Cycle on Jul 19, 2026

I am going to check the active research threads and watches for this topic to see if there are any specific questions or tripwires that have been previously set up.

I will call the `market_view` tool to retrieve the latest financial performance data, valuation, and market consensus for Nvidia (NVDA) to ground our analysis.

I will also retrieve macro public market details to see what the rates, inflation, and general environment look like. This will help see if macroeconomic trends are influencing hyperscaler capex behavior.

Now I will run some web searches to understand the current state of Nvidia's AI capex story. It's late July 2026. Let's see if there are updates on major hyperscalers' capex forecasts or spending, any recent reports on Nvidia's Blackwell, Rubin, or supply chain (TSMC, CoWoS, HBM), and any state-level data center moratoriums or power grid issues. Let's search for "hyperscaler capex Q2 2026 earnings" or "Microsoft Meta Alphabet Amazon capex Q2 2026" or "Nvidia capex story intact July 2026". Let's also check if there are recent earnings reports or updates from the major hyperscalers. Wait, the `Hyperscaler Capex Guidance Track (July 2026 Earnings)` thread mentions "Track the quarterly capex guidance and actual capex spending reported by the major AI hyperscalers (Microsoft, Meta, Alphabet, and Amazon) during the July 2026 earnings calls. This is the single most critical forward-looking signal for the AI capex story." The current date is July 19, 2026. The major earnings calls for Q2 2026 (or calendar Q2 2026) are likely happening now or in the next week or two (late July / early August). Let's see if any have already reported or if there are pre-earnings expectations and reports. Let's do a search.

Let's search for "hyperscaler capex Q2 2026 earnings" or "Google Q2 2026 earnings capex", "Microsoft Q4 FY2026 earnings capex" or "Meta Q2 2026 earnings capex" or "Amazon Q2 2026 earnings capex" to see if any of these companies have reported in July 2026. Note that today is Sunday, July 19, 2026. Let's see if any earnings have occurred in the last week or two, or if they are upcoming. Let's do a search for "Alphabet Q2 2026 earnings date" or "Microsoft Q4 FY2026 earnings date". Let's run a search.

Ah, Alphabet reports its second-quarter 2026 financial results on Wednesday, July 22, 2026 (or Tuesday July 21, 2026). Wait! Today is Sunday, July 19, 2026, so Alphabet hasn't reported Q2 2026 earnings yet (they report in 2-3 days, on July 22, 2026, or July 21, 2026). Let's find out if there are other major occurrences in the AI capex world. What about New York's July 14, 2026 moratorium on large AI data centers? Let's search for "New York data center moratorium July 2026" or "New York data center moratorium July 14 2026" to see what that regulatory hurdle was, as mentioned in the thread `State-Level AI Data Center Moratorium Expansion Track` ("following New York's July 14, 2026 moratorium"). Let's check this.

This is a major and highly relevant news item. Let's fetch the Reuters or CNBC article or the Governor's Executive Order to get the details of this New York moratorium, which occurred on July 14, 2026. This is a crucial risk factor for the physical data center infrastructure capex. Let's fetch the Reuters or CNBC article. Let's fetch `https://www.reuters.com/world/new-york-becomes-first-state-impose-data-center-moratorium-2026-07-14/` first, or if that fails, try others. Let's see.

This is a highly relevant, concrete development that occurred on July 14, 2026. Let's see if there are other states or federal bodies imposing similar restrictions or if there is research on how this affects hyperscaler capex and Nvidia's sales. Wait, let's look at the other active threads and watches: 1. `Hyperscaler Capex Guidance Track (July 2026 Earnings)` (id: `019eb0cd-20a0-71c6-b91e-92d528f6792e`). It is currently July 19, 2026. Let's check when the major hyperscalers are reporting their Q2 2026 earnings. Let's run a search for "Microsoft Q4 FY2026 earnings date", "Meta Q2 2026 earnings date", "Amazon Q2 2026 earnings date". This will tell us if any have reported or are about to. Let's search.

