Cycle on Jul 17, 2026

I will start by listing the open threads and reading the existing notes for this topic to understand our current baseline and any outstanding questions.

I will list the current research notes to understand the existing findings on this topic.

I will call `market_view` for Nvidia (NVDA) and the macro backdrop to anchor our assessment of Nvidia's AI capex story with the latest public financial and market data.

I will perform a web search to check for the latest news on Nvidia's AI capex, big tech hyperscaler capex updates, and any recent earnings reports or market indicators concerning Nvidia's Blackwell/Rubin demand in July 2026.

Let's fetch the first few results or do more targeted searches to understand what the recent developments in June and July 2026 are regarding Nvidia's AI capex story, and whether there are any new earnings or updates. Wait, let's look at the date: July 17, 2026. This means we are in the middle of earnings season for Q2 2026 (calendar Q2), which is happening now or about to happen. Let's search for "Nvidia AI capex story intact July 2026" or "TSMC Q2 2026 earnings" or "ASML Q2 2026 earnings" (which usually happen around mid-July) or "hyperscaler capex Q2 2026". Let's do a search.

This is incredible. Yesterday (July 16, 2026), TSMC (Taiwan Semiconductor Manufacturing Company) reported its Q2 2026 earnings. Let's fetch the TSMC earnings articles to get the exact numbers, quotes, and details. Let's do a `web_fetch` on: 1. `https://seekingalpha.com/article/4922631-nvidia-the-tsmc-earnings-release-was-a-warning-shot` 2. `https://www.techtimes.com/articles/320696/20260716/tsmc-posts-record-quarter-ai-chip-demand-pushes-full-year-growth-outlook-past-40.htm` 3. `https://247wallst.com/investing/2026/07/16/nvidias-75-margins-vs-tsmcs-foundry-gamble-two-paths-to-ai-dominance-one-winner/` Let's fetch `https://seekingalpha.com/article/4922631-nvidia-the-tsmc-earnings-release-was-a-warning-shot` first.

Let's fetch `https://www.techtimes.com/articles/320696/20260716/tsmc-posts-record-quarter-ai-chip-demand-pushes-full-year-growth-outlook-past-40.htm` to see the details of TSMC's record quarter, its capex guidance, and its AI chip outlook.

This is extremely rich and useful data. Let's do another search to see if there are any other recent articles specifically on Nvidia's Blackwell GPU shipments, potential delays, or if the capex story is fully intact. Let's search for "Nvidia Blackwell shipment schedule July 2026" or "Nvidia Blackwell delays 2026". Let's do a search.

This is incredibly interesting. There are rumors of a delay regarding Nvidia's upcoming Rubin and Kyber products, which SemiAnalysis alleged was a 12-month delay in the Kyber NVL144 rack system, but Nvidia has officially denied these rumors, stating that its chip roadmap is intact. Let's fetch: 1. `https://wccftech.com/nvidia-quashes-rubin-kyber-rack-delay-rumors-says-chip-roadmap-is-intact/` 2. `https://www.tomshardware.com/pc-components/gpus/nvidias-kyber-rack-for-rubin-ultra-slips-to-2028` 3. `https://www.barrons.com/articles/nvidia-stock-price-rubin-chips-1988209e` (or equivalent if we can find it) Let's fetch `https://wccftech.com/nvidia-quashes-rubin-kyber-rack-delay-rumors-says-chip-roadmap-is-intact/` first.

Let's fetch `https://www.tomshardware.com/pc-components/gpus/nvidias-kyber-rack-for-rubin-ultra-slips-to-2028` to get the technical specifics of what SemiAnalysis claimed about the Kyber NVL144 rack system and why they thought it was delayed, as well as Nvidia's response. This is a crucial detail for understanding the debate around Nvidia's roadmap.

Ah, the page for Tom's Hardware loaded a paywall or subscription prompt. Let's do a search on Google to see if we can find more details about SemiAnalysis's claims regarding the PCB midplane problems or Kyber rack delays, or let's search for "SemiAnalysis Kyber rack Nvidia Rubin delay midplane". This will help us understand the debate.

Let's fetch `https://www.cnbc.com/2026/07/06/nvidia-kyber-rack-system-delays-manufacturing-taiwan-rubin-chips-.html` or `https://www.datacenterdynamics.com/en/news/nvidia-pushes-kyber-release-to-2028-following-manufacturing-concerns-report/` to read the exact details of the SemiAnalysis report and Nvidia's denial. Let's try `https://www.cnbc.com/2026/07/06/nvidia-kyber-rack-system-delays-manufacturing-taiwan-rubin-chips-.html` first.

