Cycle on Sep 9, 2026

I will begin by listing the open threads and reading the existing notes for this topic to understand the current state of our research and see if there are any active watches or questions we need to address.

I will now read the existing notes for this topic to understand what has already been established regarding Nvidia's AI capex, financial guarantees, joint ventures, and other facets.

I should check the current market view for NVDA to see its latest public financial snapshot. This is a critical step for a topic with the public-markets datasource enabled. I will call `market_view` with kind='ticker' and id='NVDA'. I'll also call `market_view` with kind='macro' to get the current macroeconomic backdrop, which influences capex. Let's start with NVDA.

Let's see what's happening with "Is Nvidia's AI capex story still intact?" Let's search the web for recent developments, news, and reports about Nvidia's capex, hyperscaler capex, or AI infrastructure demand. I'll search for "Nvidia capex", "hyperscaler capex 2026", "Nvidia AI infrastructure demand", "Nvidia earnings", or "Nvidia Q3 FY2027" (or Q2 FY2027 since their quarter ended July 31, 2026 is likely Q2 FY2027, given their fiscal year runs ahead). Let's do a web search.

The results show that Nvidia's earnings for the quarter ending July 31, 2026 (which was Q2 FY2027) were reported. Let's do some more target searches on Nvidia's Q2 FY27 earnings, Colette Kress's comments, Jensen Huang's comments, and the hyperscaler capex trend. Let's look for "NVIDIA Q2 FY27 earnings transcript" or "NVIDIA Q2 FY2027" or "NVIDIA Q2 2027 earnings" to see what they reported and what the guide was. Wait, let's look at the REX Shares article: `https://www.rexshares.com/nvidia-earnings/`. Let's fetch that page first to see what it says about Q2 FY27. It says: "Q2 2026 combined hyperscaler capex: $166.0B (+87% YoY, +27% QoQ)." Wait, let's fetch this URL and search for other articles. Let's use `web_fetch` on `https://www.rexshares.com/nvidia-earnings/`.

This is extremely helpful! The REX Shares article provides the exact Q2 FY2027 numbers reported on August 26, 2026: - Q2 FY27 Revenue: $96.22B (vs $91.0B guide midpoint, beat by 5.7%) - Q2 FY27 EPS (non-GAAP): $2.22 (vs $2.09 expected, beat by 6.2%) - Q3 FY27 Revenue Guide: $108.0B (vs $103.8B expected, +4.0% vs street) - Data Center Revenue: $89.0B (92% of total) - Gross margin was 75.0% (wait, our market view says 74.7% TTM or GAAP, but non-GAAP or GAAP for the quarter was 75.0% or 74.7%). - Combined Q2 2026 hyperscaler capex (Amazon, Alphabet, Microsoft, Meta) was $166.01B, up 87% YoY and 27.1% QoQ (from $130.60B in Q1 2026). - Amazon Q2 2026 capex: $54.21B (+22.6% QoQ from $44.20B) - Alphabet Q2 2026 capex: $44.92B (+25.9% QoQ from $35.67B) - Microsoft Q2 2026 capex: $35.80B (+15.9% QoQ from $30.88B) - Meta Q2 2026 capex: $31.08B (+56.7% QoQ from $19.84B) Let's do a search on other elements of the capex story. What about the $1.3 trillion capex projection for 2027? Let's search for "Colette Kress" "1.3 trillion" or "Kress" "1.3 trillion" or "Kress" "backlogs". Let's search for "Nvidia" "1.3 trillion" capex. This will help us verify the sources and comments.

Let's fetch the Crypto Briefing article or the AlphaStreet Q2 2027 earnings call transcript to see the exact quotes from Colette Kress regarding the $1.3 trillion capex and $2 trillion backlog. Let's search for "NVIDIA Q2 2027 earnings call transcript" or "Nvidia Q2 2027 transcript" to get the exact transcript. Let's do a web search.

This is incredible! We have the exact transcript of NVIDIA's Q2 2027 Earnings Call (held on August 26, 2026). Let's fetch the Motley Fool transcript URL `https://www.fool.com/earnings/call-transcripts/2026/08/31/nvidia-nvda-q2-2027-earnings-call-transcript/` or the PDF from Q4cdn `https://s201.q4cdn.com/141608511/files/content_files/TRANSCRIPT_-NVIDIA-Corp-NVDA-US-Q2-2027-Earnings-Call-26-August-2026-5_00-PM-ET.pdf` to retrieve exact quotes. Let's write a query to search specifically within these documents, or use `web_fetch` to get the PDF or Motley Fool transcript. Let's do a fetch of the Motley Fool transcript first. If that's too long, let's fetch parts or search for specific sections. Let's see if we can fetch the Motley Fool transcript. Since it's a transcript, it might be large, but let's try to fetch it. Wait, the PDF is also a good option. Let's fetch the Motley Fool transcript.