Okay, so let's map out the upcoming earnings dates for the major hyperscalers: - Alphabet (GOOGL): Wednesday, July 22, 2026 (or July 21, 2026, depending on sources, but confirmed for July 22, 2026 by IG and Alphabet's investor site). - Microsoft (MSFT): Wednesday, July 29, 2026. - Amazon (AMZN): Thursday, July 30, 2026. - Meta (META): Presumably around late July / early August (usually reports same week as Microsoft/Amazon/Alphabet). Let's search for "Meta Q2 2026 earnings date". Let's run a search.

These dates are confirmed. They are ALL in late July: - Alphabet (GOOGL): July 22, 2026 - Meta (META): July 29, 2026 - Microsoft (MSFT): July 29, 2026 - Amazon (AMZN): July 30, 2026 Wait, are there any other major updates that have happened in early/mid July 2026, or late June 2026, regarding Nvidia's AI capex sustainability? Let's search for "Nvidia capex sustainability" or "Nvidia Blackwell demand July 2026" or "Nvidia Blackwell delays" or "Kyber rack manufacturing delays" or "TSMC CoWoS capacity 2026". Let's run a search.

Let's check the CNBC article `https://www.cnbc.com/2026/07/06/nvidia-kyber-rack-system-delays-manufacturing-taiwan-rubin-chips-.html` or other similar sources on the Kyber NVL144 delay. Wait, there's a CNBC article from July 6, 2026: "The Kyber NVL144 has been pushed to 2028 due to difficulties in manufacturing a key circuit board: SemiAnalysis. The reported delay adds to..." This is extremely relevant. Let's fetch this URL to verify the details. Let's do a fetch.

Ah, the fetch failed with a proxy error. Let's try to search or find another link that has the same details, or search for "Nvidia Kyber NVL144 delay SemiAnalysis July 2026" and fetch another site, or use different keywords. Let's do a search first.

Excellent. Let's fetch some of these articles, which will give us deep insight into the Kyber NVL144 delay. Let's try fetching `https://www.datacenterdynamics.com/en/news/nvidia-pushes-kyber-release-to-2028-following-manufacturing-concerns-report/` or `https://www.tomshardware.com/pc-components/gpus/nvidias-kyber-rack-for-rubin-ultra-slips-to-2028`. Let's try Tom's Hardware first.

Ah, the Tom's Hardware fetch returned a page that is mostly subscription and signup text, rather than the article itself. Let's try fetching `https://www.datacenterdynamics.com/en/news/nvidia-pushes-kyber-release-to-2028-following-manufacturing-concerns-report/` or `https://www.techtimes.com/articles/320033/20260709/nvidia-kyber-nvl144-slips-2028-physics-not-software-defeats-backup-plan.htm`. Let's try Datacenter Dynamics first.

This is a spectacular find. This is a very recent, highly specific update from July 6, 2026. Let's analyze the details of this: - **Kyber NVL144 delay**: Nvidia's Kyber rack-scale architecture has been delayed by more than 12 months, pushing its release date to 2028 (originally slated for 2027). - **PCB midplane issues**: The delay is caused by difficulties in manufacturing the printed circuit board (PCB) midplane that connects the modules together. - **Alternative design scrapped**: Nvidia's alternative plan—placing two Oberon racks back-to-back—was scrapped after hyperscalers/cloud companies pushed back due to its "odd design and heavy operational burden." - **NVL576 impacted**: The manufacturing issues also impact the planned NVL576 system (connecting 8x Oberon racks over CPO between NVSwitches), which will likely be delayed or limited to small volumes. - **Rubin Ultra modified**: Just a week before (late June/early July 2026), SemiAnalysis reported that Nvidia canceled the four-die Rubin Ultra design in favor of a two-chiplet design, also due to "manufacturing execution concerns," delivering around half the performance of the four-die Rubin Ultra. - **Nvidia's response**: "Our roadmap is intact." This perfectly addresses the watch `Nvidia Q2 FY2027 Earnings - Rubin Ultra / Kyber NVL144 Timeline Update` (and the prior findings regarding TSMC's historic $64B capex surge colliding with Kyber rack manufacturing delays). Wait! Let's check another thread/watch: `State-Level AI Data Center Moratorium Expansion Track` (id: `019f6f55-61e5-71a9-898a-2a6088793817`). New York passed the moratorium on July 14, 2026 (Executive Order 62). Let's check if other states (like Virginia, Georgia, Ohio, Texas) or federal regulatory agencies have announced similar moratoriums or restrictive grid/power regulations recently. Let's do a search for "data center moratorium Georgia Ohio Texas Virginia July 2026" or "restrictive grid power regulations data centers 2026". Let's run a search.