This is very interesting. Let's summarize: 1. **TSMC Q2 2026 Earnings (July 16, 2026)**: - Record-breaking quarter: $40.2 billion in revenue, up 36% YoY. Net profit surged 77.4% NT$706.56B. - Raised full-year revenue growth outlook to "above 40%" (from "above 30%"). - Raised its 2026 capital expenditure (CapEx) budget to $60 to $64 billion, up from $52 to $56 billion. Underlining that "the CapEx in the next three years will be even more significantly higher than the past three years." - CoWoS advanced packaging remains the primary bottleneck with lead times extending 52 to 78 weeks. - Advanced nodes (7nm and below) represented 77% of all wafer revenue, with N3 at 30% and N5 at 33%. - N2 (2nm) achieved its first commercial revenue contribution (3% of wafer revenue). - Agentic AI is emerging as a second demand channel, driving significant CPU core demand (shifting CPU-to-GPU server deployment ratio from 1:8 toward 1:1 for some agentic workloads). 2. **Nvidia's Product Roadmap and the Kyber Rack Controversy (July 2026)**: - On July 6, 2026, SemiAnalysis reported that Nvidia's next-generation Kyber NVL144 rack-scale architecture (designed to house its 2027 Rubin Ultra chips) has been delayed by more than 12 months to 2028. - The reported delay is caused by manufacturing difficulties with the PCB midplane (a specialized, multi-layer printed circuit board). - SemiAnalysis also claimed that a stopgap solution (bolting two current-generation racks together, "NVL72x2") was cancelled due to heavy pushback from CSPs and hyperscalers over design and operational burdens. They argued this leaves Nvidia with "no proven solution to expand the scale-up world size for Rubin Ultra." - Nvidia immediately and officially denied these rumors, stating that its AI and chip roadmap remains intact ("Our roadmap is intact"). - Analysts (such as Paul Triolo of DGA-Albright Stonebridge Group) notes that Nvidia has faced similar challenges before (like Blackwell design flaws/delays that were resolved before volume production) and that power supply is the real binding constraint for US AI data center expansion anyway, meaning a minor hardware delay might align better with power grid readiness. - Furthermore, Nvidia's current-generation Rubin systems (the base Vera Rubin platform, which ships in Oberon racks) are in full production and will begin shipping in Fall 2026 to eight partners (AWS, Azure, Google Cloud, etc.). - SemiAnalysis itself projects Nvidia's data-center compute revenue will run 20% above Wall Street consensus in the second half of fiscal 2027, showing that near-to-medium-term demand remains blockbuster despite the long-term debate. Wait, let's look at the existing notes and slugs. We have: - `nvidia-record-q1-results-and-trillion-dollar-order-book` (Nvidia's AI Capex Story Remains Intact Supported by Robust Blackwell Pricing and Blockbuster Financial Execution) - `big-tech-capex-reaches-710b-in-2026` (Big Tech's AI Capex Reaches Historic Scale as 2026 Infrastructure Sprint Approaches $690 Billion) - `custom-silicon-and-datacenter-hardware-competition` (Custom Silicon and Data Center Hardware Competition Escalates to Challenge Nvidia's Moat) Let's double-check if there are any other notes we should update. The topic is: "Is Nvidia's AI capex story still intact?" This is a perfect fit for: 1. `nvidia-record-q1-results-and-trillion-dollar-order-book` (which we can update to reflect the TSMC earnings, the Kyber rack controversy, the current status of Rubin and Blackwell, and the overall robustness of the capex cycle). 2. `big-tech-capex-reaches-710b-in-2026` (which we can update if there's new data on hyperscaler spending or TSMC's massive capex boost). Wait, let's see if there are other details we should search for. Let's do a search on ASML or other indicators. Or let's search for "hyperscaler capex Q2 2026 earnings" to see if Microsoft, Google, Meta, or Amazon have reported yet. (Usually they report in late July or early August, so they might not have reported Q2 yet as of July 17, 2026, but let's check). Let's search.

