Cycle on Jun 6, 2026

I will start by checking the existing threads and active watches for this topic to see what questions or catalysts are currently open.

I will pull the public-markets data for Nvidia (NVDA) to examine its current financial trajectory, valuation, and recent performance.

I see that NVDA's stock recently fell 6.2% to $205.10, and there's a note about "Nasdaq 100's Worst Day Since Trump Tariff Shock" on June 5, 2026. The broader market experienced a downturn. Despite this, Nvidia reported extraordinary numbers for the quarter ending April 30, 2026: - Revenue of $81.61B (versus $68.13B in the prior quarter, and $46.74B three quarters ago) - Net income of $58.32B (on $81.61B revenue, representing a massive profit margin) - TTM Revenue of $253.49B (+85.2% YoY) - Market Cap is $5.20T Let's do some web searches to understand the current macro/industry environment around AI capex and Nvidia, specifically checking if the "AI capex story is still intact" in mid-2026. We need to look for recent news on hyperscaler capital expenditures (Microsoft, Google, Meta, Amazon) and Nvidia's Blackwell GPU shipments or any potential headwinds/delays, or changes in demand. Let's run a web search. I will search for "Nvidia AI capex demand 2026" or "hyperscaler capex AI 2026". Or look at the recent Broadcom earnings plunge mentioned in the prior findings: "On June 3, 2026, Broadcom (AVGO) reported its fiscal Q2 2026 earnings... Broadcom shares plunged by 12.6%...". Let's check Broadcom's market view as well to see if there are details, and search for the context of that drop. Let's first search for general news on AI capex.

I should fetch some of these articles, specifically "AI Capex 2026: The $690B Infrastructure Sprint - The Futurum Group" and "Big Tech's $405B Bet: Why AI Stocks Are Set Up for a Strong 2026" or Goldman Sachs' "Tracking Trillions" to understand what the current sentiment is regarding hyperscaler capex and whether there are any cracks in the thesis. Let's also do a web search about the recent Broadcom earnings plunge on June 3, 2026, and any recent reports of cooling AI capex or Nvidia's performance in late May / early June 2026. Let's do a search for "Broadcom earnings June 2026 plunge" or "AVGO Q2 2026". Or search for "AI capex cooling 2026" or "Nvidia Blackwell delays 2026". Let's run a search.

This is very interesting. Broadcom (AVGO) reported its earnings on June 3, 2026, and its stock plunged. Let's fetch the CNBC article `https://www.cnbc.com/2026/06/03/broadcom-avgo-earnings-report-q2-2026.html` or Reuters article `https://www.reuters.com/business/broadcom-tumbles-revenue-miss-clouds-ai-boom-bets-2026-06-04/` to understand why Broadcom fell. Was it due to a change in the overall AI capex narrative, or was it company-specific (e.g., weak software sales, unchanged AI chip forecast, or missing extremely high expectations)? Let's also check if there is any broader macro impact. Let's fetch the Reuters article.

This is highly relevant. Broadcom's drop is a classic case of extremely high expectations meeting a market that wanted perfection. The company reiterated its $100 billion AI revenue forecast for fiscal 2027, rather than raising it. However, the $16 billion current-quarter AI chip sales forecast would mark an over three-fold jump from $5.2 billion a year ago. And Broadcom's CEO Hock Tan said the company expects to ship more than 10 gigawatts' worth of AI chips in 2027, and they are comfortable with memory supply, having secured supply for 2026 and 2027. Now, let's look at the actual capex numbers of the big tech players (Microsoft, Alphabet, Meta, Amazon) to see if their AI capex story is still intact. Let's search for "hyperscaler AI capex 2026" or "Microsoft Meta Alphabet Amazon capex Q1 2026". Let's first search for "Big Tech capex Q1 2026" or "hyperscaler capex Q1 2026". Let's run a search.

The Yahoo Finance article `'Magnificent 7' earnings rush reveals AI spending surge, with hyperscaler capex set to reach $725 billion in 2026` looks very relevant, as does the Tom's Hardware article `Google, Microsoft, Meta, and Amazon capex spending to hit $725 ...`. Let's search for "Magnificent 7 earnings rush reveals AI spending surge" or "hyperscaler capex set to reach $725 billion in 2026" to find the exact details. Let's fetch the Yahoo Finance article if we can, or search for its full text. Let's do a search for the Yahoo Finance URL or headline.