Let's do a search on `https://www.ncsl.org/fiscal/which-states-are-banning-data-centers` or `https://www.multistate.us/insider/2026/2/20/state-data-center-legislation-in-2026-tackles-energy-and-tax-issues` or `https://www.interconnectedcapital.com/research/data-center-moratoriums` to get more details on the state-level bills and moratoriums. Let's search for "Interconnected Capital data center moratorium tracker 2026" or "multistate data center legislation 2026". Let's run a search.

Let's fetch `https://finance.yahoo.com/sectors/technology/articles/ai-data-center-bans-rapidly-123102981.html` or `https://www.businessinsider.com/data-center-bans-moratoriums-opposition-map-2026-6` to get more details on how widespread these bans are. Let's try fetching the Business Insider article first if we can, or search for Yahoo Finance. Let's do a search to get a direct Yahoo Finance or another fetchable link. Let's search for "AI data center bans are rapidly multiplying across the US Yahoo Finance".

Let's fetch `https://www.tomshardware.com/tech-industry/artificial-intelligence/ai-data-center-bans-are-rapidly-multiplying-across-the-us-69-jurisdictions-block-new-builds-with-four-moves-noted-as-permanent` if it is readable, or try searching for "AI data center bans are rapidly multiplying across the US" on other platforms. Let's do a fetch on the Tom's Hardware link first. Wait, Tom's Hardware pages can sometimes be subscription-blocked, but let's see. Let's fetch it.

