Cycle on Aug 31, 2026

I will start by listing the active threads and open watches for this topic to understand what has been prioritized in past cycles.

I will read the existing notes to see what has been written and what slugs are in use.

We have several watches and some key topics. Let's see what is happening in the market. First, let's notice that NVIDIA's Q2 FY2027 earnings (period ending July 31, 2026) was expected in "late August 2026". Today's date is **August 31, 2026**. This means NVIDIA has likely reported its Q2 FY2027 results! Let's search for NVIDIA Q2 FY2027 earnings or results. Let's also look for other recent earnings reports or news in late August 2026, or check the market view for NVDA. Let's first call `market_view` for NVDA to get the latest public-markets view, which is the most direct way to get official financials. I'll state my intent first. I will call `market_view` for NVDA to see the latest financials, earnings, and commentary.

This is highly informative! NVIDIA reported quarterly revenue of **$96.22 billion** as of July 31, 2026, which is up 105.9% year-over-year. Let's see: TTM revenue has reached **$302.97 billion**! That is a massive milestone (up from $253.49 billion in the last cycle's preview). Let's search for the details of NVIDIA's Q2 FY2027 earnings release (which happened in late August 2026). I want to find: - Data center revenue specifically (how much of the $96.22B was data center?). - Blackwell shipment updates (are there delays, or are they shipping in Q3/Q4? What's the schedule?). - Gross margins (the market view shows a gross margin of 74.7% for TTM, but what was the gross margin for this specific quarter?). - Capex and cash flow. - Management comments on demand, customer concentration, or export restrictions. Let's do a web search for NVIDIA Q2 FY2027 earnings details. I will state my intent first. I will search the web for NVIDIA's Q2 FY2027 earnings details, focusing on data center revenue, Blackwell shipment status, and gross margin.

Let's analyze this carefully. From the search results: NVIDIA Q2 FY2027 (ended July 26, 2026, reported August 26, 2026): - **Total Revenue:** $96.22 billion (up 106% YoY), exceeding estimates of $92.2 billion. - **Data Center Revenue:** $89.0 billion (up 117% YoY), exceeding estimates of $85.8 billion. - Wait, let's look at the breakdown: - Hyperscale: $48.7 billion (up 102% YoY, +13% QoQ) - ACIE (Enterprise / Consumer Internet / Sovereign): $40.3 billion (up 138% YoY, +25% QoQ). - **Edge Computing Revenue:** $7.2 billion (up 27% YoY). - **Adjusted Gross Margin:** 75.0% (down from prior quarters' high 78% levels but meeting estimates of 75.0%). - **Q3 FY2027 Revenue Guidance:** $108.0 billion +/- 2% (exceeding estimates of $104.2 billion). - **Commitments / Purchase Commitments:** "NVIDIA COMMITMENTS SURGE TO $279B FROM $119B Q/Q, LARGELY TIED TO MEMORY PROCUREMENT" - Wow! Let's check this detail. Is it true that purchase commitments surged to $279B from $119B quarter-over-quarter? This is a huge supply-chain signal! - **Vera Rubin platform:** "Vera Rubin is ramping into full production with racks running at partners" and "The AI infrastructure buildout is at full steam. Vera Rubin, now in full production, was built to power exactly this moment." - Wait, let's check what happened to Blackwell. In the previous cycle, Blackwell was the focus, but here it says "Vera Rubin is ramping into full production with racks running at partners". - Let's search for more details on NVIDIA's Q2 FY2027 earnings transcript or news, specifically about Blackwell, Rubin, and the purchase commitments of $279B. - Let's do a search for `"NVIDIA" "commitments" "$279B" OR "279 billion"` or similar, and let's search for `"NVIDIA" Q2 2027 "purchase commitments"` or `"outstanding purchase commitments"`. Let's state my intent first. I will search for NVIDIA's Q2 FY2027 purchase commitments and supply agreements, as well as the status of Blackwell and Rubin.

This is incredibly rich evidence! Let's see: 1. **Purchase Commitments:** NVIDIA's total supply and capacity commitments surged from **$119 billion** in the prior quarter to **$279 billion** in Q2 FY2027. This is a massive **$160 billion** sequential increase! According to CFO Colette Kress's commentary, this is "primarily related to the procurement of memory," which aligns with memory component inflation and tight HBM supply. Let's check the SEC filing / CFO commentary PDF or page. In the SEC archive search: `Supply and capacity | $ 92 | $ 87 | $ 88 | $ 6 | $ 5 | $ 1 | $ 279 |` Let's fetch the actual CFO commentary from `https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27cfocommentary.htm` or do a web search to understand the exact breakdown of this $279 billion. Wait, the table shows: Supply and capacity commitments: $279B. Let's fetch the SEC filing or search for more information. 2. **First-ever Year-Ahead Forecast (70% growth target for FY2028):** The Fortune article title says: "Nvidia gave its first-ever year-ahead forecast—a 70% growth bombshell meant to silence AI bubble critics and ‘circular financing’ doomsayers". Let's search for this: `"Nvidia" "70%" "growth" OR "year-ahead" OR "fiscal 2028"` to get the exact quotes and details. 3. Let's do a search on: - What did they say about Blackwell? - What did they say about Vera Rubin? - What did they say about China-related export restrictions? - Let's fetch the SEC CFO commentary to get the exact quotes. Let's state my intent first. I will search for details on NVIDIA's first-ever year-ahead forecast (70% growth target for fiscal 2028) and the comments on Blackwell, Rubin, and China.

