Cycle on Jul 14, 2026

I will start by listing the topic's open threads to see if there are any specific questions, watches, or objectives we are tracking for this topic. I will then pull the public market views for Nvidia (NVDA) and related entities to check the latest quantitative data.

I will pull the public market view for Nvidia (`NVDA`) to get its current financials, valuation, and recent price movement, as well as checking the macro backdrop. This will ground our quantitative analysis.

I will perform a web search to see what the latest news and reports are regarding Nvidia's AI capex, the hyperscaler spending plans, and if there are any signs of a slowdown or structural shifts in July 2026. This will help answer whether the AI capex story remains intact.

I will search for the specific URL `https://alcapitaladvisory.com/research/intelligence/ai-infrastructure.html` or similar pages to fetch up-to-date analysis on the 2026 hyperscaler capex and Nvidia's share of it. I will also do a search on Google Cloud's CoreWeave payments and Oracle's financing, as we have watches on those. Let's start with a fetch of the AL Capital Advisory page if possible, or do a broader search for "AL Capital Advisory" "57" "every $1" to see what the direct source says.

I will fetch the page `https://alcapitaladvisory.com/research/intelligence/ai-infrastructure.html` to get the latest, detailed numbers on the hyperscaler capex and Nvidia's share of it.

I will perform a search on Google Cloud's CoreWeave payments and Oracle's financing to see if there are any updates or if we should register/resolve threads. I see that we have several open watches. Let's search for "Google cloud CoreWeave payments 920M" or similar terms.

I will search for "Google Cloud" "CoreWeave" "$920M" or "920 million" to see if there is a specific contract or payment watch event we should track, or if it has any news. I will also search for Oracle FY2027 financing and Alphabet Q2 2026 earnings. Let's do a search for Oracle's financing first.

I will now search for "Google Cloud" "CoreWeave" "920M" or "920 million" to see if we can find any references to that capacity payment or if it's a known contract. I'll search with more flexible terms.

I will fetch `https://datacenters.economictimes.indiatimes.com/news/ai-compute-infrastructure/google-leases-spacex-ai-compute-capacity-for-920-million-monthly/131544492` or `https://mlq.ai/news/spacex-signs-920m-per-month-deal-to-lease-110000-nvidia-gpus-to-google-ahead-of-ipo/` to confirm the details of Google's $920 million monthly lease, which was originally registered as a watch for CoreWeave but seems to actually be with SpaceX (for xAI compute capacity). Let's fetch the MLQ AI page.

I will write a note updating the Nvidia AI capex story using our stable slug `nvidia-record-q1-results-and-trillion-dollar-order-book`. I will also update the big tech capex story using the stable slug `big-tech-capex-reaches-710b-in-2026`. Let's check the existing threads and watches again. We have a watch: - `Google Cloud $920M/month CoreWeave capacity payments commence` (ID `019eecbc-83f2-7338-8da9-83c6a4f4bd83`). However, our research shows that the $920M/month payment is actually Google Cloud paying **SpaceX** for GPU compute capacity (110,000 GPUs) commencing October 2026, not CoreWeave! This is a fascinating correction to the prior understanding. I will record an event or resolve this thread/watch, but first I will write the findings. Let's write the note updating Nvidia's AI Capex Story (`nvidia-record-q1-results-and-trillion-dollar-order-book`).

I will write a second note under the slug `big-tech-capex-reaches-710b-in-2026` to update the Big Tech capex story with the updated $725B figure and the newly revealed Google-SpaceX $920M/month deal. I will also incorporate the structural shift to an energy-constrained environment.

I will write a third note about specialized GPU clouds under the stable slug `gpu-specialized-clouds-and-strategic-co-investment` to reflect how SpaceX's S-1 filings and entry into the GPU-leasing market (with Google and Anthropic deals totaling over $2B/month) represent a massive new structural challenge for neoclouds like CoreWeave.

I will now record an event on our watch `Google Cloud $920M/month CoreWeave capacity payments commence` (ID `019eecbc-83f2-7338-8da9-83c6a4f4bd83`). The watch expected Google Cloud to pay ~$920 million/month to CoreWeave. However, our research has revealed that this $920 million/month capacity payment is actually Google Cloud paying **SpaceX** (SPCX) for leasing 110,000 Nvidia GPUs starting in October 2026. I will record this event, explaining the outcome, and resolve the watch since we have identified the true counterparty and verified the contract details.

I will now resolve the watch thread using `resolve_thread`.

I have updated the findings and resolved the relevant watch thread. I will now call `finish_cycle` to complete this research cycle.

I have successfully completed the research cycle for topic **019e95ad-bd47-7cf5-bd57-eb159d26bbbe** ("Is Nvidia's AI capex story still intact?"). ### Summary of Findings: 1. **Nvidia's AI Capex Story Remains Exceptionally Robust:** - **Hyperscaler Spend Acceleration:** The "Big-5" hyperscalers (Amazon, Microsoft, Alphabet, Meta, and Oracle) are confirmed to spend approximately **$725 billion** on AI infrastructure in 2026 (a ~64% YoY increase from 2025's $443B). - **Nvidia Capital Capture:** Nvidia captures **57.4¢ of every $1** spent by hyperscalers (up from 39¢ a year ago). - **Financial Performance:** Nvidia's Q1 FY2027 revenue reached **$81.61 billion** (+85.2% YoY), with data center revenue at **$75.2 billion** (+92% YoY). The demand base is broadening, as the newly introduced "ACIE" (AI Clouds, Industrial & Enterprise) segment reached $37 billion, nearly matching the traditional Hyperscale segment ($38B). - **Product Demand:** Q2 FY2027 revenue is guided to **$91.0 billion** (±2%). Blackwell B200 and Vera Rubin GPU architectures carry massive order backlogs of **12 to 18 months**, constrained primarily by TSMC's CoWoS-L advanced packaging capacity. 2. **The Structural Shift to an Energy-Constrained Bottleneck:** - The primary gating factor for AI infrastructure has shifted from GPU availability to reliable electrical power at scale. - North American colocation vacancy is at an all-time low of **~1.4%** as of year-end 2025. - Goldman Sachs documents an **11 GW US capacity shortfall** today, with Morgan Stanley projecting it to reach **~49 GW by 2028**. This has driven massive long-term contracts for nuclear baseload power (e.g., Constellation Energy's 20-year PPAs). 3. **SpaceX Disrupts the GPU-Leasing Market (Google Capacity Watch Resolved):** - In its amended S-1 filing ahead of its IPO, SpaceX disclosed a blockbuster Cloud Service Agreement with Google Cloud. - **The $920M/Month Watch Resolved:** Google Cloud will pay SpaceX **$920 million per month** from October 2026 through June 2029 to lease approximately 110,000 Nvidia GPUs (serving as "bridge capacity" for Gemini Enterprise). This resolves the prior watch which mistakenly expected these payments to go to CoreWeave. - Combined with a separate **$1.25 billion-per-month Anthropic contract** for its Colossus 1 data center, SpaceX's AI infrastructure contracts exceed **$2 billion in monthly run-rate revenue** (~$25 billion annualized), creating massive valuation and margin pressure on pure-play neoclouds like CoreWeave (which fell 7.1% on the disclosure). All findings have been integrated into the topic's notes under their stable slugs (`nvidia-record-q1-results-and-trillion-dollar-order-book`, `big-tech-capex-reaches-710b-in-2026`, and `gpu-specialized-clouds-and-strategic-co-investment`).