This is great. Let's do a search for the Yahoo Finance article itself by searching for `inurl:725-billion-in-2026-224901707` or similar to find the full page, or search for "Magnificent 7 earnings rush reveals AI spending surge" to fetch it. Wait, let's look at the PDF `https://www.elementpointe.com/wp-content/uploads/2026/05/2026.05.06-May-Monthly-Piece_CA.pdf` which was updated in May 2026. Let's fetch that PDF or search for its content, or search for the Yahoo Finance article. Let's do a search for the exact Yahoo Finance URL: `https://finance.yahoo.com/markets/article/magnificent-7-earnings-rush-reveals-ai-spending-surge-with-hyperscaler-capex-set-to-reach-725-billion-in-2026-224901707.html`. Let's fetch this URL! Since it's a Yahoo Finance article, let's fetch it first.

Wow, this is an excellent, detailed, and highly reliable source. Let's analyze the numbers reported in April 2026 for the calendar year 2026 capex of the Big Four hyperscalers (Microsoft, Alphabet, Meta, Amazon): - **Combined Capex**: Going into the quarter (Q1 2026, reported in late April 2026), the high end of estimates was around $670 billion. After the reports, it is closer to **$725 billion** for 2026. - **Meta Platforms (META)**: Raised its 2026 capex forecast to **$125 billion to $145 billion** (up $10 billion at both ends). Meta cited "expectations for higher component pricing this year and, to a lesser extent, additional data center costs to support future year capacity." - **Alphabet (GOOGL)**: Expects full-year 2026 capex of **$180 billion to $190 billion** (up $5 billion at both ends). Alphabet also stated it expects capex spending to increase **"significantly"** in 2027. - **Microsoft (MSFT)**: Expects calendar year 2026 capex to reach **$190 billion** (including $25 billion due to higher component pricing). Microsoft's annualized rate of AI capex in January was closer to $150 billion. - **Amazon (AMZN)**: Announced in January that it expected its capital expenditures in 2026 to approach **$200 billion**, and told investors on its earnings call that its plan remained "largely the same." This means the combined capex of just these four hyperscalers is approaching **$700B - $725B** (specifically, if we add them up: Meta ~$135B, Alphabet ~$185B, Microsoft ~$190B, Amazon ~$200B, that is $710B!). Let's check if there are any updates in May or June 2026, or any other reports about Nvidia's Blackwell GPU demand, shipments, or any delays. Let's search for "Nvidia Blackwell shipment June 2026" or "Blackwell demand June 2026". Let's run a search.

Let's look at the TrendForce article on iconnect007 or TrendForce directly about NVIDIA high-end shipments in 2026. Let's do a search for "TrendForce NVIDIA high-end shipments 2026 Blackwell Rubin". Or let's search for "TrendForce Blackwell to drive 70% of NVIDIA high-end GPU shipments in 2026". Let's run a search.

Let's fetch the TrendForce press center article: `https://www.trendforce.com/presscenter/news/20260408-13003.html` to get the key facts and numbers about NVIDIA's GPU shipment mix and any delays/issues. This is dated April 8, 2026, which is very fresh. Let's fetch it.

This is a very clear and key finding: - **TrendForce Report (April 8, 2026)**: High-end GPU shipments will see a shift in mix in 2026. The **Blackwell series is projected to grow markedly to account for over 70%** (specifically 71%, up from an earlier estimate of 61%) of NVIDIA's high-end GPU shipments in 2026. This is led by the GB300/B300 series. - **Rubin Delays**: The next-generation **Rubin series faces delay risks** (due to HBM4 validation, transitioning network interconnects from CX8 to CX9, managing higher power consumption, and optimizing liquid cooling). Its share of NVIDIA's 2026 shipments is expected to decline from 29% to 22%. - **Hopper Series**: Geopolitical factors (U.S.-China export policies) add uncertainty to H200 shipments, and Hopper's share is forecast to drop from 10% to 7%. - **Growth Rate**: High-end GPU shipment growth is slightly revised downward from 26.8% to 26% (still very high growth) due to the Rubin delays. - **Inference/LPUs**: NVIDIA is expanding into AI inference, with LPU solutions (likely related to their Groq acquisition / development we saw in prior findings: `[[nvidia-groq-20b-acquihire-inference]]`) expected to reach "several hundred thousand units in 2026, with a target to double in 2027." Let's check if there are other articles or reports on Nvidia's recent earnings or performance. From the market view, NVDA's TTM revenue is $253.49B, and in the latest quarter ending April 30, 2026, they did $81.61B in revenue, with a net income of $58.32B (operating income of $53.54B and EBITDA of $70.80B). This means their profit margin is an incredible 63% (operating margin 65.6%). Let's search for "Nvidia Q1 FY2027 earnings May 2026" or "Nvidia earnings May 2026" to confirm the exact details of their latest earnings call and what they said about Blackwell shipments and demand. Let's run a search.