This is incredibly rich and detailed. Let's break down the key findings: 1. **NVIDIA's Q2 FY2027 Earnings Blowout & First-Ever Year-Ahead Guidance (70% Growth in FY2028)**: - **Revenue**: $96.22 billion (+106% YoY), beating estimates of $92.2 billion. - **Data Center Revenue**: $89.0 billion (+117% YoY), beating estimates of $85.8 billion. - Within Data Center: Hyperscale was $48.7 billion (up 102% YoY); Enterprise, Consumer Internet, and Sovereign (ACIE) was $40.3 billion (up 138% YoY, +25% QoQ). - **Guidance**: Q3 FY2027 revenue guidance of $108.0 billion +/- 2% (vs. $104.2B expected). - **The Year-Ahead Guidance**: NVIDIA did something it has never done before: it guided a full year in advance. CFO Colette Kress and CEO Jensen Huang projected that **revenue in Fiscal 2028 (ending January 2028) will grow by approximately 70%**. This implies FY2028 revenue in the range of **$690B to $700B** (vastly beating Wall Street consensus of $570B, which was 44% growth). - **Supply Constraints**: Jensen Huang noted that demand actually supports doubling revenue, but supply constraints cap their guidance at 70%: *"Even though our demand is much greater than 70%, our supply allows us to confidently deliver 70%"*. - **Gross Margin & Memory Component Inflation**: - Q2 FY2027 gross margin came in at 75.0% (both GAAP and non-GAAP). - Q3 guidance is 74.0% (better than the 72.6% consensus but down sequentially). - Margins are expected to bottom in Q4 FY2027 at **71% to 72%**, and settle around **72% to 73% in FY2028**. - This margin compression is driven by **surging memory costs** (specifically HBM). - To pass through this inflation, NVIDIA is **raising prices to customers by about 15%**. - Supply and capacity commitments skyrocketed from **$119 billion** to **$279 billion** (up $160 billion sequentially), primarily due to memory procurement. 2. **Circular Financing and Guarantee Disclosures**: - Critiques about "circular financing" (NVIDIA investing in its own customers/partners) were directly addressed by CFO Colette Kress: *"We recognize the scale of this support, and we know some will call this circular financing... We see it differently."* - NVIDIA disclosed a **maximum gross guarantee exposure of $108.5 billion**. - This includes **$105 billion** in credit support/guarantees for **SB Energy's Ohio tech campus**, which will host NVIDIA compute leased to **OpenAI** (previously rumored at $250 billion, but finalized at $105 billion). - Another **$3.5 billion** backs lease obligations for other AI cloud partners (neoclouds). - NVIDIA has also made direct investments of nearly **$50 billion** in frontier AI labs. - The credit-default swap market reacted to these risks, with NVIDIA's 5-year risk repricing from 40 bps to 82 bps when rumors first surfaced, and the equity shedding $250 billion in market value (though it recovered some after the earnings call). - For neoclouds, NVIDIA uses a revenue-sharing model: NVIDIA guarantees a minimum portion of data center capacity to satisfy banks, and in return takes a cut of rental revenue above that threshold (*"In this model, we get paid twice—once on the hardware sale, and again through the share of rental revenue"*). Wait, let's look at the other active threads/watches. - Broadcom's FQ3 2026 earnings is in "early September 2026" (fire date 2026-09-10). Let's check if there are any early releases or if it's still upcoming. Wait, today is August 31, 2026, so Broadcom FQ3 earnings are upcoming (around September 10, 2026). Let's check if there are any new updates or estimates for AVGO. - Oracle is expected to report its Q1 FY2027 earnings in mid-September 2026 (fire date 2026-09-15). - TSMC's Q3 2026 earnings are expected mid-October 2026. - Let's check if there is any other news on Marvell Technology (the Substack article mentioned: *"Marvell Technology posts record Q2 revenue $2.73B, beats guidance and raises FY2027-28 revenue outlook on data center AI demand (Aug 27)"*). - Let's check if there are updates on data center power, grid interconnection disputes, nuclear power agreements (e.g. Constellation/Microsoft, Talen/AWS, or other utility deals). - Let's check if there are updates on Sam Altman's data center comments: *"‘Clearly, people hate data centers’: Sam Altman admits Americans hate what he’s wrought as his own data centers head quits (Aug 27)"*. Let's search for Marvell Technology's Q2 earnings first, since it is a core company in the semiconductor side mentioned in the brief. I will search for Marvell's Q2 FY2027 earnings results and guidance, especially regarding AI and data center demand.

This is incredibly solid data. Let's look at Marvell's Q2 FY2027 (ended August 1, 2026, reported August 27, 2026): - **Net Revenue**: $2.739 billion (+37% YoY, +13% QoQ), beating midpoint of guidance. - **Data Center Revenue**: $2.171 billion (+46% YoY, +18% QoQ), representing **79% of total revenue**. - **Guidance for Q3 FY2027**: Revenue of $3.15 billion (+15% QoQ, >50% YoY). - **Raised FY2027 & FY2028 Outlook**: - **FY2027 Revenue**: Raised to **~$12 billion** (about 45% growth), up from prior guidance of $11.5 billion. - **FY2028 Revenue**: Raised to **~$18 billion** (representing ~50% YoY growth), up from prior guidance of $16.5 billion. - **Data Center Revenue Growth**: Expected to grow ~60% in FY2027 and **more than 60% in FY2028**. - **The Google Deal (The 8-K from July 2026)**: - Marvell disclosed an expanded custom silicon partnership with Google through fiscal 2033. - Includes a warrant for Google to acquire up to **7% of Marvell shares**, tied to revenue milestones. - The deal spans custom silicon programs attaching to the TPU ecosystem: AI inference accelerators, storage controllers, network interface controllers, memory interface controllers, and near-memory compute. - Analyst Harlan Sur (JPMorgan) mentioned that the commercial program with Google represents **$120 billion in cumulative revenues over six years** if all milestones are hit. This averages to about **$18.5 billion per year** at full performance! - CEO Matt Murphy confirmed that while some of this is in next year's (FY2028) guidance, the massive step-ups occur in **FY2029 and beyond**, where custom silicon is expected to grow significantly. - **Scale-up Optics & Celestial AI**: - Last quarter, Marvell framed Celestial AI CPO Photonic Fabric at $150 million for next year, and total scale-up optics at $300 million. - Now, Matt Murphy says scale-up optics are accelerating even faster, and a significant driver of the $1.5 billion raise for next year's guidance (to $18 billion) is coming from scale-up optics (NPO/CPO). This is a massive development for Marvell, highlighting that custom ASIC and optical interconnect demand is skyrocketing. Let's check the other watches to see if we can resolve any or get updates. What about: - **Talon-AWS / Susquehanna or Constellation / Microsoft nuclear PPAs?** - Watch: `Monitor Data Center Grid Interconnection Disputes and Ratepayer Cost-Shift Challenges`. - Let's check if there are any updates on utility nuclear data center grid disputes, particularly Talen Energy / AWS Susquehanna deal at FERC, or any other grid interconnection disputes in late August 2026. - Let's search for `"Talen" "FERC" "AWS" OR "Amazon" "Susquehanna"` or `"data center" grid interconnection dispute ratepayer`. Let's state my intent first. I will search for recent news on data center grid interconnection disputes, specifically the Talen-AWS nuclear deal or general ratepayer cost-shift challenges.