This is exceptionally valuable. Let's see: 1. **Big Tech's AI Capex**: - Combined capital expenditure for the "Big Four" (Alphabet, Microsoft, Meta, Amazon) is projected at **$725 billion** in 2026 (up 77% from $410 billion in 2025). Some estimates include Oracle and put the total above $725 billion (or $800B+ for 2027 per Citi). - Citi is projecting combined 2027 capital expenditure across Alphabet, Meta, and Amazon to reach **over $800 billion** (specifically $801 billion), reflecting that the spending is accelerating, not slowing down, heading into 2027. - Amazon is raising $25 billion through a bond sale to fund its nearly $200 billion in AI and AWS infrastructure investment in 2026. - Alphabet, Microsoft, Meta, and Amazon are approaching their Q2 2026 earnings season (late July / early August). Wait, let's look at the threads list. We have: - `019f10c8-53dd-729c-926a-27e735d10910`: `Alphabet Q2 2026 Earnings and Dilution Impact Tracking` (fire date July 25, 2026). This is still in the future (today is July 17, 2026). - `019ed2f6-f136-7051-a5b0-fa3b26ec02e0`: `Nvidia Q2 FY2027 Earnings Release` (fire date August 20, 2026). This is also in the future. - `019edd49-1ccb-73d3-9fc0-80ae266f32c4`: `Nebius Group Q2 2026 earnings and margin trajectory` (fire date August 15, 2026). In the future. Wait! Are there any threads we can resolve or watches that have fired? None of the listed watches have fired yet, because their fire dates are in late July, August, October, November, December 2026, or 2027/2028. Let's review the watches: - `Qualcomm Dragonfly C1000 commercial availability` (2028-01-15) - `Qualcomm HBC Gen 1 and AI250 commercial sampling` (2027-06-15) - `Alphabet Q2 2026 Earnings and Dilution Impact Tracking` (2026-07-25) - `Alphabet $40B ATM Stock Offering Execution` (ongoing) - `OpenAI Jalapeño chip initial production deployment` (2026-12-15) - `Oracle $40B Debt and Equity Financing Plan for FY2027` (ongoing) - `Anthropic public S-1 filing and IPO debut` (2026-11-15) - `Nebius Group Q2 2026 earnings and margin trajectory` (2026-08-15) - `Nvidia Q2 FY2027 Earnings Release` (2026-08-20) - `Meta Platforms potential equity raise announcement and pricing` (one-shot) - `Nebius 850 MW active power capacity by YE 2026` (2026-12-31) - `CoreWeave quarterly capex tracking` (ongoing) - `Vera Rubin platform broader commercial shipments in H2 2026` (2026-11-01) - `RTX Spark laptops and desktops retail launch` (2026-10-01) Wait, can we register any new watches? We can register a watch for Nvidia's Kyber rack development and Rubin Ultra shipment schedule, since SemiAnalysis reports it is delayed to 2028 but Nvidia says its roadmap is intact. We can set up a watch to track the first official commercial shipment or verification of Rubin Ultra / Kyber NVL144 systems, expected in 2027/2028. Let's write two very solid updates to our notes: 1. `nvidia-record-q1-results-and-trillion-dollar-order-book` -> We will update this note to include: - TSMC's record Q2 2026 earnings on July 16, 2026, which reported $40.2B revenue (36% YoY growth), raised full-year growth outlook to "above 40%", and raised 2026 capex to $60-$64B. This directly confirms that Nvidia's AI capex story remains highly intact and is structurally supported. - The debate on Nvidia's next-gen Kyber NVL144 rack-scale architecture (for 2027's Rubin Ultra). SemiAnalysis's July 6 report claimed a 12-month delay to 2028 due to PCB midplane manufacturing challenges, which Nvidia officially denied, stating its roadmap is intact. - Nvidia's current-generation Rubin systems (Oberon racks) are in full production and shipping in Fall 2026. - The emerging "agentic AI" demand channel which increases CPU-to-GPU ratios to 1:1 for some workloads, providing a secondary demand growth avenue. - Cite the new market view for NVDA: `/markets/NVDA/2026/07/17` showing TTM revenue of $253.49B, gross margins of 74.1%, and a $5.00T market capitalization. 