Cycle on Jul 26, 2026

I will start by listing the active threads for this topic to understand what specific questions or watches have been set up, and then check the current market views and search for recent updates regarding Nvidia's AI capex story.

I will fetch the current market view for Nvidia (ticker: NVDA) and the macro backdrop to see how public markets and recent earnings data are shaping up for the company.

I will now search for recent news and discussions regarding "Nvidia AI capex" and "hyperscaler capex Q2 2026 earnings" to check if Microsoft, Google (Alphabet), Meta, or Amazon have reported or updated their capex guidance recently (especially since late July 2026 is when Q2 2026 earnings are released). I'll also look for articles on Blackwell or Rubin platforms.

The search results show that Alphabet (Google) just reported Q2 2026 earnings on July 22, 2026. Let's look closely at the details: - Q2 revenue: $119.8B (+24% YoY) - Q2 capex: $44.9B (more than double YoY) - Q2 free cash flow (FCF): -$5.9B (swung to negative from positive $10.4B last year) - Full-year 2026 capex guidance raised to $195B–$205B (up from $180B–$190B a quarter ago) - Guided 2027 capex to rise significantly. - Alphabet's other income included a ~$99.0B net gain on equity securities (from Anthropic's $965B valuation, since Alphabet holds ~14% which is worth ~$135B). Wait, the net gain on equity securities was $99.03 billion, which lifted net income to $112.11 billion. Let's do a search on other major players to see if Microsoft, Meta, and Amazon have reported yet. The thread `Big Tech Q2 2026 Earnings Capex Revisions` (id: `019f96aa-dbf3-77a5-b4d5-41accd80bf93`) says: "Monitor Microsoft, Amazon, and Meta's Q2 2026 earnings releases (scheduled for late July / early August 2026) to verify if they follow Alphabet's template of raising full-year 2026 capex guidance and signaling further increases for 2027." And: "Fires when all three of Microsoft, Amazon, and Meta have reported Q2 2026 earnings. Check for updated 2026 capex guidance and 2027 commentary." Wait, today is Sunday, July 26, 2026. Let's check the date of Microsoft, Meta, and Amazon's Q2 2026 earnings. The article "Meta and Microsoft walk into earnings week with $145 billion question marks hanging over them" (published Jul 24, 2026) says: "Meta and Microsoft report on July 29 with the same problem in different clothes... One has to explain as much as $145 billion in planned capital spending, the other has to prove a $37 billion AI run rate can carry a much larger buildout." So Microsoft and Meta report on July 29, 2026. Let's see when Amazon reports. Let's do a search for Amazon's Q2 2026 earnings date.

So Microsoft and Meta report on July 29, and Amazon reports on July 30. This means our watch `Big Tech Q2 2026 Earnings Capex Revisions` (id: `019f96aa-dbf3-77a5-b4d5-41accd80bf93`) has not fired yet, but Alphabet's earnings on July 22, 2026, have set a clear template: - Alphabet raised full-year 2026 capex guidance to as high as $205 billion (up from $180B–$190B) and signaled that 2027 capex will rise significantly. - Alphabet's Q2 capex was $44.9 billion, more than double year-over-year. - This massive spending is driven by AI servers (roughly 60% of capex), including Google's own TPUs and Nvidia GPUs. - This massive spending caused Alphabet's free cash flow to swing to negative $5.9 billion in Q2 2026 (from positive $10.4B in Q2 2025), which has triggered some market anxiety regarding depreciation risks and the sustainability of the AI capex boom. Let's check if there are any other recent updates on Nvidia's results or supply chain that we can include. Wait! Let's search for "Nvidia Kyber NVL144", "Vera Rubin", "RTX Spark", or other Nvidia hardware related to the active watches. Let's search for "Nvidia" and "Kyber" or "NVL144" or "Rubin" to see if there's any news. Or "RTX Spark". Let's do that.

Very interesting. We have some developments regarding Nvidia's upcoming Kyber NVL144 rack-scale systems and the Vera Rubin platform: - Earlier this month (July 2026), research firm SemiAnalysis reported that Nvidia's Kyber NVL144 architecture (designed to connect 144 Rubin Ultra GPUs using a copper-based NVLink 7 fabric) faced "major setbacks" and would be delayed by over 12 months, pushing its launch from late 2027 to 2028. - Nvidia's CEO Jensen Huang officially denied these reports, stating that the company is on track to produce "giant amounts" of Vera Rubin-based products and that its chip roadmap remains intact. - This relates directly to our watch: `Nvidia Kyber NVL144 rack commercial shipment verification` (id: `019f6d77-03de-7a2b-bf6d-aacefbf423de`). Let's fetch the full article on Tom's Hardware or Yahoo Finance to get concrete quotes regarding this. Let's fetch the Tom's Hardware article: `https://www.tomshardware.com/tech-industry/artificial-intelligence/nvidias-huang-vows-to-deliver-giant-amounts-of-vera-rubin-company-says-that-our-roadmap-is-intact` or Yahoo Finance: `https://finance.yahoo.com/technology/ai/articles/nvidia-ceo-jensen-huang-dismisses-233017030.html` (wait, the URL in the search results is `https://finance.yahoo.com/technology/ai/articles/nvidia-ceo-jensen-huang-dismisses-233017030.html` or similar. Let's fetch the Tom's Hardware one). Let's do a search to get the exact URL for the Tom's Hardware article. It was: `https://www.tomshardware.com/tech-industry/artificial-intelligence/nvidias-huang-vows-to-deliver-giant-amounts-of-vera-rubin-company-says-that-our-roadmap-is-intact`. Let's fetch it.