This is extremely helpful! Let's fetch the CNBC live updates article for NVDA Q1 2027 earnings (reported on May 20, 2026): `https://www.cnbc.com/2026/05/20/nvidia-nvda-earnings-report-q1-2027.html` and let's check Intellectia's blog or other sources to get the exact quotes from the Q1 FY2027 earnings call in late May 2026. Let's search for "CNBC Nvidia Q1 2027 earnings report live updates" or fetch the CNBC URL: `https://www.cnbc.com/2026/05/20/nvidia-nvda-earnings-report-q1-2027.html`. Let's fetch it first.

This is an incredibly rich article! Let's pull out the key points: 1. **Financial Results (Q1 FY2027 ending April 30, 2026)**: - **Revenue**: $81.62 billion (vs $78.86 billion estimated) - **EPS**: $1.87 adjusted (vs $1.76 estimated) - **Data Center Revenue**: Accounted for 92% of total sales, nearly doubling from the previous year to $75.2 billion (out of $81.62B total). - **Buyback**: Nvidia announced an $80 billion share buyback program and increased its dividend. - **Guidance**: Guidance beat estimates (though the stock fell slightly on the call, which is a common pattern for NVDA post-earnings due to profit-taking/lofty expectations). 2. **Jensen Huang's Statements on Demand**: - "Demand has gone parabolic. The reason is simple: Agentic AI has arrived." - "Nvidia is the only platform that runs every frontier AI model," mentioning Anthropic, OpenAI, SpaceXAI, Meta, and Google's Gemini. - The chipmaker is powering every hyperscaler, supporting core data processing, ML workloads, internal AI services, and public cloud demand. - Huang highlighted the Vera CPU as a "major new growth driver," opening up a $200 billion revenue opportunity. 3. **Competitive Landscape and Customer ASICs**: - Nvidia acknowledged in its 10-Q filing that some of its customers are developing their own ASICs and products optimized for certain workloads that may not require all the features of Nvidia's data center systems. - This directly relates to Broadcom's business (AVGO helping Google and Meta design custom ASICs) and explains why AVGO is seeing strong custom AI chip growth (though AVGO's stock fell because they didn't raise their $100B FY2027 AI revenue guidance). 4. **Groq Acquisition/LPUs**: - Huang said on the call that Nvidia's custom AI chip from its $20B acquisition of Groq's tech "will be a niche product for some time." - This is an update on the Groq acquihire finding `[[nvidia-groq-20b-acquihire-inference]]`. Let's do a search on "Nvidia Blackwell demand 2026" or "Nvidia Blackwell backlog 2026" to get more concrete numbers about Blackwell's backlog and shipment schedule. Let's see: "Products are expected to start shipping this quarter, having been delayed due to an unexpected but now resolved design flaw. In August, Nvidia... Nvidia's Blackwell GPUs sold out for 12 months - DCD". Wait, let's search for "Nvidia Blackwell sold out 12 months" or check the Datacenter Dynamics article `https://www.datacenterdynamics.com/en/news/nvidias-blackwell-gpus-sold-out-for-12-months/` to get the details of Blackwell's supply/demand mismatch. Let's run a search.

This is incredibly valuable! Let's look at the article "The Blackwell Cliff That Wasn't" on Investing.com: `https://www.investing.com/analysis/the-blackwell-cliff-that-wasnt-200681435`. This article seems to address the core of the user's question: "Is Nvidia's AI capex story still intact?" and specifically talks about Blackwell's backlog and whether there's a cliff. Let's do a search to fetch it or get its text. Let's search for "The Blackwell Cliff That Wasn't".

This is incredible. Let's fetch the RIA (Real Investment Advice) article: `https://realinvestmentadvice.com/resources/blog/the-blackwell-cliff-that-wasnt/` to get the full story of "The Blackwell Cliff That Wasn't" which is dated June 3, 2026! This is exactly what we need to answer the user's prompt about whether Nvidia's AI capex story is still intact. Let's fetch it.