This is fantastic! Let's analyze the key details from the transcript of NVIDIA's Q2 2027 earnings call: 1. **Top-line and Capex Story**: - Total Revenue: $96.2 billion, up 106% YoY. - Data Center Revenue: $89.0 billion, up 117% YoY. - Q3 Revenue Guidance: $108.0 billion (plus or minus 2%), driven by the ACIE (AI, cloud, industrial, and enterprise) segment. - FY2028 Revenue Guidance: Management expects YoY growth of approximately 70%, which represents a "supply-constrained outlook." - Hyperscaler Capex: Colette Kress stated: "With cloud industry backlog now greater than $2 trillion, CapEx by the top 5 hyperscalers is expected to reach nearly $800 billion in 2026 and $1.3 trillion in 2027." - This absolutely confirms the capex story is intact and scaling even higher. The previous number for 2026 was $710 billion or $610 billion, and is now projected to reach "nearly $800 billion in 2026 and $1.3 trillion in 2027" by the top five customers! - Note that this matches our existing note: `[[big-tech-capex-reaches-710b-in-2026]]` which was titled "Big Tech's AI Capex Surpasses $800 Billion in 2026 on Track for $1.3 Trillion in 2027 as Nvidia Projects 70% Growth". We can update and refine this note. 2. **Supply Constraints and Memory Costs**: - "Inventory increased to $31.6 billion as the company prepared for the Vera Rubin platform launch." - "Q4 Gross Margin Guidance: Expected to bottom in the 71% to 72% range due to significant price increases in memory components." - Kress stated: "The magnitude of the price increase has exceeded our prior expectations and are headed even higher into next year," referring to extreme pricing conditions in memory. - This aligns with our watch `Nvidia Q4 FY2027 Gross Margin Bottoming (71%-72%)` (thread ID: `01a0409b-bb2f-7d8f-a4a6-e30c93fff0e0`) and `HBM Memory Component Pricing and Supply Bottlenecks` (thread ID: `019fbaba-2a7b-7e80-b885-9f29c3a7f0ee`). 3. **Segment breakdown**: - Hyperscale Revenue: $49.0 billion, up 13% sequentially. - ACIE Segment (AI, cloud, industrial, and enterprise) Revenue: $40.0 billion, up 25% sequentially and 138% YoY, as NeoCloud partners bring capacity online. - Sovereign AI Revenue (regional NeoClouds) grew 35% sequentially and more than tripled YoY. - Grace CPU Revenue exceeded $5.0 billion on a TTM basis. - Networking Revenue increased 18% sequentially, Spectrum-X grew 2.6x YoY. - Edge Computing: $7.2 billion, up 27% YoY. - Q3 Vera Rubin Mix: Expected to account for ~20% of data center revenue. - Automotive trailing 12-month revenue: $8.0 billion. 4. **Financing and Revenue Sharing**: - "The company introduced a revenue-sharing structure for NeoClouds, providing a minimum revenue guarantee to secure financing in exchange for a portion of future rental income." - This is a massive new development in their off-balance-sheet financing and demand-engineering model! Let's look closer at this "revenue-sharing structure for NeoClouds, providing a minimum revenue guarantee to secure financing in exchange for a portion of future rental income." This is highly relevant to `[[off-balance-sheet-ai-infrastructure-financing-models]]` and `[[nvidia-record-q1-results-and-trillion-dollar-order-book]]`. Let's do a search on this specific revenue-sharing program and minimum revenue guarantees for NeoClouds. This is a very interesting structural shift! Let's search for: `"minimum revenue guarantee" OR "revenue-sharing" OR "revenue sharing" "NeoClouds" OR "neocloud" OR "Nvidia"`.

Wow, this is a massive, breaking story! Let's look at the search results: "NVIDIA Pauses $36 Billion Revenue-Sharing Financing Model Amid Antitrust Scrutiny, Partner Pushback." (Published August 28, 2026). "Nvidia pauses revenue-sharing deals with AI cloud companies, WSJ reports" "Nvidia Pauses Revenue-Sharing Deals With AI Cloud Companies" (WSJ) Let's do a web search on this WSJ report or other articles about "Nvidia pauses revenue-sharing deals" to understand exactly what happened, when, and why. I'll search for: `"Nvidia pauses revenue-sharing" OR "Nvidia Pauses $36 Billion Revenue-Sharing"` to retrieve the most detailed articles.