Ah! The web_search budget is spent, which is a good reminder to focus on the information we have fetched and synthesized. Let's review what we have: 1. **Nvidia's financial trajectory & market view**: - Market Cap: $4.91 trillion (as of July 19, 2026). - TTM Revenue: $253.49 billion (up 85.2% YoY). - Gross Margin: 74.1%. - Latest quarter ending April 30, 2026: reported $81.61B revenue (beat est. of $81B+ or similar) and $58.32B net income, EPS $1.87 vs $1.77 est. - Next earnings expected around August 25, 2026 (for Q2 FY2027) with revenue guidance of $91.0 billion. - 10-year Treasury is at 4.57% and Fed funds rate is at 3.63%. 2. **Hyperscaler Q2 2026 Earnings & Capex Outlook**: - The major hyperscalers are all scheduled to report their Q2 2026 earnings in late July: - Alphabet (GOOGL): July 22, 2026. - Meta (META): July 29, 2026. - Microsoft (MSFT): July 29, 2026. - Amazon (AMZN): July 30, 2026. - Prior to these earnings, estimates for 2026 capex spending across these players have surged to historic levels, with some firms like Yahoo Finance and Introl Blog reporting that combined CapEx of the "Big Four" (Amazon, Alphabet, Meta, Microsoft) is expected to reach $600B to $725B in 2026 (a massive increase of 36% to 70% YoY). For instance, Amazon is guided to ~$200B, Microsoft to ~$190B, Alphabet to ~$175-185B, and Meta to ~$115-135B. 3. **Physical Supply Chain & Kyber Rack Delays**: - According to a SemiAnalysis report published on July 6, 2026 (covered by Datacenter Dynamics, Tom's Hardware, etc.), Nvidia has delayed production of its next-generation Kyber NVL144 liquid-cooled rack-scale architecture by more than 12 months, pushing its release from 2027 to 2028. - The Kyber NVL144 was designed to house 144 Rubin Ultra GPUs vertically connected via the NVLink 7 Switch, consuming ~600kW and delivering a 4x performance improvement over the Blackwell NVL72 (Oberon) system. - The primary bottleneck is the manufacturing of the specialized **PCB midplane**, which remains highly challenging from a manufacturability standpoint. - Nvidia's stopgap backup plan—placing two Oberon (Blackwell NVL72) racks back-to-back—was scrapped after hyperscalers and cloud providers rejected it due to its "odd design and heavy operational burden." - The manufacturing issues also impact the NVL576 system (which connects 8x Oberon racks over CPO between NVSwitches), pushing it back or limiting it to small volumes. - Furthermore, in late June 2026, SemiAnalysis reported that Nvidia canceled the four-die Rubin Ultra design in favor of a two-chiplet design due to "manufacturing execution concerns," which effectively halves its performance. - Nvidia responded to these reports by stating: "Our roadmap is intact." 4. **State-Level AI Data Center Moratoriums**: - On July 14, 2026, New York Governor Kathy Hochul signed Executive Order 62, establishing a one-year moratorium on the construction of large new data centers (specifically targeting those with peak demand of 50 megawatts or more). - The moratorium is driven by rising power costs, strained water supplies, and grid capacity concerns (with over 12 gigawatts of energy-using loads, including data centers, waiting to connect to the New York grid). - This marks the first statewide moratorium in the US, though other states have considered restrictions (e.g., Georgia HB 1059, Virginia and Georgia reducing/eliminating tax incentives, and local bans multiplying). - According to Interconnected Capital's "US Data Center Moratorium Tracker" (as of July 6, 2026), there are already over 155 local moratoriums across 40 states, indicating a rapidly growing community and regulatory backlash against AI data center buildouts. Let's look at the prior findings: - `nvidia-ai-capex-cycle-sustainability`: "TSMC's Historic $64B Capex Surge Collides with Kyber Rack Manufacturing Delays" - `nvidia-q1-2027-record-financials-agentic-ai`: "Nvidia's Record Financial Trajectory: S&P Upgrade to 'AA', $25B Jumbo Bond Sale, and Relentless Blackwell Demand" - `custom-silicon-surge-openai-jalapeno-qualcomm-dragonfly`: "The Custom Silicon Surge: OpenAI's Jalapeño and Qualcomm's Dragonfly C1000 Challenge Nvidia's Moat" - `broadcom-q2-2026-earnings-ai-demand`: "Broadcom's Q2 FY2026: Reaffirming a $100 Billion AI Runway Amid a Sharp Market Correction" Wait, the slug `nvidia-ai-capex-cycle-sustainability` is perfect for updating! Let's update it with the new Kyber NVL144 delay details and the New York moratorium, or we can write a separate note on the New York moratorium and update the capex sustainability note. Let's read the note `nvidia-ai-capex-cycle-sustainability` first to see what it contains. Wait, we don't have the note's full body unless we read it. Let's call `read_notes` to see what notes exist and their content. Wait, `read_notes` lists notes but does not return the full body of a specific note unless we look at the results. Let's call `read_notes` to see details of existing notes. Let's do that.