This is an absolute goldmine of information! Let's carefully synthesize what we've discovered: 1. **The "Blackwell Cliff" Concern**: Wall Street had feared a "Blackwell Cliff" — an "air pocket" or buying pause between the current Blackwell chip generation and the next-generation Vera Rubin chip platform. The worry was that hyperscalers would front-load Blackwell orders and then pause, leaving excess capacity. 2. **The Reality (The "Blackwell Cliff That Wasn't")**: - Nvidia's Blackwell backlog is **3.6 million units**, which is **fully sold out through late 2026** (mid-to-late 2026). - Supply, not demand, remains the primary constraint on Nvidia's growth. - The next-generation **Vera Rubin platform** is already in its sampling phase, with first production shipments expected in **Q3 2026** and volume ramping through **Q4 2026 into 2027**. This means the transition window is measured in months, not years. - At GTC (likely GTC 2026), Jensen Huang confirmed **over $1 trillion in combined Blackwell and Vera Rubin purchase orders through 2027** — which is **double** his $500 billion projection from the 2025 conference! 3. **Hyperscaler Capex Updates**: - In late April 2026 earnings reports, the big four hyperscalers (Microsoft, Alphabet, Meta, Amazon) raised their combined 2026 capex forecast from $670 billion to **closer to $725 billion**. - Meta raised its 2026 capex forecast to **$125B–$145B**. - Alphabet expects **$180B–$190B** for 2026 (and says 2027 will increase "significantly"). - Microsoft expects **$190B** for calendar year 2026. - Amazon expects around **$200B**. - Alphabet also recently did an **$80 billion capital raise** to fund its rapidly expanding AI infrastructure, with **Berkshire Hathaway committing $10 billion** to the deal (at a 6-8% discount). This indicates massive capital raising to fund the AI infrastructure sprint. 4. **Broadcom's Custom Chip Context**: - Broadcom (AVGO) recently dropped over 14% on June 4, 2026, because it reiterated, rather than raised, its **$100 billion AI revenue forecast for fiscal 2027**. While this disappointed a market that wanted perfection, its fundamental business remains highly robust: current-quarter AI chip sales forecast is $16 billion (up over 3x from $5.2 billion YoY) and it expects to ship more than 10 gigawatts of AI chips in 2027. - This reinforces that custom ASICs (which Broadcom designs for Google, Meta, etc.) are growing fast as alternatives to Nvidia's expensive chips, which Nvidia itself acknowledged in its latest 10-Q filing. 5. **Nvidia's Latest Earnings (May 20, 2026, for Q1 FY2027)**: - Revenue: **$81.62 billion** (vs $78.86 billion estimated), representing huge year-over-year and sequential growth (TTM revenue is now $253.49B). - Data Center Revenue: **$75.2 billion**, accounting for 92% of total sales. - Net Income: **$58.32 billion** (giving a profit margin of ~63%!). - Jensen Huang: "Demand has gone parabolic. The reason is simple: Agentic AI has arrived." Let's check if there are any other specific notes or threads we should write. First, we should update/refine the existing notes: - `hyperscaler-capex-surge-2026` (Hyperscaler Capex Surge: The $700 Billion AI Infrastructure Sprint) - `nvidia-q1-2027-record-financials-agentic-ai` (Nvidia's Record Q1 FY2027: Parabolic Demand Driven by Agentic AI) - `broadcom-q2-2026-earnings-ai-demand` (Broadcom's Q2 FY2026: Reaffirming a $100 Billion AI Runway for FY2027) - `nvidia-groq-20b-acquihire-inference` (Nvidia's $20 Billion Groq Acquihire: Securing the Agentic Inference Market) - wait, is there an update here? Yes, Jensen Huang's comment that the custom Groq LPUs "will be a niche product for some time". Let's check the exact slugs and content of the prior findings to make sure we refine them correctly. I can use `read_notes` to see the existing notes' content. Let's call `read_notes`. This is a very good step to ensure we carry the exact context forward.