Ah, that Tom's Hardware page has a paywall or cookie wall and the main text is missing. Let me try fetching another URL, like `https://wccftech.com/nvidia-quashes-rubin-kyber-rack-delay-rumors-says-chip-roadmap-is-intact/` or `https://finance.yahoo.com/technology/ai/articles/nvidia-ceo-jensen-huang-dismisses-233017030.html`. Let's fetch the Wccftech one first. Let's do a quick search for the exact URL or use the one from the search results. The search results showed `https://wccftech.com/nvidia-quashes-rubin-kyber-rack-delay-rumors-says-chip-roadmap-is-intact/`. Let's fetch it.

This is extremely valuable. The SemiAnalysis report claimed that the quad-die Rubin Ultra was scaled back to a dual-die solution, and Kyber racks featuring Rubin Ultra chips were delayed to 2028. However, Nvidia officially denied this on July 6, 2026, stating that its AI chip roadmap remains intact and disputing the SemiAnalysis delay report. Let's do a search on Alphabet's Q2 2026 earnings transcript or reports to understand what they said on the call regarding their capex, their spending on GPUs vs TPUs, and their demand for compute. Let's search for "Alphabet Q2 2026 earnings transcript capex" or similar.

This is extremely interesting. Let's analyze the landscape right now. 1. **Alphabet's Q2 2026 Earnings (reported on July 22, 2026):** - **Revenue:** $119.8B (+24% YoY). - **Google Cloud Revenue:** $24.8B (+82% YoY), which represents a major acceleration. Operating income for Cloud more than tripled to $8.8B (operating margin rose to 35.6%). Cloud backlog grew by more than $50B sequentially to $514B. - **Capex:** $44.9B in Q2 2026 alone, up 100% YoY (from $22.4B in Q2 2025) and up 26% sequentially (from $35.7B in Q1 2026). The vast majority of this was spent on technical infrastructure to support AI. Specifically, 60% of this was spent on servers (primarily GPUs and TPUs) and 40% on data centers and networking. - **Free Cash Flow:** Swung to negative $5.9B in Q2 2026 (from positive $10.4B in Q2 2025), reflecting the intense cash burn of the capex ramp. - **Full-Year 2026 Capex Guidance:** Raised to $195B–$205B (up from $180B–$190B guided in Q1, and $175B–$185B guided in Q4 2025). This represents a massive increase and a clear signal that Alphabet is accelerating capacity delivery. - **2027 Capex Guidance:** CFO Anat Ashkenazi stated on the call that capex is expected to "increase significantly in 2027." - **Chip Strategy (TPU vs. GPU):** Alphabet is pursuing a full-stack hybrid approach, using both its in-house TPUs (TPU 8t and 8i) and NVIDIA's GPUs (including the new Vera Rubin platform). Interestingly, Google began recognizing revenue from direct sales of its TPU chips to customer data centers for the first time in Q2 2026, though the vast majority of this revenue is expected to realize in 2027. They also reported that they are supply-constrained and are using third-party capacity (e.g., renting Nvidia GPUs from SpaceX for $920 million/month and expanding third-party capacity in Q3 as a bridging strategy). - **Market Reaction:** Despite the revenue and earnings beat, Alphabet's stock sank (down 4.24% in after-hours to $327.4) due to investor anxiety over the capex surge, FCF compression, and upcoming depreciation risks. 2. **Nvidia's Supply Chain & Product Roadmap Developments:** - **Vera Rubin & Kyber NVL144 Racks:** Earlier in July 2026, rumors circulated (originated by SemiAnalysis) that Nvidia's upcoming Kyber NVL144 architecture (designed to connect 144 Rubin Ultra GPUs) faced "major setbacks" and would be delayed by over 12 months to 2028, and that the quad-die Rubin Ultra was scaled back to a dual-die solution. - **Nvidia's Denial:** Nvidia immediately and officially denied these reports on July 6, 2026, stating that its AI chip roadmap remains intact and disputing the SemiAnalysis delay report. CEO Jensen Huang affirmed that the company is on track to produce "giant amounts" of Vera Rubin-based products. - **Integration of Vera Rubin:** In Alphabet's Q2 earnings call, CEO Sundar Pichai explicitly highlighted their deployment of the "new NVIDIA Vera Rubin platform" alongside Google's TPU 8t and 8i, indicating that hyperscalers are already preparing to integrate Nvidia's next-generation platform. 3. **Status of Active Watches:** - **Big Tech Q2 2026 Earnings Capex Revisions (id: `019f96aa-dbf3-77a5-b4d5-41accd80bf93`):** This watch is set to fire when all three of Microsoft, Amazon, and Meta have reported Q2 2026 earnings. Microsoft and Meta report on July 29, 2026, and Amazon reports on July 30, 2026. Therefore, the watch has not fired yet, but Alphabet's report on July 22 provides the initial catalyst. - **Nvidia Kyber NVL144 rack commercial shipment verification (id: `019f6d77-03de-7a2b-bf6d-aacefbf423de`):** Since this is a one-shot watch for late 2027, we can record the event of Nvidia's denial of Kyber NVL144 delays as a temporary update, but the watch itself remains open since the fire date is late 2027. Wait, `record_event` is for when a watch has *fired* (resolves: true/false). Since the watch is "Verify if Nvidia begins commercial shipments of the Kyber NVL144 rack-scale systems by late 2027 or if SemiAnalysis's predicted delay to 2028 is borne out", this watch has not fully resolved, but we can update our findings in the note. - Wait, let's see if we should record any events. Is there any watch that has fired? - Let's check `Nvidia Q2 FY2027 Earnings Release` (id: `019ed2f6-f136-7051-a5b0-fa3b26ec02e0`): Fire date is August 20, 2026. Nvidia has not reported yet. - Let's check `Nebius Group Q2 2026 earnings and margin trajectory` (id: `019edd49-1ccb-73d3-9fc0-80ae266f32c4`): Fire date is August 15, 2026. Not reported yet. - Let's check `Anthropic public S-1 filing and IPO debut` (id: `019ee796-b599-775e-b0db-e799779d71a4`): Fire date is November 15, 2026. Not yet. - Let's check if there are any other watches. None of them have their fire dates reached yet. Let's read the notes we have to make sure we update the correct slugs. Let's call `read_notes` first.