This is a massive story that perfectly ties into the "Nvidia's AI capex story still intact?" brief, especially as it relates to their complex off-balance-sheet financing, demand-engineering models, and neocloud partnerships. Let's carefully piece together what we have learned: 1. **The Q2 FY2027 Earnings Report (August 26, 2026)**: - Nvidia delivered another blowout quarter with $96.22 billion in revenue (up 106% YoY), beating their own guide midpoint of $91.0 billion by 5.7% and consensus of $92.07 billion by 4.5%. - Data Center revenue came in at $89.0 billion (representing 92.5% of total revenue), driven by strong contributions from both hyperscalers ($49 billion, up 13% QoQ) and ACIE (AI, cloud, industrial, enterprise) customers ($40 billion, up 25% QoQ). - Q3 FY2027 revenue guidance was set at $108.0 billion (plus or minus 2%), which is +4.0% above the street consensus of $103.8 billion. - For the full fiscal year 2028, Nvidia expects revenue growth of approximately 70%—representing a "supply-constrained outlook." - In the earnings call, CFO Colette Kress stated that the top five hyperscalers are expected to spend nearly $800 billion on capex in 2026 and $1.3 trillion in 2027, backed by a cloud industry backlog of greater than $2 trillion. - However, gross margins are expected to bottom in the 71%-72% range in Q4 FY2027 due to "significant price increases in memory components." Kress noted that "the magnitude of the price increase has exceeded our prior expectations and are headed even higher into next year." 2. **The Revenue-Sharing & Credit-Support Model Controversy (August 27, 2026)**: - Just one day after the earnings call, on August 27, 2026, the Wall Street Journal reported that Nvidia had paused some of its deals under its new revenue-sharing financing program. - Under this model, Nvidia provided credit support and minimum revenue guarantees (offering to rent back compute capacity if the cloud provider couldn't sell it) to help small AI cloud providers (NeoClouds) secure financing to purchase Nvidia chips. - In return, Nvidia would profit twice: first from the initial hardware sale, and second by taking a 50% cut of future cloud rental revenues earned above an agreed cost threshold. - The total commitments tied to these agreements reached $36 billion over a typical six-year duration, as disclosed in a recent quarterly filing. - **Why the pause?** - **Antitrust Scrutiny**: Nvidia employees expressed concerns to customers that the initiative could draw intense antitrust scrutiny, as it involved Nvidia dictating how customers do business (e.g., telling cloud providers they could only rent chips to approved customers, and preferring capacity to be spread among multiple smaller firms rather than one large customer). - **Circular Financing Concerns**: Investor scrutiny has mounted over Nvidia pumping money back into the AI ecosystem, stoking fears over "circular deals" that could artificially inflate demand. - **Partner Pushback**: Some potential partners were irked by the level of control Nvidia sought in the early weeks of the program. Let's check the existing notes to see which ones we should update. The note `[[big-tech-capex-reaches-710b-in-2026]]` is titled "Big Tech's AI Capex Surpasses $800 Billion in 2026 on Track for $1.3 Trillion in 2027 as Nvidia Projects 70% Growth". Let's check its contents using `read_notes` or since we already have its preview/title, we can update it directly. Wait! Let's fetch the actual content of `[[big-tech-capex-reaches-710b-in-2026]]` first so we can properly revise it without losing its context. Wait, can we read the note? Let's check if there is a tool to read a specific note. Ah, `read_notes` lists the notes but doesn't return the full body. Let's look at the parameters of `read_notes`. It takes `topic_id` and `limit`. Let's see if we can get the full body. Ah, `read_notes` returns slug, title, preview, updated_at, and source_count. It does not return the full body of all notes unless we call it, or maybe it returns the full body? Let's check. Wait, the description says "Returns slug, title, preview, updated_at, and source_count for each note". Let's check if we can fetch the note file directly or if we have to write the update. Wait, since we are in a linux workspace, are the notes stored locally in our workspace directory? Let's check our workspace files! Our workspace is `/var/lib/hey-lefty/notes/019e89b4-58b3-7e3b-a0c2-6a68bd3242db/019e95ad-bd47-7cf5-bd57-eb159d26bbbe`. Let's check if there are files in this directory. Let's list the directory contents using standard shell commands? No, we don't have shell tools in our declarations! The only tools we have are: `list_threads`, `record_thread`, `resolve_thread`, `record_watch`, `record_event`, `write_note`, `read_notes`, `web_search`, `web_fetch`, `market_view`, and `finish_cycle`. Wait, let's verify if there are other declarations. No, those are the only ones! So we don't have a direct tool to read the full body of a note unless we can find it, or we can just write a comprehensive update to the notes using `write_note`. Let's see: `write_note` takes `topic_id`, `slug`, `body`, and `rationale`. Wait, let's look at the existing notes: 1. `big-tech-capex-reaches-710b-in-2026`: "Big Tech's AI Capex Surpasses $800 Billion in 2026 on Track for $1.3 Trillion in 2027 as Nvidia Projects 70% Growth" 2. `off-balance-sheet-ai-infrastructure-financing-models`: "Off-Balance-Sheet Joint Ventures and the New AI Infrastructure Financing Model: Nvidia's $500 Billion Wall Street Platform and the $105 Billion OpenAI Guarantee" 3. `nvidia-server-price-hikes-and-margin-protection`: "Nvidia's AI Margin Protection: HBM Cost Inflation Triggers Gross Margin Reset as Samsung Mounts Custom HBM4E Challenge" Let's check if there are any other notes. Yes: - `nvidia-record-q1-results-and-trillion-dollar-order-book`: "Nvidia's AI Moat: $164.5 Billion in Financial Guarantees, the Finalized $105 Billion OpenAI Deal, and the $500 Billion Wall Street Compute Platform" - `gpu-specialized-clouds-and-strategic-co-investment`: "Specialized GPU Clouds Face Competitive Disruption and Strategic Consolidation as Nvidia Takes 9.3% Stake in Nebius" Let's check if there are any watches that have fired. Let's look at our open threads: 1. `Nvidia Q3 FY2027 Earnings Release and Gross Margin Performance` (thread ID: `01a069d0-b1a2-7ba9-b116-e1dd31f2d79f`): "Monitor Nvidia's Q3 FY2027 earnings release, specifically verifying if revenue hits the guided $108 billion and if gross margins land in the guided 74% range." This is a dated catalyst for November 17, 2026. Today is September 9, 2026, so this has NOT fired yet. 2. `Nvidia FY2028 Gross Margin Settling (72%-73%)` (thread ID: `01a06ef8-fdde-7ef6-bbfe-0b4764e55dde`): "Verify if Nvidia's full-year FY2028 gross margin settles in the guided 72%-73% range...". This is for February 25, 2028, so it has NOT fired. 3. `Nvidia FY2028 Revenue Growth (Guided ~70%)` (thread ID: `01a0409b-c90b-7527-a2c7-f7064fc17243`): "Verify if Nvidia's full-year FY2028 revenue meets or exceeds the guided 70% growth rate as supply constraints are resolved." This is for February 25, 2028. But wait! The guidance was officially given in the Q2 FY2027 earnings call on August 26, 2026 ("Management expects year-over-year growth of approximately 70%, representing a supply-constrained outlook"). This watch is to *verify if they meet it*, so it will fire in 2028. 4. `Nvidia Q4 FY2027 Gross Margin Bottoming (71%-72%)` (thread ID: `01a0409b-bb2f-7d8f-a4a6-e30c93fff0e0`): "Verify if Nvidia's gross margins bottom out in the guided 71%-72% range in Q4 FY2027...". This is for February 25, 2027, so it has NOT fired. 5. `HBM Memory Component Pricing and Supply Bottlenecks` (thread ID: `019fbaba-2a7b-7e80-b885-9f29c3a7f0ee`): "Monitor HBM component pricing...". This is an ongoing watch. Has it fired? Let's check. In the Q2 FY2027 earnings call, Colette Kress explicitly stated that memory price increases have exceeded expectations and are headed higher, leading to a bottoming of gross margins in Q4. This is a significant event! We can record an event for this ongoing watch to note the escalation. 6. `Nvidia $500B Wall Street Compute Financing Platform Capital Deployment` (thread ID: `019fee38-1dbd-73a3-8fa2-113b552feff5`): "Monitor the official launch, terms, and capital deployment of the first compute financing transaction...". This is ongoing. In the Q2 FY2027 call, Kress mentioned that the top five customers are expected to spend $1.3T in 2027, and this month (August 2026) Nvidia helped arrange $500 billion in financing from major U.S. financial institutions (Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, KKR). Let's see if we can find more details about this $500B arrangement. Let's do a web search. Let's search for: `"Nvidia" "$500 billion" "Apollo" OR "BlackRock" OR "Blackstone" OR "Brookfield" OR "Goldman" OR "KKR"`. This is highly relevant to our off-balance-sheet financing note. Let's search.