Excellent. I see the existing notes and their titles. Let's update/refine them based on the latest findings we gathered: 1. **`hyperscaler-capex-surge-2026`**: This is the perfect place to integrate the updated capex numbers from late April 2026 (the $725 billion combined capex for 2026, including Meta raising to $125B-$145B, Alphabet expecting $180B-$190B, Microsoft at $190B, Amazon at $200B). We can also add the detail about Alphabet's massive $80 billion capital raise, with Berkshire Hathaway anchoring a $10 billion private placement at a 6-8% discount, signaling both the immense capital requirements and the funding risks/margin of safety demanded by major allocators. 2. **`nvidia-q1-2027-record-financials-agentic-ai`**: We can update this with the specific numbers from their May 20, 2026 earnings report (Q1 FY2027) which we got from CNBC: $81.62B revenue (vs $78.86B est), $1.87 adjusted EPS (vs $1.76 est), Data Center revenue nearly doubling to $75.2B (92% of sales), and an $80 billion share buyback program. We can also add the "Blackwell Cliff That Wasn't" detail: Blackwell's backlog of 3.6 million units is fully sold out through late 2026, and Jensen Huang's confirmation of over $1 trillion in combined Blackwell and Vera Rubin purchase orders through 2027 (doubling his $500B projection from 2025). 3. **`broadcom-q2-2026-earnings-ai-demand`**: We can update this with the details of its stock plunge on June 4, 2026 (slumping over 14% to erase $315B in market value). We can clarify that this was due to Broadcom reiterating, rather than raising, its $100 billion AI revenue forecast for fiscal 2027, which disappointed a market priced for perfection. We can highlight that its core business remains extremely robust (current-quarter AI chip sales forecast of $16B, up over 3x YoY from $5.2B, and Hock Tan expecting to ship over 10GW of AI chips in 2027). We can also connect this to Nvidia's 10-Q filing where Nvidia acknowledged that its own customers are developing custom ASICs (with Broadcom's help) optimized for specific workloads that compete with Nvidia's systems. 4. **`nvidia-groq-20b-acquihire-inference`**: We can add a brief update about Jensen Huang's comments during the Q1 FY2027 earnings call in late May 2026, where he stated that Nvidia's custom AI chips from the Groq acquisition (the Groq LPUs) "will be a niche product for some time." Let's carefully write these notes. First, let's write/update `hyperscaler-capex-surge-2026`. Wait, the guidelines state: "No YAML frontmatter — the title is the first `# ` heading and the slug is the slug parameter you pass. If the finding refines one of the prior findings auto-injected above, reuse that slug to update; otherwise pick a fresh slug." "Every substantive finding must: - name and link the actual thing (the story, article, paper, or page that was posted/published) — not just "the discussion", - carry 1-2 REPRESENTATIVE VERBATIM QUOTES, copied exactly, each kept with the link it came from (an HN comment permalink, the article URL, etc.), - then say what it means." Let's draft the body for `hyperscaler-capex-surge-2026`. Let's see: what are the sources for `hyperscaler-capex-surge-2026`? - Yahoo Finance article by Myles Udland: `'Magnificent 7' earnings rush reveals AI spending surge, with hyperscaler capex set to reach $725 billion in 2026` (https://finance.yahoo.com/markets/article/magnificent-7-earnings-rush-reveals-ai-spending-surge-with-hyperscaler-capex-set-to-reach-725-billion-in-2026-224901707.html) - RIA article: `The Blackwell Cliff That Wasn't` (https://realinvestmentadvice.com/resources/blog/the-blackwell-cliff-that-wasnt/) Let's get some verbatim quotes from these articles. From Yahoo Finance: > "Going into the quarter, the high end of estimates put the group's AI spending at around $670 billion this year. As of Wednesday night’s reports, that number is closer to $725 billion." > "Meta was the first to announce its plans for 2026, raising its capex forecast to $125 billion to $145 billion, up $10 billion at both ends." > "Alphabet told investors it now expects full-year capex to fall in the range of $180 billion to $190 billion, up $5 billion at both ends. Looking out to 2027, the company told investors it expects capex spending to increase 'significantly.'" > "Microsoft told investors on Wednesday night’s earnings call that it expects capex spending for calendar year 2026 to reach $190 billion..." From RIA: > "Yesterday, Alphabet announced a hefty $80 billion capital raise to fund its rapidly expanding AI infrastructure. Buried in that announcement was a detail worth examining: Berkshire Hathaway committed $10 billion to the deal at a discount to Alphabet’s current market price." > "When a well-regarded and disciplined capital allocator demands a margin of safety on a $10 billion investment, and the counterparty agrees, both sides are telling you they see some funding risk in the market." Let's write the note for `hyperscaler-capex-surge-2026`. I will call `write_note`. Let's do it.