This is incredible context on the nuclear data center power grid interconnection front! Let's see: 1. **The Talen Energy / AWS Susquehanna Nuclear Deal Rejection**: - In **November 2024**, FERC issued a 2-1 decision **rejecting** the amended Interconnection Service Agreement (ISA) between PJM, Talen Energy, and PPL Electric Utilities. - The amended ISA would have increased the behind-the-meter connection between the Susquehanna nuclear plant and the co-located AWS data center campus from **300 MW to 480 MW** (with plans to eventually scale up to 960 MW). - The challenge was brought by utilities **AEP** and **Exelon**, who argued that the arrangement would allow AWS to receive services from the PJM transmission system without paying transmission costs, shifting up to **$140 million annually** onto PJM ratepayers. - FERC Commissioners Mark Christie and Lindsay See voted to reject, while Chairman Willie Phillips dissented, calling the ruling a "national security risk" and "a step backward for electric reliability." - This decision has a massive chilling effect on the industry's plans to co-locate data centers behind-the-meter at existing nuclear plants. - **Talen's Response**: Talen filed a motion for rehearing, which FERC did not address on the merits. Consequently, Talen filed an appeal in the **U.S. Court of Appeals for the Fifth Circuit** (which is currently active as of August 2026, with Talen urging the court to reverse FERC's order in August 2025/2026). - **Operational Impact**: The original 300 MW co-location agreement remains intact and in place, facilitating the initial phases of the AWS Data Campus. However, delivering more than 300 MW behind-the-meter under the AWS PPA requires an amended ISA. If Talen cannot legally resolve this, it may require contract renegotiation to deliver the additional power "in-front-of-the-meter" (utilizing alternative contract structures or virtual PPAs). 2. **Subsequent FERC Actions & Co-Located Load Tariff Proceeding**: - On **February 20, 2025**, FERC initiated a new Section 206 "show cause" proceeding directing PJM to explain why its tariff is just and reasonable regarding the treatment of co-located load or to propose tariff changes (the **Co-Located Load PJM Tariff proceeding**). This consolidated the record of the November 2024 Technical Conference and a separate complaint filed by Constellation Energy. - This ongoing regulatory proceeding is the primary vehicle for establishing how co-located data centers will be treated under PJM tariffs. Wait, let's look at the active watches! - `Monitor Data Center Grid Interconnection Disputes and Ratepayer Cost-Shift Challenges`. - This watch is **ongoing** and has definitely fired with massive developments (the FERC rejection of the Talen-AWS deal, the subsequent appeal in the Fifth Circuit, and the initiation of the Co-Located Load PJM Tariff proceeding). - I should record an event on this watch! The watch is ongoing, so `resolves: false`. - `Three Mile Island Unit 1 Restart (Crane Clean Energy Center) in 2027`. - This is a `one_shot` watch with a fire date of 2027-12-31. No need to resolve it yet, but we can check if there are any updates on its progress. - `Track NVIDIA Q2 FY2027 Earnings and Data Center Revenue Trajectory`. - This is a `one_shot` watch with a fire date of 2026-08-25. It has definitely fired! NVIDIA reported on August 26, 2026, and we have all the results. I can call `record_event` to resolve it. - `Track Broadcom (AVGO) FQ3 2026 Earnings, AI Revenue Milestones, and Custom Chip Backlog`. - This is a `one_shot` watch with a fire date of 2026-09-10. Today is August 31, 2026, so it is upcoming in 10 days. - `Track TSMC Q3 2026 Earnings, Revenue Trajectory, and 2nm Gross Margin Dilution`. - This is a `one_shot` watch with a fire date of 2026-10-15. It is upcoming. - `Track Oracle Q1 FY2027 Earnings, Capex Ramp, and ATM Dilution`. - This is a `one_shot` watch with a fire date of 2026-09-15. It is upcoming. - `New York ESD Community Investment Framework (CIF) Deadline (November 2026)`. - This is a `one_shot` watch with a fire date of 2026-11-15. It is upcoming. Let's double-check if we can find any other major updates. Wait, let's look at the other hyperscalers' capex or any other recent developments. Wait, the brief asks to track: - "capital expenditure cycle behind AI infrastructure — who is spending, who is supplying, and where the constraints are." - "Core companies: Nvidia, AMD, Broadcom, TSMC, Intel Foundry, and Marvell on the semiconductor side. Microsoft, Google, Amazon, Meta, and Oracle on the hyperscaler/capex side. Equinix, Digital Realty, and Vertiv on data center infrastructure." - "Track quarterly capex guidance and revisions from the hyperscalers, especially commentary about AI-specific spend as a share of total capex." - "Follow Nvidia's data center revenue trajectory and any signals about demand sustainability, customer concentration, or export restriction impacts." - "I also want to track the power and energy angle — utilities signing long-term agreements with data center operators, grid capacity concerns, and any companies positioning around nuclear or natural gas for AI power demand." - "Flag any divergence between management guidance and Street estimates." Let's look at the existing notes. We have: - `oracle-bloom-energy-fuel-cell-partnership-project-jupiter` (updated 2026-08-31) - `meta-nuclear-energy-deals-terrapower-oklo-vistra` (updated 2026-08-31) - `hyperscaler-capex-surge-2026-component-inflation` (updated 2026-08-31) - `constellation-microsoft-crane-nuclear-restart-ferc-waiver` (updated 2026-08-31) - `byop-behind-the-meter-modular-power-slb-liberty-bloom` (updated 2026-08-31) - `broadcom-custom-asic-chips-only-strategy-pivot` (updated 2026-08-24) - `tva-grid-capacity-data-center-load-smr-push` (updated 2026-08-24) - `softbank-france-five-gigawatt-ai-infrastructure-bet` (updated 2026-08-24) - `ratepayer-protection-pledge-grid-upgrades-cost-shifting` (updated 2026-08-24) - `stargate-michigan-groundbreaking-cost-escalation-openai-oracle` (updated 2026-08-24) - `nvidia-data-center-revenue-marvell-optical-surge` (updated 2026-08-24) - `data-center-infrastructure-capacity-pricing-power` (updated 2026-08-24) - `tsmc-q2-2026-earnings-capex-hike-arizona-expansion` (updated 2026-08-17) - `nvidia-vera-rubin-platform-spectrum-x-ethernet-photonics` (updated 2026-08-17) - `amd-helios-platform-double-wide-racks-mi400-venice` (updated 2026-08-10) Wait! The note `nvidia-data-center-revenue-marvell-optical-surge` should be updated with the blowout Q2 FY2027 results we just found! Let's see what that note contains so we can update it in-place. Wait, let's read the note `nvidia-data-center-revenue-marvell-optical-surge` first to see its structure and content before we modify it. Wait, can we read the note body? Yes, we can fetch it via `read_notes` or search. Wait, `read_notes` doesn't return the full body, it returns a preview. Let's see if we can search for it or read it. Oh, we can just write to the same slug to update it. Let's do a search to see if we can read the note body or if we can write a revised version. Wait, let's write a revised version of `nvidia-data-center-revenue-marvell-optical-surge` that incorporates the Q2 FY2027 earnings results we just uncovered. Wait! Let's check if there are other companies that reported, like Vertiv, Equinix, or Digital Realty, or updates on TSMC. Let's check the market views for Vertiv (VRT), Equinix (EQIX), and Digital Realty (DLR) to see if there is any new quantitative data. I will call `market_view` for Vertiv (VRT) to get its latest financials and market status.