This is phenomenal! Let's analyze the $500 billion Wall Street deal: - Announced on Monday, August 10, 2026. - Nvidia partnered with six major Wall Street asset managers: Apollo Global Management, Blackstone, BlackRock, Brookfield Asset Management, Goldman Sachs, and KKR. - The purpose is to establish "compute financing platforms" to mobilize over $500 billion of third-party capital for the buildout of AI infrastructure over time. - Under these agreements, these firms will create dedicated pools of capital at attractive rates for Nvidia customers (including frontier AI labs, enterprises, and AI clouds) to build out data centers and acquire Nvidia hardware. - Jensen Huang, CEO of Nvidia, stated: "We began by building chips; today, we are helping create a new class of productive, investable infrastructure: AI factories." He added that "Nvidia compute is uniquely suited for this role... extending its useful life and improving its economics over time." - This is a massive confirmation of the off-balance-sheet financing platform we are tracking! This perfectly updates `[[off-balance-sheet-ai-infrastructure-financing-models]]` and `[[nvidia-record-q1-results-and-trillion-dollar-order-book]]`. Wait, let's look at the other major piece of news: Nvidia's $36 Billion Revenue-Sharing program with NeoClouds was paused on August 27, 2026 (just one day after their Q2 FY2027 earnings call). - Under the revenue-sharing model, Nvidia provided credit support and minimum revenue guarantees (offering to rent back compute capacity if the cloud provider couldn't sell it) to help small AI cloud providers (NeoClouds) secure financing. - In return, Nvidia would profit twice: first from the initial hardware sale, and second by taking a 50% cut of future cloud rental revenues earned above an agreed cost threshold. - The scale of the program was $36 billion in total commitments over a typical six-year duration. - However, Nvidia paused this program due to concerns over: 1. **Antitrust Scrutiny**: Employees expressed concerns that dictating how customers do business (e.g., telling cloud providers they could only rent chips to approved customers and preferring capacity to be spread among multiple smaller firms) could draw antitrust scrutiny. 2. **Circular Financing Concerns**: Investor scrutiny has mounted over Nvidia pumping money back into the AI ecosystem, stoking fears over "circular deals" that could artificially inflate demand. 3. **Partner Pushback**: Some potential partners were irked by the level of control Nvidia sought. This is a massive development! Let's write/update the notes. Wait, let's review the required note structure and critical rules: - One finding = ONE coherent subject tracked under one stable slug across cycles. - Reuse slugs of existing notes to update them in place. - Let's check our list of existing notes and slugs: - `big-tech-capex-reaches-710b-in-2026`: We should update this to reflect the new numbers from Q2 FY2027: - Aggregate capex by top 5 hyperscalers is now projected to reach nearly $800 billion in 2026 and $1.3 trillion in 2027 (as announced by CFO Colette Kress on August 26, 2026). - Combined Q2 2026 capex for the top 4 (Amazon, Alphabet, Microsoft, Meta) reached a record $166.01B (up 87% YoY and 27.1% QoQ). - Nvidia's own Q2 FY2027 revenue reached $96.22B (up 106% YoY), with Data Center revenue at $89.0B (up 117% YoY). - Nvidia's Q3 FY2027 revenue guidance is $108.0B. - Nvidia's full-year FY2028 revenue growth is guided at ~70% (supply-constrained). - `off-balance-sheet-ai-infrastructure-financing-models`: We should update this note to include: - The official announcement on August 10, 2026 of Nvidia's $500 billion partnership with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to establish "compute financing platforms" to fund AI factories. - The launch and subsequent pausing of the $36 billion NeoCloud revenue-sharing and credit-support program on August 27, 2026, amid antitrust concerns, partner pushback, and circular financing fears. - `nvidia-server-price-hikes-and-margin-protection`: We should update this note to include: - Q4 FY2027 gross margins are guided to bottom in the 71%-72% range due to extreme pricing conditions in memory components. - Colette Kress's quote: "The magnitude of the price increase has exceeded our prior expectations and are headed even higher into next year." - This highlights the margin pressures from HBM cost inflation. Let's check if there are other notes we should update or if we should write new ones. Wait, we should write notes that tell the story through concrete evidence, citing our sources with verbatim quotes. Let's prepare the bodies for the notes. First, let's update `big-tech-capex-reaches-710b-in-2026`. Wait, let's make sure we write a highly detailed, professional, and cited note. Let's see: what are the sources for the capex numbers? - Motley Fool transcript of Nvidia Q2 2027 Earnings Call: `https://www.fool.com/earnings/call-transcripts/2026/08/31/nvidia-nvda-q2-2027-earnings-call-transcript/` - REX Shares article on Nvidia Earnings Q2 FY27: `https://www.rexshares.com/nvidia-earnings/` - Nvidia's market view on `/markets/NVDA/2026/09/09` Let's draft the body for `big-tech-capex-reaches-710b-in-2026`: ```markdown # Big Tech's AI Capex Surpasses $800 Billion in 2026 on Track for $1.3 Trillion in 2027 as Nvidia Projects 70% Growth The massive capital expenditure cycle driving artificial intelligence infrastructure has accelerated to historic proportions. During NVIDIA Corporation's (NVDA) fiscal Q2 2027 earnings call on August 26, 2026, CFO Colette Kress revealed that capital expenditures by the top five hyperscalers are now projected to reach nearly $800 billion in 2026 and surge to $1.3 trillion in 2027, supported by a massive cloud industry backlog exceeding $2 trillion. Nvidia's own results and guidance demonstrate that the AI capex story remains fully intact, even as the company operates under severe supply constraints. For fiscal Q2 2027, Nvidia reported record total revenue of $96.22 billion (up 106% YoY), beating its own guidance midpoint of $91.0 billion by 5.7%. Data Center revenue reached $89.0 billion (representing 92.5% of total revenue, up 117% YoY), driven by a sequential increase of 13% in hyperscaler revenue ($49.0 billion) and 25% in the ACIE (AI, cloud, industrial, and enterprise) segment ($40.0 billion). Nvidia guided Q3 FY2027 revenue to $108.0 billion (plus or minus 2%), representing a 4.0% beat against street expectations of $103.8 billion. Looking further ahead, management issued an initial fiscal year 2028 revenue growth guide of approximately 70% YoY, which Kress characterized as a "supply-constrained outlook." This relentless demand is mirrored in the quarterly capital spending of Nvidia's largest customers. In the calendar second quarter of 2026, the four largest hyperscalers—Amazon, Alphabet, Microsoft, and Meta—spent a combined record of $166.01 billion on capital expenditures, representing an 87% year-over-year increase and a 27.1% sequential jump from Q1 2026 ($130.60 billion). ## Sources - [NVIDIA Corp. Q2 2027 Earnings Call Transcript](https://www.fool.com/earnings/call-transcripts/2026/08/31/nvidia-nvda-q2-2027-earnings-call-transcript/) — "With cloud industry backlog now greater than $2 trillion, CapEx by the top 5 hyperscalers is expected to reach nearly $800 billion in 2026 and $1.3 trillion in 2027." - [NVIDIA Earnings Q2 FY27: Revenue, Data Center, AI Capex](https://www.rexshares.com/nvidia-earnings/) — "NVIDIA reported revenue of $96.22 billion for fiscal Q2 2027, 5.7% above the company’s own guide." - [NVIDIA Corporation (NVDA) Market View](/markets/NVDA/2026/09/09) — Nvidia financial quality, quality & returns, and quarterly trajectory showing TTM revenue of $302.97B. ``` Wait, let's verify if the TTM revenue is indeed $302.97B. Yes, the market view says: "Revenue (TTM): $302.97B (+105.9% YoY)". Let's call `write_note` for `big-tech-capex-reaches-710b-in-2026`. Let's do that.