Excellent! Now let's write/update the next note: `nvidia-q1-2027-record-financials-agentic-ai`. Let's see what we need to include: - Q1 FY2027 earnings (reported on May 20, 2026): - Revenue: $81.62 billion (vs $78.86B estimated) - EPS: $1.87 adjusted (vs $1.76 estimated) - Data center revenue: $75.2 billion (92% of total sales), nearly doubling YoY. - $80 billion share buyback program and increased dividend. - Jensen Huang's statements: - "Demand has gone parabolic. The reason is simple: Agentic AI has arrived." - "Nvidia is the only platform that runs every frontier AI model." - The "Blackwell Cliff That Wasn't": - Blackwell backlog is 3.6 million units, fully sold out through late 2026. - Supply, not demand, remains the primary constraint on growth. - Jensen Huang confirmed over **$1 trillion** in combined Blackwell and Vera Rubin purchase orders through 2027 (double the $500 billion projection from 2025). - Vera Rubin is already sampling, with first production shipments expected in Q3 2026 and ramping in Q4 2026/2027, compressing the transition window to months rather than years. - TrendForce Report (April 8, 2026): - Blackwell is projected to account for **71%** of NVIDIA's high-end GPU shipments in 2026 (led by GB300/B300). - Rubin faces delay risks (due to HBM4 validation, CX8 to CX9 interconnect transition, high power usage, and liquid cooling optimization), which drops its 2026 share to 22%, meaning Blackwell will dominate 2026 high-end shipments. Let's pull together the quotes: From CNBC: > "This was an extraordinary quarter. Demand has gone parabolic," Huang said as he closed out the earnings call. "The reason is simple: Agentic AI has arrived." > "Nvidia is the only platform that runs every frontier AI model," Huang said, mentioning Anthropic, OpenAI, SpaceXAI, Meta and Google’s Gemini. From RIA: > "At the recent GPU Technology Conference (GTC), Nvidia CEO Jensen Huang confirmed there are over $1 trillion in combined Blackwell and Vera Rubin purchase orders through 2027. That was double his $500 billion projection at the 2025 conference." > "Blackwell’s backlog of 3.6 million units sold out through late-2026, with supply, not demand, remaining the binding constraint on Nvidia’s growth." From TrendForce: > "Meanwhile, the Blackwell series is projected to grow markedly from 61% to 71%, solidifying its leading position in the market." > "This downward revision is primarily due to shipment delays facing the Rubin series. In addition to the time required for HBM4 validation, challenges include transitioning network interconnects from CX8 to CX9, managing significantly higher power consumption, and optimizing performance under more advanced liquid cooling solutions." Let's write this note under the slug `nvidia-q1-2027-record-financials-agentic-ai`. This is a direct update of the prior finding. Let's call `write_note`.

Excellent! Both notes updated successfully. Now let's write/update the third note: `broadcom-q2-2026-earnings-ai-demand`. Let's see what we need to include: - Broadcom reported its fiscal Q2 2026 earnings on June 3, 2026. - The stock fell by over 14% on Thursday, June 4, 2026, wiping out over $315 billion in market value. - The primary cause was Broadcom reiterating, rather than raising, its **$100 billion AI revenue forecast for fiscal 2027**, combined with custom AI chip sales guidance of $16 billion for the current quarter being slightly below whisper numbers (though still a massive increase from $5.2 billion a year ago). - Despite the drop, analysts raised price targets and Hock Tan reported robust long-term outlooks: - Expects to ship over **10 gigawatts** of AI chips in 2027. - Secured memory chip supply for 2026 and 2027, making them comfortable amid supply crunches. - Broadcom's business of custom ASICs (for Google, Meta, etc.) represents a major alternative to Nvidia's expensive chips. - This custom ASIC threat is real: Nvidia acknowledged in its latest 10-Q filing that its own customers are developing custom ASICs optimized for specific workloads that do not require all the features of Nvidia's data center systems, which could impact Nvidia's long-term demand. Let's pull together the quotes: From Reuters: > "Broadcom shares slumped more than 14% on Thursday, dragging chip peers lower, after the company's results fell short of lofty expectations around demand for its custom AI chips business." > "But Broadcom faces tougher competition and its eye-watering gains have left little room for error, analysts said, adding the decision to reiterate, rather than raise, its $100 billion AI revenue forecast for fiscal 2027 disappointed investors." > "To be sure, Broadcom CEO Hock Tan said the company now expects to ship more than 10 gigawatts' worth of AI chips in 2027, a slight increase from previous estimates." From CNBC (Nvidia 10-Q): > "Some of our customers are developing their own ASICs and other products, including designs optimized for certain workloads that may not require all of the features and functionality our data center systems provide," Nvidia said in the filing... Let's write this note under the slug `broadcom-q2-2026-earnings-ai-demand`. Let's call `write_note`.