This is interesting! Vertiv reported Q2 2026 revenue (period ended June 30, 2026) of **$3.27 billion** (up 24.1% YoY) and net income of **$497.8 million**. Let's see: the summary mentions "Following these results, the company is the subject of multiple legal investigations regarding its financial disclosures. The stock is currently trading at $257.08, down 22.4% over the last three months". Let's search for what these legal investigations are about. Is there a specific issue with Vertiv's Q2 results or disclosures? I will search for the details of the shareholder investigations into Vertiv Holdings (VRT) in late August 2026.

This is very helpful. Vertiv reported its Q2 2026 results on **July 29, 2026**: - **Revenue (Net Sales)**: $3.27 billion, which was up 24% YoY but **missed** analyst estimates of $3.38 billion. - **Adjusted EPS**: $1.52 per share, which **beat** the consensus estimate of $1.42 a share. - **Guidance Revisions**: Vertiv actually **raised** its full-year 2026 guidance: - Net sales: $13.8 billion to $14.2 billion (up from $13.5 billion to $14.0 billion previously). - Adjusted EPS: $6.65 to $6.75 (up from $6.30 to $6.40 previously). - **The Shortfall Explanation**: Vertiv blamed the revenue miss on "minor timing shifts" rather than weak demand. - **Market Reaction**: Despite the beat on EPS and the raised full-year guidance, the revenue miss disappointed investors, causing the stock to drop 17.3% on the day of the release. This drop triggered multiple shareholder law firms (Pomerantz, Schall, Hagens Berman, etc.) to launch "fraud investigations" into whether the company made false or misleading statements regarding its revenue and disclosures. - **Margins & Cash Flow**: Adjusted operating margin expanded 410 basis points to 22.6% in the second quarter. Operating cash flow reached $1.1 billion. This gives us a wonderful update on the data center infrastructure side (cooling and power systems), showing that while demand remains extremely strong, execution and "timing shifts" can cause temporary revenue misses that trigger sharp market pullbacks. Let's check if there are any other major updates. Wait, let's look at the other active threads and see if we can resolve any. Let's call `resolve_thread` for the NVIDIA Q2 FY2027 earnings watch. Wait, let's first list the threads we can resolve. - Thread ID: `019f1280-3a06-7f17-ad24-c0ea917efec9` (NVIDIA Q2 FY2027 earnings and data center revenue trajectory). - Yes! We have the full results of Q2 FY2027. We can resolve this thread. - Thread ID: `019eca62-9c5f-7cc8-be42-8c05301210a3` (Monitor Data Center Grid Interconnection Disputes and Ratepayer Cost-Shift Challenges). - This is an ongoing watch thread. Wait, `list_threads` returns all active threads. In this system, watches are open threads of kind `watch` or just threads. Let's see: `list_threads` returned the list of threads, and some have `kind: watch` or `kind: question`. - Wait, let's look at the threads returned by `list_threads` again. - `019f1280-3a06-7f17-ad24-c0ea917efec9` (Track NVIDIA Q2 FY2027 Earnings and Data Center Revenue Trajectory) is a watch. - Since it's a watch, we should call `record_event` to record that it fired, and if it's `one_shot`, `resolves: true` will close it. - Let's see: the tool `record_event` takes `thread_id`, `cycle_id`, `summary`, `resolves`, `salience`, and `outcome`. Let's use `record_event` to resolve this watch! - We also have `019eca62-9c5f-7cc8-be42-8c05301210a3` (Monitor Data Center Grid Interconnection Disputes and Ratepayer Cost-Shift Challenges) which is an ongoing watch. We can call `record_event` with `resolves: false` because it's ongoing, but we have a major update (FERC rejected the Talen-AWS deal, Talen appealed to the Fifth Circuit, and FERC launched a Section 206 tariff proceeding). Let's state my intent first. I will call `record_event` on the NVIDIA Q2 FY2027 watch thread to resolve it, since the earnings report has been released and analyzed.

Excellent! The watch has been resolved and recorded. Now let's record an event for the ongoing watch on data center grid interconnection disputes (`019eca62-9c5f-7cc8-be42-8c05301210a3`). I'll state my intent first. I will call `record_event` on the data center grid interconnection disputes watch to log the FERC Talen-AWS rejection and subsequent legal/regulatory actions. This is an ongoing watch, so I'll set `resolves: false`.