Now let's work on `off-balance-sheet-ai-infrastructure-financing-models`. We have two massive developments here: 1. **The $500 Billion Wall Street Compute Financing Platform**: - Announced on August 10, 2026. - Nvidia partnered with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR. - Designed to mobilize over $500 billion of third-party capital to establish "compute financing platforms." This treats compute infrastructure as an investable asset class (similar to real estate or toll roads), creating dedicated pools of capital for Nvidia's customers to fund data center buildouts and acquire Nvidia chips. 2. **The Paused $36 Billion NeoCloud Revenue-Sharing Program**: - Announced in early July 2026 and paused on August 27, 2026 (WSJ report). - Under this credit-support model, Nvidia provided minimum revenue guarantees and credit support to small AI cloud providers (NeoClouds) to help them secure debt financing for Nvidia hardware. - In return, Nvidia took a 50% cut of their rental revenues earned above an agreed threshold. The total commitments under this model reached $36 billion over six-year durations. - The program was paused due to concerns that Nvidia's strict controls (e.g., dictating which customers they could lease to, and requiring distribution across smaller firms) would trigger antitrust scrutiny. Additionally, investors and regulators raised concerns about "circular financing" deals where Nvidia pumps money back into its own ecosystem to artificially inflate chip demand. Let's write a comprehensive, cited, and detailed note for `off-balance-sheet-ai-infrastructure-financing-models`. Let's draft this note carefully. ```markdown # Off-Balance-Sheet Joint Ventures and the New AI Infrastructure Financing Model: Nvidia's $500 Billion Wall Street Platform and the $105 Billion OpenAI Guarantee As the capital expenditures required to build frontier AI infrastructure balloon toward $1.3 trillion (see [[big-tech-capex-reaches-710b-in-2026]]), the traditional model of relying solely on corporate balance sheets has given way to highly structured, off-balance-sheet financing vehicles. Nvidia has positioned itself at the absolute center of this structural shift, engineering two massive financial initiatives in August 2026 to unlock third-party capital while actively managing demand. ### The $500 Billion Wall Street Private Equity Alliance On August 10, 2026, NVIDIA announced historic partnerships with six of the world's premier alternative asset managers—Apollo Global Management, Blackstone, BlackRock, Brookfield Asset Management, Goldman Sachs, and KKR—to establish independent "compute financing platforms." The coalition aims to mobilize over $500 billion of third-party capital over time to finance data center buildouts and hardware acquisitions for hyperscalers, frontier AI labs, and specialized AI clouds. This model treats Nvidia's full-stack accelerated computing systems as a scarce, yielding infrastructure asset class—akin to commercial real estate, power grids, or toll roads—against which capital providers can independently underwrite debt. By creating dedicated pools of capital at attractive rates, Nvidia ensures that its customers can access compute at scale without straining their own balance sheets or inducing immediate dilution. ### The Pausing of the $36 Billion NeoCloud Revenue-Sharing Model While Nvidia's institutional private equity alliance has forged ahead, its more aggressive attempt to finance smaller specialized GPU clouds ("NeoClouds") has run into severe regulatory and commercial headwinds. On August 27, 2026, the Wall Street Journal reported that Nvidia had paused some deals under its new revenue-sharing and credit-support financing initiative, less than two months after launching it. Under this model, Nvidia provided credit support and minimum revenue guarantees—effectively committing to lease back compute capacity if a NeoCloud customer was unable to rent it out. This guaranteed buyer structure allowed highly leveraged startups to secure bank debt to purchase Nvidia chips. In return, Nvidia stood to profit twice: first from the initial hardware sale, and subsequently by taking a 50% cut of any rental revenues earned above an agreed cost threshold. Nvidia's quarterly SEC filing revealed that total commitments tied to these six-year revenue-sharing agreements had reached $36 billion. However, the program was abruptly paused due to a combination of internal and external pressures: 1. **Antitrust Scrutiny**: Nvidia employees raised concerns that the strict operational controls the company sought to impose could draw antitrust investigations. Under the proposed terms, Nvidia dictated how customers did business, including telling cloud providers they could only lease chips to approved customers and requiring that capacity be distributed among multiple smaller AI startups rather than a single large client. 2. **Circular Financing Fears**: The program stoked mounting investor anxiety regarding circular transactions. Critics argued that by providing financial guarantees, credit support, and direct equity investments to its own customers, Nvidia was effectively recycling its hardware profits to artificially inflate demand for its chips. 3. **Partner Pushback**: Several potential NeoCloud partners pushed back against the level of control Nvidia sought in the early weeks of negotiations, prompting the company to step back and rethink the program's structure. ## Sources - [NVIDIA Partners With Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to Establish AI Compute Infrastructure Financing Platforms](https://nvidianews.nvidia.com/news/nvidia-partners-with-apollo-blackrock-blackstone-brookfield-goldman-sachs-and-kkr-to-establish-ai-compute-infrastructure-financing-platforms-to-mobilize-over-500-billion-of-third-party-capital) — "Memorandums of understanding signed with six of the world’s premier financial institutions to create these partnerships aim to establish the first compute financing platforms of their kind at global scale..." - [Nvidia, Wall Street asset managers partner on $500B AI push](https://www.cnbc.com/2026/08/10/nvidia-wall-street-asset-managers-500-billion-ai-push.html) — "Nvidia is attempting to turn its artificial intelligence chips into Wall Street’s newest asset class, partnering with six large asset managers on a $500 billion financing push..." - [Nvidia pauses revenue-sharing deals with AI cloud companies, WSJ reports](https://finance.yahoo.com/news/nvidia-pauses-revenue-sharing-deals-223140237.html) — "Nvidia has paused some deals in a new financing initiative that offered credit support to AI cloud companies in exchange for a share of revenue, the Wall Street Journal reported on Thursday, citing people familiar with the matter." - [NVIDIA Pauses $36 Billion Revenue-Sharing Financing Model](https://techstrong.ai/articles/nvidia-pauses-36-billion-revenue-sharing-financing-model-amid-antitrust-scrutiny-partner-pushback/) — "The scale of the venture was revealed in a recent quarterly filing, which detailed $36 billion in total commitments tied to agreements spanning a typical six-year duration." ``` Let's call `write_note` for `off-balance-sheet-ai-infrastructure-financing-models`. Let's do that.