Excellent. Now let's write/update the fourth note: `nvidia-groq-20b-acquihire-inference`. Let's see what we need to include: - In December 2025, Nvidia executed a $20 billion asset acquisition and "acquihire" of the development team at Groq, alongside licensing its LPU (Language Processing Unit) tech. - During the Q1 FY2027 earnings call on May 20, 2026, CEO Jensen Huang provided a key update on this acquisition. - Huang stated that Nvidia's custom AI chips resulting from the Groq acquisition (the Groq LPX/LPU chips) "will be a niche product for some time." - This suggests that while Nvidia is actively developing LPUs and custom ASICs to secure the inference market and counter competitors like Broadcom, these products are not expected to drive significant near-term revenue compared to Nvidia's core GPU lines (Blackwell and Hopper). - However, TrendForce's report on April 8, 2026, noted that NVIDIA is actively expanding into AI inference, with demand for its new LPU solutions expected to reach "several hundred thousand units in 2026, with a target to double in 2027." This indicates a steady, albeit initially niche, ramp of the Groq-derived technology. Let's pull together the quotes: From CNBC: > "Nvidia’s custom AI chip that was the result of its $20 billion acquisition of Groq’s tech 'will be a niche product for some time,' Huang said on the earnings call Wednesday." > "The new Groq chips in LPX are an example of what’s known as application-specific integrated circuits, or ASICs. They’re lower power chips programmed for more specific tasks." From TrendForce: > "Notably, in addition to reinforcing its leadership in high-end AI training, NVIDIA is actively expanding into AI inference applications. Demand for its new LPU solutions is expected to reach several hundred thousand units in 2026, with a target to double in 2027." Let's update `nvidia-groq-20b-acquihire-inference`. Let's call `write_note`.

Amazing! All four notes have been updated with precise, fresh, and deeply-researched information from the latest earnings calls, reports, and market movements of May/June 2026. Let's register a watch for future cycles. What forward-looking catalysts or tripwires did this cycle surface? 1. **The launch of the Vera Rubin platform**: The next-generation Vera Rubin architecture is currently sampling, and first production shipments are expected in Q3 2026, with volume ramping in Q4 2026 into 2027. We should register a watch for the announcement of the first production shipments of Vera Rubin GPUs. - Watch Type: `dated_catalyst` or `live_process` or `tripwire` - Class: `one_shot` or `ongoing` - Let's make it a `one_shot` watch for Q3/Q4 2026. 2. **Apple's September 2026 launch of its revamped Siri**: This was mentioned in the prior findings (`[[apple-outsources-siri-to-nvidia-blackwell]]`). We could register a watch for Apple's Siri launch in September 2026 to see if there is any impact on Blackwell demand or cloud workloads. Let's look at the prior findings list. It says "Apple Outsources Siri Cloud Workloads to Nvidia Blackwell GPUs: Apple is preparing for a landmark September 2026 launch of its revamped Siri...". This is a great dated catalyst! Let's register a watch for it. 3. **Broadcom's FY2027 AI Revenue Guideline**: Broadcom reiterated its $100 billion AI revenue forecast for fiscal 2027. We should watch for any future earnings calls (e.g., Q3 FY2026 in September 2026) where Broadcom might revise this forecast upward or downward, or report on its 10GW shipment progress. Let's call `record_watch` to register these watches. This is a very clean way to set up future cycles. Let's register a watch for "Nvidia Vera Rubin First Production Shipments" (expected Q3/Q4 2026). Let's do that.

Ah, I missed `watch_type`. Let me check the declaration: `watch_type` is an enum of `[dated_catalyst, scheduled_release, metric_threshold, live_process, tripwire, thesis_breaker]`. I will set `watch_type: "dated_catalyst"`. Let's call it again with the correct parameters.

Excellent! The watch has been registered successfully. Let's also register a watch for Apple's Siri launch in September 2026. This is a very interesting catalyst that is highly relevant to the Blackwell demand. Let's do that.

Excellent! Both watches have been registered. Let's double-check if there are any other open threads or actions required. I didn't see any threads returned by `list_threads` earlier, so there are no threads to resolve (`resolve_thread`). I have written/updated all four notes (`hyperscaler-capex-surge-2026`, `nvidia-q1-2027-record-financials-agentic-ai`, `broadcom-q2-2026-earnings-ai-demand`, and `nvidia-groq-20b-acquihire-inference`). Now I am ready to finish the cycle. I will call `finish_cycle`. Let's do it.