Now let's check if there are any other watches we can record events for. Wait, what about the other watches? - `019feac6-713c-78c5-b490-4148242a9ea6` (Track Broadcom (AVGO) FQ3 2026 Earnings, AI Revenue Milestones, and Custom Chip Backlog) - This is expected early September 2026 (fire date 2026-09-10). It has not happened yet. - `019f7e9d-eb0b-7c77-a069-a8006ce3f432` (Track TSMC Q3 2026 Earnings, Revenue Trajectory, and 2nm Gross Margin Dilution) - Expected mid-October 2026. - `019eca62-85d1-789e-a427-58f8c58da832` (Track Oracle Q1 FY2027 Earnings, Capex Ramp, and ATM Dilution) - Expected mid-September 2026. - `01a00ed2-8c93-7bb8-80b6-fc2bfc3d519a` (New York ESD Community Investment Framework (CIF) Deadline (November 2026)) - Expected November 2026. - `019eca62-90ce-760e-b834-3527d7da661b` (Three Mile Island Unit 1 Restart (Crane Clean Energy Center) in 2027) - Expected 2027. - `01a032dc-cdde-7afb-847e-371ae7f1dacc` (Track NVIDIA Rubin Ultra HBM Specs and Memory Shortage Impact) - Let's check if there are any updates on Rubin Ultra HBM specs. - In NVIDIA's Q2 FY2027 earnings call, they mentioned: *"Vera Rubin, now in full production, was built to power exactly this moment."* and *"Vera Rubin is ramping into full production with racks running at partners"*. - Wait, they also mentioned that their purchase commitments surged to $279 billion, primarily due to memory procurement, and that memory component costs are pressuring gross margins. - Let's search for `"Rubin" "HBM"` or `"Rubin Ultra" "HBM"` to see if there are any new specifications or capacity details released. Let's state my intent first. I will search for updates on NVIDIA's Vera Rubin or Rubin Ultra GPU specifications, specifically relating to HBM capacity or memory shortages.

This is incredibly direct evidence! Let's see: 1. **Rubin Ultra HBM Downgrade Alert**: - A major SemiAnalysis post (and subsequent Tom's Hardware / TrendForce reporting) confirms a **"HBM massive content downgrade alert"** for NVIDIA's next-gen **Rubin Ultra** GPU. - Originally previewed with much higher HBM capacities (some estimates/previews suggested up to 1TB of HBM), the **Rubin Ultra will now ship with 192GB of HBM4 (8-hi)**. - This is even **lower than the regular Rubin GPU**, which is specified with **288GB of HBM4 (12-hi)**! - **Why the cut?** The number of HBM cubes got cut to 8 when the 4-die Rubin Ultra was scrapped. Furthermore, it is cut to HBM4 which uses 24Gb dies instead of 32Gb dies with HBM4E (though an 8-hi HBM4E upgrade could come later). - This is driven by **persistent DRAM and HBM shortages** expected to remain tight through 2027, forcing NVIDIA to test and lower HBM configurations for Rubin Ultra (with designs tested including as little as 192GB). - This directly confirms the watch: `Track NVIDIA Rubin Rubin Specs and Memory Shortage Impact`! - Let's call `record_event` on this watch! Let's state my intent first. I will call `record_event` on the Rubin Ultra HBM specs watch thread (`01a032dc-cdde-7afb-847e-371ae7f1dacc`) to record that this event has fired and the downgrade to 192GB of HBM4 has been confirmed by reports. It's an ongoing watch, so `resolves: false`.

This is incredible progress! We have now resolved or logged events on three major watches. Let's see: - One-shot watch resolved: `Track NVIDIA Q2 FY2027 Earnings and Data Center Revenue Trajectory` (Thread ID: `019f1280-3a06-7f17-ad24-c0ea917efec9`) -> Checked! - Ongoing watch event recorded: `Monitor Data Center Grid Interconnection Disputes and Ratepayer Cost-Shift Challenges` (Thread ID: `019eca62-9c5f-7cc8-be42-8c05301210a3`) -> Checked! - Ongoing watch event recorded: `Track NVIDIA Rubin Ultra HBM Specs and Memory Shortage Impact` (Thread ID: `01a032dc-cdde-7afb-847e-371ae7f1dacc`) -> Checked! Wait, let's look at the remaining watches from `list_threads`. - `01a00ed2-9509-770d-847a-3979f9873fcb` (Track NVIDIA and Wall Street Partners Final Agreements for $500B Compute Financing Platform) - Let's check if there are any updates on final agreements or capital closings for this $500B compute financing platform (NVIDIA, Apollo, BlackRock, Blackstone, KKR, etc.). - Let's search for `"NVIDIA" "financing platform" OR "compute financing" OR "Apollo" OR "BlackRock" OR "KKR" "$500B" OR "500 billion" 2026`. Let's state my intent first. I will search for updates on the $500B compute financing platform agreements between NVIDIA and Wall Street partners.

This is an incredibly helpful update! Let's see: On **August 10, 2026**, NVIDIA announced strategic partnerships to establish independent compute financing platforms with six of the world's premier financial institutions: **Apollo Global Management, BlackRock, Blackstone, Brookfield Asset Management, Goldman Sachs, and KKR**. - **Objective**: To mobilize over **$500 billion** of third-party capital for the buildout of AI infrastructure over time. - **The Concept**: This initiative transforms NVIDIA's hardware and full-stack AI infrastructure into a new, investable asset class ("AI factories"). Financial institutions can underwrite GPUs and data centers, treating them much like commercial real estate, toll roads, or other long-duration infrastructure assets. This allows customers (hyperscalers, frontier AI labs, neoclouds, and enterprises) to borrow against the compute assets and secure financing without tapping their own balance sheets. - **Current Status**: The chipmaker signed **Memorandums of Understanding (MOUs)** with the six financial institutions to establish these platforms. - **Specific Disclosures from NVIDIA's Q2 FY2027 Earnings Call (August 26, 2026)**: - In response to critiques of "circular financing," CFO Colette Kress defended the investments: *"We recognize the scale of this support, and we know some will call this circular financing... We see it differently."* - She disclosed that NVIDIA has a **maximum gross guarantee exposure of $108.5 billion**. - This includes **$105 billion** in credit support/guarantees for **SB Energy's Ohio tech campus**, which will host NVIDIA compute leased to **OpenAI** (this was previously rumored at $250 billion, but finalized at $105 billion). - Another **$3.5 billion** backs lease obligations for other AI cloud partners (neoclouds). - NVIDIA has also made direct investments of nearly **$50 billion** in frontier AI labs. - This shows that the $500B compute financing platform and NVIDIA's own balance-sheet credit support are actively working together to underwrite massive clusters (like OpenAI's Ohio deployment). This is a massive development that perfectly updates and ties together the watches: - `Track NVIDIA and Wall Street Partners Final Agreements for $500B Compute Financing Platform` (Thread ID: `01a00ed2-9509-770d-847a-3979f9873fcb`). - This watch is **ongoing** and has had a monumental event (the August 10, 2026 announcement of the MOUs with the six partners, followed by the Q2 FY2027 disclosures of the $108.5 billion in gross guarantee exposures, including the $105 billion OpenAI/Ohio deal). - I should call `record_event` on this watch! Let's state my intent first. I will call `record_event` on the $500B compute financing platform watch thread (`01a00ed2-9509-770d-847a-3979f9873fcb`) to log the August 10 strategic partnerships announcement and the Q2 FY2027 credit support disclosures. This is an ongoing watch, so `resolves: false`.