Now let's work on `nvidia-server-price-hikes-and-margin-protection`. Let's update this note to reflect the latest guidance from the Q2 FY2027 earnings call regarding the gross margin reset and the extreme pricing conditions in memory components. Wait, let's look at the specific details we retrieved from the earnings call transcript: - Gross Margin was 75.0% for Q2 FY2027 (both GAAP and non-GAAP), which was largely unchanged from the previous quarter. - Q4 Gross Margin Guidance is expected to bottom in the 71% to 72% range due to "significant price increases in memory components." - Colette Kress stated: "The magnitude of the price increase has exceeded our prior expectations and are headed even higher into next year," referring to extreme pricing conditions in memory. - This represents a significant margin compression from their historic 75%-76% levels, but they are planning server price hikes of >15% starting in early 2027 to protect margins. - They also have the upcoming Samsung HBM4E qualification and the SK Hynix HBM4 strategic partnership to help ease bottlenecks over time. Let's draft this note. ```markdown # Nvidia's AI Margin Protection: HBM Cost Inflation Triggers Gross Margin Reset as Samsung Mounts Custom HBM4E Challenge While NVIDIA Corporation continues to deliver record top-line growth (see [[big-tech-capex-reaches-710b-in-2026]]), the highly profitable gross margin profile of its AI hardware has run into a significant near-term bottleneck. During its fiscal Q2 2027 earnings call on August 26, 2026, Nvidia guided that its gross margins will bottom in the 71% to 72% range in Q4 FY2027—a notable decline from the 75.0% reported in Q2. ### Severe Memory Cost Inflation This margin compression is being driven primarily by extreme pricing pressures in the High Bandwidth Memory (HBM) supply chain. As Nvidia prepares to ramp its next-generation Vera Rubin platform, memory suppliers (including SK Hynix, Micron, and Samsung) have aggressively raised component prices, catching Nvidia's management off guard. CFO Colette Kress warned on the earnings call: > "The magnitude of the price increase has exceeded our prior expectations and are headed even higher into next year." To offset these rising memory component costs and protect its long-term gross margin profile, Nvidia is planning to implement server price hikes of over 15% starting in early 2027. This aggressive pricing strategy is intended to push gross margins back toward the guided 72% to 73% range for full-year FY2028, and eventually back above 74% as supply chain dynamics stabilize. ### The HBM4/HBM4E Competitive Landscape The primary supply-chain bottleneck remains the yield and availability of custom HBM memory. To diversify its supply chain away from its heavy reliance on SK Hynix, Nvidia is actively working to qualify Samsung's custom 8-layer NVHBM (HBM4E) chip. This qualification, expected to enter high-volume commercial shipments by late 2027, represents a critical catalyst to ease memory supply bottlenecks and alleviate the cost inflation currently squeezing Nvidia's gross margins. ## Sources - [Nvidia (NVDA) Q2 2027 Earnings Call Transcript](https://www.fool.com/earnings/call-transcripts/2026/08/31/nvidia-nvda-q2-2027-earnings-call-transcript/) — "The magnitude of the price increase has exceeded our prior expectations and are headed even higher into next year," and "Q4 Gross Margin Guidance – Expected to bottom in the 71% to 72% range due to significant price increases in memory components." - [NVIDIA Earnings Q2 FY27: Revenue, Data Center, AI Capex](https://www.rexshares.com/nvidia-earnings/) — Detail on Q2 75.0% gross margin and forward cost pressures. ``` Let's call `write_note` for `nvidia-server-price-hikes-and-margin-protection`. Let's do that.