I have completed the research cycle and updated the topic's findings. Here is a summary of the key findings regarding whether Nvidia's AI capex story is still intact: ### 1. **Hyperscaler Capex Continues to Accelerate (Now Projected at $725 Billion)** The core narrative supporting the artificial intelligence boom—aggressive capital expenditure by the world's largest technology companies—remains fully intact and has actually accelerated in mid-2026. Combined infrastructure spending from the big four hyperscalers (Microsoft, Alphabet, Meta, and Amazon) is now projected to reach approximately **$725 billion** for calendar year 2026, up from previous high-end estimates of $670 billion: * **Meta Platforms (META):** Raised its 2026 capital expenditure outlook to a range of **$125 billion to $145 billion** (up $10 billion at both ends), citing higher component pricing and additional data center capacity needs. * **Alphabet (GOOGL):** Lifted its full-year 2026 capex forecast to **$180 billion to $190 billion** (up $5 billion at both ends) and signaled that expenditures will increase "significantly" in 2027. * **Microsoft (MSFT):** Projects calendar year 2026 capex to hit **$190 billion**, representing a steep climb from an annualized run-rate of $150 billion earlier in the year. * **Amazon (AMZN):** Maintained its massive capital outlay projections, which are expected to approach **$200 billion** for the year. To fund this massive infrastructure footprint, Alphabet announced a massive **$80 billion capital raise**, with Warren Buffett's Berkshire Hathaway committing **$10 billion** to the deal (purchasing Class A and Class C shares at a 6% and 8% discount, respectively). ### 2. **Nvidia's Record Q1 FY2027 & the Dissolution of the "Blackwell Cliff"** Nvidia's financial results for its fiscal first quarter of 2027 (ended April 30, 2026, reported on May 20, 2026) prove that the massive AI capex surge is translating directly into unprecedented financial performance: * **Revenue:** Reached **$81.62 billion** (vs. $78.86 billion estimated), representing sequential growth from $68.13 billion in the prior quarter and a massive TTM revenue of $253.49 billion. * **Data Center Revenue:** Surged to **$75.2 billion**, accounting for 92% of total sales and nearly doubling from the prior year. * **Net Income & Margins:** Net income reached **$58.32 billion**, representing an astonishing net profit margin of **63.0%**. * **$1 Trillion Backlog:** Jensen Huang confirmed that Nvidia has accumulated **over $1 trillion in combined Blackwell and Vera Rubin purchase orders through 2027**—doubling his previous $500 billion projection from 2025. * **No "Blackwell Cliff":** Wall Street's fear of a buying "air pocket" between Blackwell and Vera Rubin has dissolved. Blackwell's backlog of **3.6 million units** is fully sold out through late 2026. Supply, not demand, remains the binding constraint on Nvidia's growth. * **Blackwell Dominance:** According to TrendForce, the more mature Blackwell platform is estimated to account for **over 70%** of NVIDIA's high-end GPU shipments in 2026, led by the GB300/B300 series. This is reinforced by slight delay risks facing the next-generation Rubin series (due to HBM4 validation, CX8 to CX9 interconnect transition, and power/liquid cooling optimization), which drops Rubin's 2026 shipment share to 22%. ### 3. **The Rise of Custom ASICs as a Parallel Track** While Nvidia's high-end training monopoly remains secure, hyperscalers are actively building parallel, cost-optimized custom silicon to diversify: * **Broadcom's $100 Billion Runway:** Broadcom (AVGO) recently experienced a sharp 14% stock drop on June 4, 2026, simply because it reiterated, rather than raised, its **$100 billion AI revenue forecast for fiscal 2027**. This represents a market expectations reset rather than a broken thesis; Broadcom's current-quarter AI chip sales forecast of $16 billion is up over 3x YoY from $5.2 billion, and Broadcom CEO Hock Tan expects to ship **more than 10 gigawatts** of AI chips in 2027. * **Nvidia's Competitive Disclosures:** In its latest 10-Q filing, Nvidia explicitly acknowledged the growing threat of custom ASICs, stating that customers are developing their own custom chips optimized for specific workloads that do not require all the features of Nvidia's data center systems. ### 4. **Nvidia's Inference Play (Groq/LPUs)** During the Q1 FY2027 earnings call, Jensen Huang tempered near-term expectations on the custom silicon resulting from its $20 billion acquisition of Groq (the Groq LPX/LPU chips), stating they "will be a niche product for some time" as they are lower-power chips designed for specific tasks. However, TrendForce reports that Nvidia's LPU solutions are still projected to reach **several hundred thousand units** in shipments in 2026 and double in 2027 as Nvidia expands aggressively into specialized enterprise AI inference. ### 5. **New Watches Registered** To keep tracking these developments in future cycles, I have registered: 1. **Nvidia Vera Rubin First Production Shipments (dated_catalyst, Q3/Q4 2026):** To monitor the transition timeline from Blackwell. 2. **Apple Revamped Siri Launch with Blackwell Workloads (dated_catalyst, September 2026):** To track the real-world impact of Apple's outsourced Siri cloud workloads on Blackwell demand. All findings have been logged as individual, detailed notes under the topic. Let me know if you would like to explore any of these areas in more detail!