2. `big-tech-capex-reaches-710b-in-2026` -> We will update this note to include: - The latest estimates showing that combined capex across Alphabet, Microsoft, Meta, and Amazon is projected at **$725 billion** in 2026 (up 77% YoY). - Citi's mid-July 2026 projection of **over $800 billion** ($801B) in combined capex for Alphabet, Meta, and Amazon in 2027, demonstrating that the infrastructure buildout is accelerating rather than plateauing. - Amazon's $25 billion bond sale in July 2026 to fund its nearly $200 billion AI and AWS infrastructure investment. - TSMC's massive capex boost to $60-$64 billion for 2026, with CEO C.C. Wei announcing that capex for the next three years will be "even more significantly higher than the past three years." Wait, let's verify if there are any other notes we should update or if we should write a new note. No, the topic is "Is Nvidia's AI capex story still intact?", so updating these two notes is exactly what is needed. Let's read the current content of `nvidia-record-q1-results-and-trillion-dollar-order-book` to make sure we integrate the updates seamlessly. I will fetch the note text. Wait, we don't have a direct tool to read a specific note's content other than `read_notes` which gives a preview. Is there another way? No, `read_notes` returns the preview, and we can overwrite it or write a comprehensive update. Let's make sure the update is detailed and fully comprehensive, carrying forward the previous context while updating it with the latest July 2026 data. Let's do a quick search on the exact details of Nvidia's Q1 FY2027 earnings (which ended April 30, 2026) to make sure we keep the key historical numbers accurate. From the `market_view` of `NVDA`: - Quarter ending 2026-04-30: revenue $81.61B, operating income $53.54B, net income $58.32B, EBITDA $71.00B. Gross margin 74.1%. - TTM Revenue: $253.49B. - EPS for the quarter ending April 30, 2026: $1.87 reported vs $1.77 est (beat). - Market cap: $5.00 trillion. Let's write the updated note for `nvidia-record-q1-results-and-trillion-dollar-order-book`. Wait, the title is "Nvidia's AI Capex Story Remains Intact Supported by Robust Blackwell Pricing and Blockbuster Financial Execution". Let's keep the title or enhance it slightly, but we must reuse the slug `nvidia-record-q1-results-and-trillion-dollar-order-book`. Let's draft the note for `nvidia-record-q1-results-and-trillion-dollar-order-book`: We will write about: - Nvidia's overall financial strength: TTM revenue of $253.49B, gross margins of 74.1%, and a $5.00T market capitalization as of July 17, 2026 (citing `/markets/NVDA/2026/07/17`). - TSMC's record Q2 2026 earnings on July 16, 2026, which serves as a powerful proxy for Nvidia's demand and supply. TSMC reported Q2 revenue of $40.2B (up 36% YoY) and raised its full-year growth outlook to "above 40%". Its 2026 CapEx was raised to $60-$64B, with management declaring that CapEx over the next three years will be "even more significantly higher" than the prior three years. This acts as a resounding validation that the AI capex cycle is accelerating, not plateauing. - The CoWoS packaging constraint: TSMC's CoWoS advanced packaging platform remains the primary bottleneck for AI accelerators, with lead times of 52 to 78 weeks, confirming that supply, not demand, remains the binding limit on Nvidia's growth. - The Kyber NVL144 Rack Controversy: On July 6, 2026, SemiAnalysis claimed Nvidia's next-generation Kyber NVL144 rack-scale architecture (slated for 2027's Rubin Ultra chips) is delayed by more than 12 months to 2028 due to PCB midplane manufacturing challenges. Nvidia officially denied this, stating its roadmap is intact. Even with this debate, Nvidia's base Vera Rubin systems (using Oberon racks) remain in full production and are on track to ship in Fall 2026. - The Agentic AI Shift: TSMC flagged agentic AI as a second demand channel, driving massive CPU core demand (shifting CPU-to-GPU ratios towards 1:1 for some workloads), which benefits Nvidia's broader product portfolio, including its newly launched Vera CPU. Let's write this note first. I will state my intent and call `write_note`.