This is amazing! We have updated all three relevant notes. Let's review if there are any other notes we should update or if there's anything else. What about the open thread watches? We identified an ongoing watch: `HBM Memory Component Pricing and Supply Bottlenecks` (thread ID: `019fbaba-2a7b-7e80-b885-9f29c3a7f0ee`). Let's see if we should record an event on it. Yes! The watch is ongoing, so we should call `record_event` with `resolves: false` and a summary of what happened. Let's look at the watch description: "Monitor HBM (High Bandwidth Memory) component pricing, manufacturing yields, and supply bottlenecks (SK Hynix, Micron, Samsung) as a key driver of hyperscaler hardware procurement costs and potential margin pressure for Nvidia and its customers." And the trigger: "Fires if major memory manufacturers report severe HBM supply shortages, yield issues on HBM4, or if component price hikes drive further revisions to hyperscaler capex budgets." During the Q2 FY2027 earnings call, Nvidia's CFO explicitly stated that memory price increases have exceeded expectations and are headed even higher, forcing them to guide Q4 gross margins down to 71%-72%. This is a direct hit on this watch! Let's call `record_event` for this watch. Let's do that.

This is excellent! The event is recorded. Let's check if there are other watches we should resolve or record events on. Wait, what about the watch `Nvidia $500B Wall Street Compute Financing Platform Capital Deployment` (thread ID: `019fee38-1dbd-73a3-8fa2-113b552feff5`)? Its description: "Monitor the official launch, terms, and capital deployment of the first compute financing transaction under Nvidia's $500 billion partnership with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR." And the trigger: "Fires when Nvidia or its financial partners announce the first finalized compute financing deal or pool deployment under the $500B MOU." Wait, has the first finalized compute financing deal or pool deployment been announced? The partnership was announced on August 10, 2026. The announcement says: "These partnerships remain subject to execution of the final agreements." So this was an MOU (Memorandum of Understanding). The actual deployment or first finalized transaction has not been announced yet, so the watch remains open and we shouldn't resolve it or fire it yet. What about the watch `Nvidia Q4 FY2027 Gross Margin Bottoming (71%-72%)` (thread ID: `01a0409b-bb2f-7d8f-a4a6-e30c93fff0e0`)? Its description: "Verify if Nvidia's gross margins bottom out in the guided 71%-72% range in Q4 FY2027, or if memory price pressures push margins lower." Since Q4 FY2027 ends in January 2027 and reports in February 2027, this has not happened yet. The guidance was just given, so the watch remains open. What about `Nvidia FY2028 Revenue Growth (Guided ~70%)` (thread ID: `01a0409b-c90b-7527-a2c7-f7064fc17243`)? Its description: "Verify if Nvidia's full-year FY2028 revenue meets or exceeds the guided 70% growth rate as supply constraints are resolved." This will report in February 2028, so it remains open. Let's double check if there are any other watches we can resolve or record events on. What about `Nvidia >15% Server Price Hikes Effective and Gross Margin Impact` (thread ID: `01a02c01-239b-7dfa-b8af-490df009beb4`)? Its description: "Verify if the >15% server price hikes are successfully implemented starting early 2027 and whether they protect or expand Nvidia's gross margin above 74% in subsequent quarters." This will fire in early 2027, so it remains open. Let's see if there are any other open threads or watches. Wait, let's look at the list of threads again. Is there any other thread? No, we have recorded the event on the ongoing memory pricing watch. Let's double-check if there are any new watches we should register. Wait! We found that Nvidia paused its $36 billion NeoCloud revenue-sharing credit-support model on August 27, 2026. This was a massive development. Should we register a watch for when Nvidia resumes, revamps, or officially cancels this program? Yes, the WSJ reported that Nvidia "could still revamp the initiative later or fold it into another program." Let's register a watch for: - Title: `Nvidia NeoCloud Revenue-Sharing Program Revamp or Cancellation` - Watch type: `tripwire` or `thesis_breaker` (or `live_process`) - Cardinality: `one_shot` - Description: `Monitor whether Nvidia officially cancels, revamps, or resumes its $36 billion NeoCloud revenue-sharing and credit-support program, or folds it into the $500 billion Wall Street platform.` Let's call `record_watch`. Let's do that.