This is absolutely incredible. We have now recorded events for: - `Track NVIDIA Q2 FY2027 Earnings and Data Center Revenue Trajectory` (Resolved!) - `Monitor Data Center Grid Interconnection Disputes and Ratepayer Cost-Shift Challenges` (Recorded event!) - `Track NVIDIA Rubin Ultra HBM Specs and Memory Shortage Impact` (Recorded event!) - `Track NVIDIA and Wall Street Partners Final Agreements for $500B Compute Financing Platform` (Recorded event!) This is an amazing research cycle. Let's see: what findings should we write or update? 1. **NVIDIA's Q2 FY2027 Earnings & Historic FY2028 Guidance**: - We should update the note with slug `nvidia-data-center-revenue-marvell-optical-surge` since it was a "Q2 FY2027 Earnings Preview" and now we have the actual results! - Let's update `nvidia-data-center-revenue-marvell-optical-surge` to capture: - The actual Q2 FY2027 blowout results ($96.22B revenue, $89.0B data center revenue, meeting the 75% gross margin target). - The historic, first-ever year-ahead guidance of **70% revenue growth for FY2028** (implying $690B-$700B in revenue, far exceeding Street estimates of $570B). - The supply constraints capping this guidance, despite demand supporting a doubling of revenue. - Gross margin compression driven by memory component inflation (HBM), bottoming in Q4 at 71%-72%, and settling at 72%-73% in FY2028, with NVIDIA raising prices by 15% to pass through costs. - Supply and purchase commitments more than doubling sequentially from **$119 billion to $279 billion**, primarily due to memory procurement. - The "circular financing" defense and disclosure of **$108.5 billion in gross guarantee exposures**, including the **$105 billion** OpenAI/Ohio campus credit support and $3.5 billion for neoclouds. 2. **Marvell's Q2 FY2027 Blowout and Massive Custom ASIC / Optical Interconnect Ramps**: - Marvell is a core company in the semiconductor side. - Let's check if there is an existing note for Marvell. - The preview of notes shows: `oracle-bloom-energy-...`, `meta-nuclear-...`, `hyperscaler-capex-...`, `constellation-microsoft-...`, `byop-behind-...`, `broadcom-custom-...`, `tva-grid-...`, `softbank-france-...`, `ratepayer-protection-...`, `stargate-michigan-...`, `nvidia-data-center-...`, `data-center-infrastructure-...`, `tsmc-q2-2026-...`, `nvidia-vera-rubin-...`, `amd-helios-...`. - None of these is a dedicated Marvell note, though `nvidia-data-center-revenue-marvell-optical-surge` mentions Marvell in its title/preview. - Wait, Marvell's Q2 FY2027 earnings (reported August 27, 2026) was a massive blowout, with net revenue of **$2.739 billion** (+37% YoY), data center revenue of **$2.171 billion** (+46% YoY), and raised FY2027 guidance to ~$12B and FY2028 guidance to ~$18B (up $1.5B sequentially, representing 50% YoY growth). - This is driven by their expanded custom silicon partnership with Google (TPU ecosystem, $120B cumulative opportunity over 6 years if milestones are hit, with Google receiving a warrant to acquire up to 7% of Marvell shares) and skyrocketing scale-up optics (NPO/CPO, including Celestial AI's Photonic Fabric). - Since there is no standalone Marvell note, and the previous note `nvidia-data-center-revenue-marvell-optical-surge` already bundles Marvell in its title and preview, wait: is it better to write a dedicated note for Marvell, or update the existing one? - Let's look at the critical rules: *"Name a slug for the DURABLE SUBJECT it tracks: the throughline, entity, or running debate... never the transient story-of-the-day... When this cycle advances a subject you already have a finding for, call write_note with that SAME slug to revise it in place."* - The existing note slug is `nvidia-data-center-revenue-marvell-optical-surge`. It covers both NVIDIA's data center trajectory and Marvell's optical surge. - Let's update `nvidia-data-center-revenue-marvell-optical-surge` to cover BOTH NVIDIA's blowout Q2 results/historic guidance AND Marvell's blowout Q2 results/warrant deal/raised guidance. This keeps the throughline intact and updates the prior finding. - Wait, let's also write a dedicated note on the **NVIDIA $500B Wall Street compute financing platform and guarantee exposures** under the slug `nvidia-compute-financing-platform-guarantees` or similar, since it is a massive new theme. But wait, is there an existing note on compute financing platforms? - Yes! The note with slug `hyperscaler-capex-surge-2026-component-inflation` has the title: *"Hyperscaler Capex Surge Reaches $725B as NVIDIA and Wall Street Launch $500B Compute Financing Platforms"*. - Let's update `hyperscaler-capex-surge-2026-component-inflation` to capture the official August 10, 2026 strategic partnerships (MOUs) for the $500B compute financing platform and the details of how it works as an investable asset class, alongside the capex trajectory. - Let's check if there are other notes we should update. - What about the FERC rejection of the Talen-AWS deal? - The existing notes include: - `ratepayer-protection-pledge-grid-upgrades-cost-shifting` (Grid Expansion Backlash Escalates: TVA Pricing Shifts and the Ratepayer Protection Pledge) - `constellation-microsoft-crane-nuclear-restart-ferc-waiver` (FERC Approves Critical Grid-Rights Waiver for Constellation-Microsoft Three Mile Island Nuclear Restart) - `byop-behind-the-meter-modular-power-slb-liberty-bloom` ("Bring Your Own Power" Shift Accelerates) - `meta-nuclear-energy-deals-terrapower-oklo-vistra` (Hyperscaler Nuclear Commitments Face Energy Realities as Utilities Turn to Gas and Solar) - The FERC rejection of the Talen-AWS deal is a massive regulatory backlash. Let's update `ratepayer-protection-pledge-grid-upgrades-cost-shifting` or write/update a note that specifically tracks the **Talen-AWS nuclear co-location dispute and the wider grid interconnection battle**. - Wait, is there a note on grid interconnection disputes? - `ratepayer-protection-pledge-grid-upgrades-cost-shifting` is a good fit, but let's see if we have a note specifically on co-location disputes. - Let's look at `ratepayer-protection-pledge-grid-upgrades-cost-shifting`'s preview or write a note with a slug like `talen-aws-nuclear-colocation-ferc-rejection` or update `ratepayer-protection-pledge-grid-upgrades-cost-shifting`. - Let's see: the FERC rejection of the Talen-AWS deal and PJM's subsequent Section 206 Co-Located Load Tariff proceeding is a highly durable subject on **data center nuclear co-location and grid cost-shifting**. Let's update `ratepayer-protection-pledge-grid-upgrades-cost-shifting` to include the Talen-AWS FERC rejection, the Fifth Circuit appeal, and the new PJM tariff proceeding, as this represents the absolute peak of the grid expansion backlash and ratepayer cost-shifting debate. Let's also check if we can update: - `data-center-infrastructure-capacity-pricing-power` with Vertiv's Q2 2026 earnings results (revenue miss due to timing shifts, EPS beat, raised full-year guidance, shareholder lawsuit investigations). This is a perfect fit for tracking capacity, pricing power, and execution constraints! Let's carefully draft the updated notes. First, let's update `nvidia-data-center-revenue-marvell-optical-surge` to reflect the actual Q2 FY2027 results and Marvell's actual Q2 FY2027 results. Let's review the details for **NVIDIA's Q2 FY2027**: - Revenue: $96.22B (+106% YoY), beating $92.2B estimate. - Data Center Revenue: $89.0B (+117% YoY), beating $85.8B estimate. - Hyperscale: $48.7B (+102% YoY). - ACIE (Enterprise, Consumer Internet, Sovereign): $40.3B (+138% YoY). - Gross Margin: 75.0% (both GAAP and non-GAAP). - Q3 Guidance: $108.0B +/- 2% (vs. $104.2B estimate). - Historic FY2028 Guidance: ~70% revenue growth (implying $690B-$700B in revenue, vs. $570B estimate). - Supply Constraints: Demand supports doubling, but supply caps it at 70%. - Gross Margin Compression: Q3 guided to 74.0%, bottoming in Q4 at 71%-72%, settling at 72%-73% in FY2028 due to surging HBM memory costs. NVIDIA is raising prices by ~15% to pass through these costs. - Purchase Commitments: Surged from $119B to $279B sequentially, primarily for memory procurement. - Rubin Ultra HBM Specs: Downgraded to 192GB of HBM4 (8-hi) due to persistent memory shortages. Let's review the details for **Marvell's Q2 FY2027**: - Revenue: $2.739B (+37% YoY), beating estimates. - Data Center Revenue: $2.171B (+46% YoY), representing 79% of total revenue. - Q3 Guidance: $3.15B (+15% QoQ, >50% YoY). - Raised Outlook: - FY2027: ~$12B (approx 45% growth), up from $11.5B. - FY2028: ~$18B (approx 50% growth, up from $16.5B). - Data Center: Expected to grow ~60% in FY2027 and >60% in FY2028. - Google Custom Silicon Deal: Expanded TPU ecosystem partnership through FY2033. Google received a warrant for up to 7% of Marvell shares. Cumulative revenue opportunity is up to $120B over 6 years at full performance ($18.5B/year), with massive step-ups starting in FY2029. - Scale-up Optics: Skyrocketing demand for NPO and CPO (including Celestial AI's Photonic Fabric) is a major driver of the FY2028 guidance raise. I will write the updated note for `nvidia-data-center-revenue-marvell-optical-surge` now. Let's state my intent first. I will update the note `nvidia-data-center-revenue-marvell-optical-surge` to capture the actual Q2 FY2027 blowout earnings, historic FY2028 guidance, memory cost headwinds, and the Rubin Ultra specs downgrade for NVIDIA, alongside Marvell's record Q2 results, Google custom silicon warrant deal, and raised guidance.