Now, let's write an update to `big-tech-capex-reaches-710b-in-2026` to reflect the latest numbers: - Combined capex across the top hyperscalers (Alphabet, Microsoft, Meta, Amazon) is projected to reach **$725 billion** in 2026 (up 77% YoY from $410 billion in 2025). - Citi’s mid-July 2026 projection of **over $800 billion** ($801B) in combined capex for Alphabet, Meta, and Amazon in 2027, demonstrating that the infrastructure buildout is accelerating rather than plateauing. - Amazon's $25 billion bond sale in July 2026 to fund its nearly $200 billion AI and AWS infrastructure investment. - TSMC's massive capex boost to $60-$64 billion for 2026, with CEO C.C. Wei announcing that capex for the next three years will be "even more significantly higher than the past three years." - Reference the upcoming Q2 2026 earnings reports in late July/early August, where market participants will watch for any signs of capex moderation or cash flow strains. Let's do this by updating the note under the slug `big-tech-capex-reaches-710b-in-2026`. I will state my intent and call `write_note`.

Let's review if there are any other notes we should update or if we should register a new watch. Wait, let's look at the Kyber NVL144 rack controversy. Since SemiAnalysis claims that Kyber NVL144 is delayed to 2028, and Nvidia says its roadmap is intact, this is a perfect trigger for a forward-looking watch! Let's call `record_watch` to register a watch on this. What type of watch? - `watch_type`: `tripwire` or `thesis_breaker` or `dated_catalyst` or `live_process` - let's make it a `thesis_breaker` or `tripwire` with cardinality `one_shot`. - Title: `Nvidia Kyber NVL144 rack commercial shipment verification` - Description: `Verify if Nvidia begins commercial shipments of the Kyber NVL144 rack-scale systems by late 2027 or if SemiAnalysis's predicted delay to 2028 is borne out.` - Entity: `NVDA` - Fire Date: `2027-12-31T00:00:00Z` Let's call `record_watch` to set this up.

This is excellent. The watch is successfully registered. Let's see if there are any other threads or actions to take. We have listed threads, read notes, updated the two relevant notes, and registered a forward-looking watch. Since we have fully addressed the brief "Is Nvidia's AI capex story still intact?" with concrete, newly surfaced evidence (TSMC's earnings from yesterday, July 16, 2026, and the Kyber rack debate from July 6, 2026), we are ready to call `finish_cycle`. Let's do that now.

I have completed this research cycle, updating our core findings with powerful, highly recent evidence from July 2026. Here is a summary of the key insights and actions taken during this cycle: ### 1. Fundamental Anchoring: TSMC's Record Q2 2026 Earnings (July 16, 2026) TSMC's earnings release functions as the ultimate real-time proxy for AI chip demand and supply, confirming that the AI capex supercycle is accelerating rather than plateauing: * **Record Financial Performance:** TSMC reported Q2 revenue of **$40.2 billion** (36% YoY growth) and a net profit surge of 77.4% to NT$706.56 billion. * **Upward Guidance Revision:** TSMC raised its full-year revenue growth outlook to **"above 40%"** (up from "above 30%"). * **Aggressive CapEx Boost:** TSMC raised its 2026 CapEx budget to **$60–$64 billion** (from $52–$56 billion), with CEO C.C. Wei noting that capital spending over the next three years will be *"even more significantly higher than the past three years."* * **CoWoS Bottleneck:** TSMC's Chip-on-Wafer-on-Substrate (CoWoS) packaging remains the primary bottleneck, with lead times of **52 to 78 weeks**, confirming that physical supply, rather than customer demand, is the binding limit on Nvidia's near-term growth. ### 2. The Kyber NVL144 Rack Controversy (July 6, 2026) A central debate has emerged regarding Nvidia's next-generation hardware execution: * **SemiAnalysis Claims:** On July 6, research firm SemiAnalysis reported that Nvidia's next-generation **Kyber NVL144 rack-scale architecture** (housing its 2027 Vera Rubin Ultra chips) has been delayed by over 12 months to 2028 due to manufacturing challenges with the multi-layer PCB midplane. * **Nvidia's Denial:** Nvidia immediately rejected the report, stating: *"Our roadmap is intact."* * **Near-Term Mitigation:** Nvidia's base Vera Rubin systems (using Oberon racks) remain in full production and are on track to begin shipping in Fall 2026 to eight major partners. * **Watch Registered:** I registered a forward-looking watch to track and verify the commercial shipment timeline of the Kyber NVL144 rack systems by late 2027. ### 3. Big Tech Capex Supercycle Scaling * **Historic Projections:** Combined 2026 capital expenditures across Alphabet, Microsoft, Meta, Amazon, and Oracle are pacing toward **$725 billion** (up 77% YoY). * **2027 Acceleration:** Citi projected that combined capital expenditures for just Alphabet, Meta, and Amazon will exceed **$801 billion** in 2027. * **Debt Funding:** Amazon initiated a **$25 billion bond sale** in July 2026 to help fund its estimated $200 billion in annual AI and AWS infrastructure investments. ### 4. Updates & Operations Completed * **Updated Note:** `nvidia-record-q1-results-and-trillion-dollar-order-book` (Nvidia's AI Capex Story Remains Intact Supported by Robust Blackwell Pricing and Blockbuster Financial Execution) — fully updated with TSMC Q2 earnings, the Kyber rack controversy, and the emerging agentic AI demand channel. * **Updated Note:** `big-tech-capex-reaches-710b-in-2026` (Big Tech's AI Capex Reaches Historic Scale as 2026 Infrastructure Sprint Approaches $725 Billion) — updated with the $725B 2026 projection, Citi's $801B 2027 projection, Amazon's bond sale, and upcoming Q2 earnings expectations. * **Registered Watch:** `Nvidia Kyber NVL144 rack commercial shipment verification` (`tripwire` / `one_shot` with a target fire date of December 31, 2027).