This is great! We have recorded the watch. Let's review if we need to open any other threads. Wait! Is there any open question we couldn't fully answer? No, we have successfully gathered primary evidence, cited it fully, and written comprehensive updates to our notes. We've also recorded a significant event on an ongoing watch and registered a new watch tracking the NeoCloud program's future. Let's do a quick final review of our findings to make sure they adhere to all guidelines: - Name and link the actual thing (done). - Carry 1-2 representative verbatim quotes, copied exactly, each kept with the link it came from (done). - Say what it means (done). - No YAML frontmatter (done). - Title is first `# ` heading (done). - Slug is stable and reused where appropriate (done). - No summary-note or roll-up note (done). - Wikilinks used to link findings (done). Everything looks absolutely perfect. I'm ready to call `finish_cycle` and conclude this research cycle.

The research cycle is now complete. Here is a summary of the findings and actions taken: ### 1. Key Findings & Note Updates We updated three standing findings under their stable slugs to incorporate the massive, breaking developments from August 2026: * **Big Tech's AI Capex Surpasses $800 Billion in 2026 on Track for $1.3 Trillion in 2027 as Nvidia Projects 70% Growth** (`big-tech-capex-reaches-710b-in-2026`): * **The Update**: Incorporated Nvidia's record Q2 FY2027 earnings (reported August 26, 2026) of **$96.22 billion in revenue** (up 105.9% YoY) and **$89.0 billion in Data Center revenue** (up 117% YoY). * **The Capex Story**: Confirmed the capex story is fully intact and scaling even higher. CFO Colette Kress revealed that capital expenditures by the top five hyperscalers are now projected to reach **nearly $800 billion in 2026 and surge to $1.3 trillion in 2027**, backed by a cloud industry backlog exceeding $2 trillion. Nvidia's initial full-year FY2028 guidance is set at ~70% YoY growth under a "supply-constrained outlook." * **Hyperscaler Spending**: Combined calendar Q2 2026 capex for the top four hyperscalers (Amazon, Alphabet, Microsoft, Meta) hit a record **$166.01 billion** (up 87% YoY and 27.1% QoQ). * **Off-Balance-Sheet Joint Ventures and the New AI Infrastructure Financing Model: Nvidia's $500 Billion Wall Street Platform and the $105 Billion OpenAI Guarantee** (`off-balance-sheet-ai-infrastructure-financing-models`): * **The $500B PE Alliance**: Detailed Nvidia's August 10, 2026 announcement of a strategic partnership with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to establish "compute financing platforms" designed to mobilize over $500 billion of third-party capital. This treats compute as a yielding infrastructure asset class to fund AI factories. * **The NeoCloud Revenue-Sharing Pause**: Detailed the dramatic pausing of Nvidia's **$36 billion NeoCloud credit-support and revenue-sharing program** on August 27, 2026 (WSJ report), just one day after the earnings call. Under this model, Nvidia provided credit support and minimum revenue guarantees in exchange for a 50% cut of future rental revenues. The program was paused due to internal antitrust concerns (dictating customer operations), partner pushback, and investor fears of "circular financing" artificially inflating chip demand. * **Nvidia's AI Margin Protection: HBM Cost Inflation Triggers Gross Margin Reset as Samsung Mounts Custom HBM4E Challenge** (`nvidia-server-price-hikes-and-margin-protection`): * **The Margin Reset**: Documented that Nvidia guided its Q4 FY2027 gross margins to bottom in the **71% to 72% range** (down from 75.0% in Q2) due to severe High Bandwidth Memory (HBM) cost inflation. * **The Quote**: CFO Colette Kress stated: *"The magnitude of the price increase has exceeded our prior expectations and are headed even higher into next year."* * **The Mitigation**: Nvidia plans to implement **over 15% server price hikes** starting in early 2027 to protect margins and push full-year FY2028 gross margins back to the guided 72%-73% range. ### 2. Thread & Watch Actions * **Recorded Event**: Triggered a status update on the ongoing watch `HBM Memory Component Pricing and Supply Bottlenecks` (thread ID: `019fbaba-2a7b-7e80-b885-9f29c3a7f0ee`) to record that Nvidia has officially guided its margins down due to memory price increases exceeding expectations. * **Registered New Watch**: Created a new watch `Nvidia NeoCloud Revenue-Sharing Program Revamp or Cancellation` (thread ID: `01a0838f-85f4-7546-9aa6-96928bb600cf`) to track whether Nvidia officially cancels, revamps, or resumes its $36 billion NeoCloud credit-support initiative, or folds it into the $500 billion Wall Street platform.