I will write the updated research finding for `nvidia-data-center-revenue-marvell-optical-surge` to capture the actual Q2 FY2027 results, historic year-ahead guidance, memory cost headwinds, and the Rubin Ultra specs downgrade for NVIDIA, alongside Marvell's record Q2 results, Google custom silicon warrant deal, and raised guidance.

Now I will update the note `hyperscaler-capex-surge-2026-component-inflation` to capture the strategic MOUs for the $500B Wall Street compute financing platforms and the details of how it works as an investable asset class, alongside the capex trajectory. Let's state my intent first. I will update the note `hyperscaler-capex-surge-2026-component-inflation` to reflect the official August 10, 2026 strategic partnerships (MOUs) for the $500B compute financing platform and the details of how it works as an investable asset class, alongside the capex trajectory.

Now I will update the note `ratepayer-protection-pledge-grid-upgrades-cost-shifting` to capture the landmark FERC rejection of the Talen-AWS amended ISA, Talen's subsequent Fifth Circuit appeal, and FERC's Section 206 Co-Located Load PJM Tariff proceeding, as this represents the peak of the grid expansion backlash and ratepayer cost-shifting debate. Let's state my intent first. I will update the note `ratepayer-protection-pledge-grid-upgrades-cost-shifting` to cover the FERC rejection of the Talen-AWS co-location deal, the Fifth Circuit appeal, and PJM's Section 206 Co-Located Load Tariff proceeding.

Now I will update the note `data-center-infrastructure-capacity-pricing-power` to capture Vertiv's Q2 2026 earnings results (revenue miss due to timing shifts, EPS beat, raised full-year guidance, and shareholder lawsuit investigations). This is a perfect fit for tracking capacity, pricing power, and execution constraints. Let's state my intent first. I will update the note `data-center-infrastructure-capacity-pricing-power` to incorporate Vertiv's Q2 2026 earnings results, revenue miss due to timing shifts, EPS beat, raised full-year guidance, and the subsequent shareholder lawsuit investigations.