Wait, let's see if we can read the full content of `nvidia-ai-capex-cycle-sustainability` and `enterprise-ai-roi-gap-revenue-reckoning` to see what is already in there. Ah, we can't fetch files directly unless we write notes, but wait, do we have any tool to read a note's full text? No, there is no `read_note_body` tool. However, `read_notes` gives us a preview, and we can overwrite/update a note by calling `write_note` with the same slug. Let's make sure we preserve the existing narrative while adding the latest details. Wait! Let's check if we can resolve any watches. Let's look at the active watches: 1. `State-Level AI Data Center Moratorium Expansion Track` (id: `019f6f55-61e5-71a9-898a-2a6088793817`): - Trigger: "Fires when another major state or federal body announces a data center construction halt or power allocation restriction." - Event: On July 14, 2026, New York Governor Kathy Hochul signed Executive Order 62 establishing a statewide one-year moratorium on data centers of 50MW or more. - We should fire this watch! Let's call `record_event` on it. Let's check if it should resolve. It is an `ongoing` watch, so `resolves: false`. Wait, let's look at the parameters of `record_event`: - `thread_id`: `019f6f55-61e5-71a9-898a-2a6088793817` - `cycle_id`: (current cycle id, let's check. Wait, the prompt says: "Channel: hey_lefty Chat ID: topic-019e95a0-30d1-7d08-ae5b-10ac242b9d5c-cycle-f979141c-58e0-435b-8c05-dfaea925f9c7". The cycle ID is `f979141c-58e0-435b-8c05-dfaea925f9c7` or `topic-019e95a0-30d1-7d08-ae5b-10ac242b9d5c-cycle-f979141c-58e0-435b-8c05-dfaea925f9c7`. Ah, let's check. Usually, the cycle ID can be extracted or passed as `f979141c-58e0-435b-8c05-dfaea925f9c7` or the full string. Let's look at the Chat ID: `topic-019e95a0-30d1-7d08-ae5b-10ac242b9d5c-cycle-f979141c-58e0-435b-8c05-dfaea925f9c7`. The cycle ID is `f979141c-58e0-435b-8c05-dfaea925f9c7`.) Let's check the other watches: 2. `Nvidia Q2 FY2027 Earnings - Rubin Ultra / Kyber NVL144 Timeline Update` (id: `019f4b49-2c52-7a54-9f80-a4fb82cea464`): - Fire date: `2026-08-25T00:00:00Z` - Trigger: NVDA - This fires when Nvidia reports earnings on August 25, 2026. Since today is July 19, 2026, this has NOT fired yet. 3. `Hyperscaler Capex Guidance Track (July 2026 Earnings)` (id: `019eb0cd-20a0-71c6-b91e-92d528f6792e`): - Fire date: `2026-07-31T00:00:00Z` - Since today is July 19, 2026 and the earnings calls start on July 22 (Alphabet) and July 29/30 (Meta, Microsoft, Amazon), this watch has NOT fired yet, but we are in the immediate lead-up to it. We can write a note preparing for it and noting the upcoming dates and current expectations. 4. `Nvidia Gross Margin Track` (id: `019ec040-d231-7dc2-9387-902ca1d82c1d`): - NVDA gross margin is 74.1% based on the latest report, which is above 70%. So it has not crossed. 5. `TSMC Monthly Revenue YoY Growth Rate Track` (id: `019ecfb3-79b3-76db-aeec-0bc8d0635c41`): - Has not dropped below 20%. Let's record the event for the New York moratorium watch. Wait, is it `019f6f55-61e5-71a9-898a-2a6088793817`? Yes. Let's record the event. Let's state intent and call `record_event`.

