on from the DeepSeek-Huawei story, so no new threads are needed. **Cycle summary — Is Nvidia's AI capex story still intact? Yes, on every observable gauge, but the challenge has moved to the software layer:** 1. **NEW — `cuda-moat-challenge-deepseek-huawei`**: DeepSeek open-sourced TileLang for Huawei's Ascend chips and built a 128-chip Ascend 950 supernode — the first open attack on CUDA's 7.5M-developer moat, landing exactly as Nvidia's own 10-Q warned rivals are building "larger developer and customer ecosystems to challenge us worldwide." 2. **`big-tech-capex-reaches-710b-in-2026`**: The demand gauge stays hot — Nebius's second hike (17–21%) landed Oct 1, JPMorgan upgraded CoreWeave on "stronger compute pricing" with 5–10pts of contribution-margin uplift on new contracts, and contracted power is up 35% in seven months. Goldman's caveat: capex growth and "upside surprises will diminish." 3. **`first-rate-hike-since-2023`**: The 10-year hit 5.26% (Sept 29) — matching the 2007 peak the prior revision measured from. 4. **`nvidia-record-q1-results-and-trillion-dollar-order-book`**: The Oct 1 market view grounds the "intact" verdict — $5.49T cap, TTM revenue $302.97B (+105.9%), 74.7% gross margin, 4/4 beats, $21.4B quarterly FCF, stock 2.9% below highs despite the rate backdrop.

The research cycle is complete. Four findings were persisted this cycle: - **New finding** — the DeepSeek-Huawei open-source software stack (TileLang, Ascend 950 supernode) as the first direct challenge to CUDA's developer moat - **Three updates** — the demand gauge (Nebius hike landed, JPMorgan pricing-power confirmation), the 10-year matching its 2007 peak at 5.26%, and NVDA's Oct 1 market snapshot ($5.49T cap, 74.7% GM, 4/4 beats) **Bottom line on the brief:** the capex story is intact on every observable gauge — rising neocloud pricing, expanding margins on renewals, record Nvidia fundamentals — but this cycle surfaced the first serious threat to the *software* layer of the moat, and the discount-rate backdrop keeps tightening. All tripwire watches (neocloud price cut, hyperscaler capex cut, GPU write-down, China approval, insurance deal) remain unfired and will be checked next cycle.