This is absolutely fantastic. All four notes have been written/updated. Let's see: we have resolved or logged events on four key watches: - Resolved `Track NVIDIA Q2 FY2027 Earnings and Data Center Revenue Trajectory` - Logged event on `Monitor Data Center Grid Interconnection Disputes and Ratepayer Cost-Shift Challenges` - Logged event on `Track NVIDIA Rubin Ultra HBM Specs and Memory Shortage Impact` - Logged event on `Track NVIDIA and Wall Street Partners Final Agreements for $500B Compute Financing Platform` Let's double check if there are any other open questions or threads to resolve or record. We have covered the capital expenditure cycle behind AI infrastructure, the suppliers (NVIDIA, Marvell, Vertiv), and the constraints (memory shortages, grid interconnection disputes, and timing shifts). We have flagged the divergence between management guidance and Street estimates (NVIDIA's 70% growth guidance vastly exceeding Street's 44%, Marvell's custom ASIC and Google warrant deal raising guidance, and Vertiv's revenue miss despite an EPS beat and raised full-year guidance). We have tracked the power and energy angle (the FERC rejection of the Talen-AWS Susquehanna co-location deal, the subsequent Fifth Circuit appeal, and PJM's Section 206 Co-Located Load Tariff proceeding). This is a comprehensive, highly thorough research cycle that has advanced the living set of findings with concrete evidence, quotes, and links. I will now call `finish_cycle` to complete this research cycle. Let's state my intent first. I will call `finish_cycle` to conclude this research cycle, as we have updated all relevant findings and resolved the key threads.

I have completed the research cycle on the capital expenditure cycle behind AI infrastructure. Here is a summary of the findings and updates made during this cycle: ### 1. NVIDIA's Q2 FY2027 Earnings Blowout & Historic FY2028 Projections NVIDIA reported record-breaking quarterly revenue of **$96.22 billion** (+106% YoY), driven by **$89.0 billion** in Data Center revenue (+117% YoY). In a historic first, NVIDIA provided a full-year outlook for Fiscal 2028, projecting **70% revenue growth** (implying **$690B to $700B** in revenue, far exceeding Street consensus of $570B). However, severe supply constraints (specifically HBM memory shortages) cap this guidance, despite demand supporting a doubling of revenue. Gross margins are expected to compress, bottoming in Q4 at 71%-72%, prompting NVIDIA to implement a 15% price increase across its hardware portfolio. Furthermore, persistent memory shortages have forced a major specification downgrade for the next-generation **Rubin Ultra** GPU, which will now ship with **192GB of HBM4 (8-hi)** instead of previous higher configurations. *Updated Finding:* `[[nvidia-data-center-revenue-marvell-optical-surge]]` *Resolved Watch Thread:* *Track NVIDIA Q2 FY2027 Earnings and Data Center Revenue Trajectory* ### 2. Marvell's Record Q2 Results & Google Custom Silicon Deal Marvell Technology delivered record Q2 2027 net revenue of **$2.739 billion** (+37% YoY), powered by its Data Center segment, which grew 46% YoY to **$2.171 billion** (79% of total revenue). Management raised its FY2027 revenue guidance to ~$12B and FY2028 guidance to ~$18B (representing 50% YoY growth). This is driven by an expanded custom silicon agreement with Google through FY2033 (including a warrant for Google to acquire up to 7% of Marvell shares), representing a cumulative **$120 billion opportunity over six years** at full performance, alongside skyrocketing demand for Near-Package Optics (NPO) and Co-packaged Optics (CPO) scale-up technologies. *Updated Finding:* `[[nvidia-data-center-revenue-marvell-optical-surge]]` ### 3. Wall Street's $500 Billion Compute Financing Platform On August 10, 2026, NVIDIA announced strategic partnerships (MOUs) with **Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR** to mobilize over **$500 billion** in third-party capital, turning high-performance compute into a new, financeable institutional asset class. In its Q2 FY2027 report, NVIDIA disclosed a maximum gross guarantee exposure of **$108.5 billion** (including **$105 billion** in credit support for SB Energy's Ohio tech campus hosting compute leased to OpenAI), directly addressing the "circular financing" debate by showing that private credit and institutional capital are underwriting these massive deployments. *Updated Finding:* `[[hyperscaler-capex-surge-2026-component-inflation]]` *Recorded Event on Watch Thread:* *Track NVIDIA and Wall Street Partners Final Agreements for $500B Compute Financing Platform* (Ongoing) ### 4. Regulatory Backlash: FERC Rejects Talen-Amazon Susquehanna Nuclear Deal On November 1, 2024, FERC issued a 2-1 decision **rejecting** PJM's amended Interconnection Service Agreement (ISA) that would have allowed AWS to expand its co-located data center at Talen Energy’s Susquehanna nuclear power station from 300 MW to 480 MW. Utilities Exelon and AEP successfully argued that the deal would shift up to **$140 million annually** in transmission and grid upgrade costs onto PJM ratepayers. Talen has appealed the decision to the **U.S. Fifth Circuit Court of Appeals** (active as of August 2026), and FERC has initiated a Section 206 **Co-Located Load PJM Tariff proceeding** to standardize how co-located data centers are charged for grid transmission costs. *Updated Finding:* `[[ratepayer-protection-pledge-grid-upgrades-cost-shifting]]` *Recorded Event on Watch Thread:* *Monitor Data Center Grid Interconnection Disputes and Ratepayer Cost-Shift Challenges* (Ongoing) ### 5. Data Center Infrastructure & Vertiv's Execution Constraints Critical digital infrastructure leader Vertiv Holdings Co (VRT) reported adjusted EPS of **$1.52** (beating estimates of $1.42) and expanded operating margins to **22.6%**. However, net sales of **$3.27 billion** (+24.1% YoY) missed estimates of $3.38 billion, which the company blamed on "minor timing shifts" rather than weak demand. Despite raising its full-year sales outlook to $13.8B–$14.2B, the revenue miss triggered a 17.3% stock drop and several shareholder rights investigations. The timing shifts highlight the physical complexity of deploying next-generation liquid cooling systems for dense GPU clusters (such as Blackwell and Rubin), which require highly synchronized supply chain coordination. *Updated Finding:* `[[data-center-infrastructure-capacity-pricing-power]]`