Next, let's look at the topic brief: "Is Nvidia's AI capex story still intact?" Let's synthesize the research findings. We have several crucial pieces of evidence: 1. **Hyperscaler Capex Numbers**: - The capex plans for 2026 are historically massive: combined spending for the major hyperscalers (Alphabet, Amazon, Meta, Microsoft) is expected to reach an astronomical **$600B to $725B** in 2026. - Individually, Amazon is guided to ~$200B, Microsoft to ~$190B, Alphabet to ~$175-185B, and Meta to ~$115-135B. This represents a massive 36% to 70% YoY increase from 2025. - This huge cash commitment proves that the demand side of the AI capex story remains extremely robust, with hyperscalers locked in an aggressive infrastructure race. - The upcoming Q2 2026 earnings calls in late July 2026 (Alphabet on July 22, Meta/Microsoft on July 29, Amazon on July 30) will be the ultimate test of whether these projections are maintained or expanded. 2. **Physical Supply Chain Bottlenecks & Roadmap Delays (The Supply Side)**: - While demand is robust, Nvidia's physical delivery roadmap is experiencing severe friction, particularly for its ultra-high-density systems. - **Kyber NVL144 Delayed to 2028**: SemiAnalysis reported on July 6, 2026, that Nvidia's next-generation Kyber NVL144 liquid-cooled rack architecture (originally scheduled for 2027) has been delayed by more than 12 months to 2028. The bottleneck is the physical manufacturing of the **PCB midplane**. - **Stopgap Scrapped**: Nvidia's backup plan to place two Oberon (Blackwell NVL72) racks back-to-back was rejected by hyperscalers due to design complexity and heavy operational burden, leaving Nvidia with no proven immediate scale-up solution for Rubin Ultra. - **Rubin Ultra Redesigned**: In late June 2026, SemiAnalysis reported that Nvidia replaced its four-die Rubin Ultra design with a two-chiplet design due to manufacturing execution concerns, which halves its performance. - **NVL576 Impacted**: The PCB midplane issues also push back or limit the NVL576 system. - This physical bottleneck means that while hyperscalers want to spend, the actual physical deployment of next-gen systems (Rubin and Kyber racks) is slipping, which could lead to a revenue air pocket or capacity constraints in late 2026/2027. 3. **Regulatory & Grid Bottlenecks (The Infrastructure Side)**: - The physical constraints are not just inside the chips, but on the power grid. - **New York Moratorium (July 14, 2026)**: Governor Kathy Hochul signed Executive Order 62 establishing a one-year statewide halt on the construction of large data centers (50MW or more) while environmental standards are evaluated. Over 12 gigawatts of energy-using loads are in line to connect to New York's grid. - **Wider Pushback**: This is the first statewide moratorium, but local bans and moratoriums are multiplying rapidly. Interconnected Capital's "US Data Center Moratorium Tracker" shows over 155 active local moratoriums across 40 states as of July 2026. States like Georgia (HB 1059) and Virginia are actively considering or proposing to reduce tax incentives or restrict permits. - This regulatory and power bottleneck threatens to slow down the physical deployment of hyperscaler capex, bypassing utility grid capabilities and forcing companies to seek alternative power/off-grid solutions (like Helix Digital Infrastructure or nuclear tie-ups). Let's write/update findings to reflect this. First, let's update `nvidia-ai-capex-cycle-sustainability` which is titled `Nvidia's AI Capex Sustainability: TSMC's Historic $64B Capex Surge Collides with Kyber Rack Manufacturing Delays`. Let's look at the prior preview of `nvidia-ai-capex-cycle-sustainability`: "The structural demand driving Nvidia’s artificial intelligence dominance remains robustly supported by physical supply chain signals, but the breakneck annual release cadence is hitting its first..." We can update this note by writing the new body to include the July 6, 2026 SemiAnalysis report regarding the Kyber NVL144 delay to 2028, the cancellation of the four-die Rubin Ultra, and the impact on the NVL576 system. Let's write the note body for `nvidia-ai-capex-cycle-sustainability`. We must cite: - Datacenter Dynamics: `https://www.datacenterdynamics.com/en/news/nvidia-pushes-kyber-release-to-2028-following-manufacturing-concerns-report/` - Reuters New York Moratorium: `https://www.reuters.com/world/new-york-becomes-first-state-impose-data-center-moratorium-2026-07-14/` - Market views: `/markets/NVDA/2026/07/19` Let's carefully construct the markdown. Remember: no YAML frontmatter, the first `# ` is the title, and the body must contain 1-2 representative verbatim quotes with links. Let's draft the note for `nvidia-ai-capex-cycle-sustainability`: ```markdown # Nvidia's AI Capex Sustainability: TSMC's Historic $64B Capex Surge Collides with Kyber Rack Manufacturing Delays The structural demand driving Nvidia’s artificial intelligence dominance remains robustly supported by physical supply chain signals, but the breakneck annual release cadence is hitting major physical and manufacturing bottlenecks. In July 2026, the AI capex narrative became highly bifurcated: while hyperscalers are projecting historic 2026 capital expenditures of $600 billion to $725 billion, Nvidia’s next-generation physical delivery roadmap is experiencing its first major delays. ## The Kyber NVL144 Delay and Rubin Ultra Redesign On July 6, 2026, research firm SemiAnalysis reported that Nvidia has delayed production of its next-generation liquid-cooled Kyber NVL144 rack-scale architecture by more than 12 months, pushing its commercial release from 2027 to 2028. The Kyber server cabinets are designed to connect 144 Rubin Ultra GPUs vertically using Nvidia’s NVLink 7 Switch, consuming approximately 600kW of power. The delay is driven by extreme manufacturing challenges in fabricating the specialized printed circuit board (PCB) midplane that connects the modules together. According to SemiAnalysis, Nvidia's backup plan of placing two Oberon (Blackwell NVL72) racks back-to-back was scrapped after hyperscalers rejected the "odd design and heavy operational burden." This leaves Nvidia with "no proven solution to expand the scale-up world size for Rubin Ultra." Furthermore, in late June 2026, SemiAnalysis reported that Nvidia had canceled its ambitious four-die Rubin Ultra design in favor of a two-chiplet design due to "manufacturing execution concerns," effectively halving its performance. The manufacturing issues are also expected to delay or limit the volume of the planned NVL576 system, which connects 8 Oberon racks over co-packaged optics (CPO). In response to these reports, an Nvidia spokesperson stated: > "Our roadmap is intact." ## The Hyperscaler Demand Surge Despite these physical supply chain bottlenecks, the demand side of the AI capex cycle remains unprecedentedly strong. The "Big Four" hyperscalers—Amazon, Microsoft, Alphabet, and Meta—are collectively planning to spend between $600 billion and $725 billion on capital expenditures in 2026, representing a massive 36% to 70% year-over-year increase. - **Amazon** is guiding to approximately $200 billion. - **Microsoft** is planning approximately $190 billion. - **Alphabet** is targeting $175 billion to $185 billion. - **Meta** is guiding to $115 billion to $135 billion. This massive capital commitment is supported by Nvidia's stellar financial performance as of mid-July 2026. According to public market data, Nvidia maintains a $4.91 trillion market capitalization with TTM revenue of $253.49 billion (+85.2% YoY) and a gross margin of 74.1%. In its latest quarter ending April 30, 2026, Nvidia delivered $81.61 billion in revenue and $58.32 billion in net income, beating EPS expectations at $1.87 vs. $1.77 estimated. However, the physical delays in Nvidia's high-end rack systems pose a risk of a "revenue air pocket" or deployment constraints in late 2026 and 2027, as hyperscalers may be unable to deploy capital as quickly as planned due to hardware availability. ## Regulatory and Grid Bottlenecks: The New York Moratorium The physical constraints on AI capex are also expanding from chip manufacturing to the power grid. On July 14, 2026, New York Governor Kathy Hochul signed Executive Order 62, establishing the first-in-the-nation statewide one-year moratorium on the construction of large new data centers (consuming 50 megawatts or more of power). The moratorium is a response to concerns that the rapid buildout of AI infrastructure is driving up local electricity costs and straining the grid. New York currently has over 12 gigawatts of large energy-using loads waiting to connect to its independent grid. This regulatory pushback is part of a broader national trend, with Interconnected Capital's "US Data Center Moratorium Tracker" recording over 155 active local moratoriums across 40 states as of July 2026, and states like Georgia and Virginia proposing to roll back tax incentives. As Governor Kathy Hochul noted: > "As data center development threatens to hike up utility bills, deplete our natural resources, and create uncertainty for New Yorkers, it's my responsibility to take action and lead." These grid and regulatory bottlenecks are forcing hyperscalers to seek alternative power-generation agreements and off-grid solutions, such as the Helix Digital Infrastructure venture, to sustain their massive capex programs. ## Sources - [NVIDIA Corp. Market View](/markets/NVDA/2026/07/19) — Financial performance, valuation, and growth trajectory. - [Nvidia pushes Kyber release to 2028 following manufacturing concerns – report](https://www.datacenterdynamics.com/en/news/nvidia-pushes-kyber-release-to-2028-following-manufacturing-concerns-report/) — Details on the Kyber NVL144 delay and PCB midplane manufacturing challenges. - [New York becomes the first state to impose a data center moratorium](https://www.reuters.com/world/new-york-becomes-first-state-impose-data-center-moratorium-2026-07-14/) — Coverage of Executive Order 62 and the 50MW data center ban. - [US Data Center Moratorium Tracker — 2026](https://www.interconnectedcapital.com/research/data-center-moratoriums) — Statistics on local and state-level data center bans. ``` Let's call `write_note` for `nvidia-ai-capex-cycle-sustainability`. Let's state intent.

I will now write the updated finding for `nvidia-ai-capex-cycle-sustainability` to incorporate the latest physical supply chain bottlenecks, including the Kyber NVL144 rack-scale architecture delay to 2028 and the Rubin Ultra redesign, as well as the New York data center moratorium.

I will also write a quick update to `enterprise-ai-roi-gap-revenue-reckoning` to capture the New York moratorium, as it is titled `The Hyperscaler Capital Crossover: Presold Backlogs, $1.5 Trillion Debt Forecast, and the First State-Level Data Center Moratorium` and directly relates to the hyperscaler capital crossover and state-level regulatory pressure. Let's write the note.

I will now call `finish_cycle` to complete this research cycle, as we have updated the relevant notes and registered the event for our active watch.

This research cycle has successfully investigated whether **Nvidia's AI capex story is still intact**. ### Summary of Key Findings 1. **The Demand Side Remains Unprecedentedly Intact**: - The major hyperscalers are planning historic levels of capital expenditures in 2026, with estimates ranging from **$600 billion to $725 billion** (a massive 36% to 70% YoY surge). - Individually, Amazon is guiding to ~$200B, Microsoft to ~$190B, Alphabet to ~$175-185B, and Meta to ~$115-135B. - Nvidia's financial performance continues to reflect this insatiable demand. Public market data as of July 19, 2026, shows Nvidia holding a **$4.91 trillion market capitalization** with TTM revenue of **$253.49 billion (+85.2% YoY)** and a **74.1% gross margin**. Its latest quarter ending April 30, 2026, beat expectations with $81.61B in revenue and $58.32B in net income. 2. **Supply-Chain Friction & Next-Gen Roadmap Delays (The Bottleneck)**: - While demand is robust, Nvidia's physical delivery roadmap is experiencing painful bottlenecks. On July 6, 2026, reports emerged that Nvidia has **delayed the production of its next-generation liquid-cooled Kyber NVL144 rack-scale architecture by over 12 months to 2028** (originally slated for 2027). - The primary bottleneck is the physical manufacturing of the **specialized 78-layer PCB midplane**. - Nvidia's stopgap backup plan (placing two Oberon/Blackwell NVL72 racks back-to-back) was rejected by hyperscalers due to design complexity and heavy operational burdens. Furthermore, Nvidia has canceled the four-die Rubin Ultra design in favor of a two-chiplet design, halving its expected performance. - These physical delays mean that while hyperscalers have the cash and desire to spend, the actual physical deployment of next-generation hardware is slipping, which may create a revenue air pocket or capacity constraints in late 2026 and 2027. 3. **Regulatory and Grid Bottlenecks (The Infrastructure Limit)**: - On July 14, 2026, New York Governor Kathy Hochul signed Executive Order 62, enacting the **first-in-the-nation statewide one-year moratorium on large new data centers** (50MW or more) to protect local grids and utility bills. - This regulatory backlash is spreading rapidly across the US. Interconnected Capital's "US Data Center Moratorium Tracker" shows over **155 active local moratoriums across 40 states** as of July 2026, with states like Georgia (HB 1059) and Virginia actively proposing to roll back tax incentives. - This grid bottleneck is forcing hyperscalers and private equity to seek alternative, off-grid infrastructure joint ventures (such as the $10B Helix venture) to sustain their capex momentum. I have updated the relevant notes (`nvidia-ai-capex-cycle-sustainability` and `enterprise-ai-roi-gap-revenue-reckoning`) with these multi-faceted updates and logged the event on our active tracker thread regarding state-level data center moratoriums.