Let's look at the notes we have: - `nvidia-record-q1-results-and-trillion-dollar-order-book`: "Nvidia's AI Capex Story Remains Intact Supported by Record Earnings and Massive Supply Chain Expansion" - `big-tech-capex-reaches-710b-in-2026`: "Big Tech's AI Capex Reaches Historic Scale as 2026 Sprint Approaches $725 Billion" - `custom-silicon-and-datacenter-hardware-competition`: "Custom Silicon and Data Center Hardware Competition Escalates to Challenge Nvidia's Moat" Let's review these and update them based on our fresh findings from Alphabet's Q2 earnings (July 22, 2026) and Nvidia's roadmap denial (July 6, 2026). Wait, let's read the full content of `nvidia-record-q1-results-and-trillion-dollar-order-book` and `big-tech-capex-reaches-710b-in-2026` first to make sure we don't overwrite or duplicate information, and that we integrate the new findings cleanly. But wait, how can we read the full content of a note? Since we don't have a specific `read_note_body` tool, let's see if we can search or read them. Wait, `read_notes` only returns a preview. Ah! Can we search or read the markdown files directly? Yes! The guidelines say: "Your workspace is at: /var/lib/hey-lefty/notes/... Long-term memory: ... History log: ... Before modifying a file, read it first. Do not assume files or directories exist." Wait, does `read_notes` have a way to return the full body or do we read them from the file system? Let's check if the notes are stored as files in our workspace, or if we can use another tool. Wait, let's check if we have standard bash or file tools. We don't see any file tools in the declarations (the declarations only show `list_threads`, `market_view`, `read_notes`, `record_event`, `record_thread`, `record_watch`, `resolve_thread`, `web_fetch`, `web_search`, `write_note`, `finish_cycle`). Wait! There are no file editing tools declared in the tool definitions. The only tools declared are: - `default_api:list_threads` - `default_api:market_view` - `default_api:read_notes` - `default_api:record_event` - `default_api:record_thread` - `default_api:record_watch` - `default_api:resolve_thread` - `default_api:web_fetch` - `default_api:web_search` - `default_api:write_note` - `default_api:finish_cycle` Ah! So we do not have file system tools like bash or cat in our tool declarations. We must use `read_notes` and `write_note` to manage notes, and we can't read the full body of a note directly through a tool except that `read_notes` gives us a preview, or wait, does `read_notes` have a parameter to return the full body? Let's check the parameters of `read_notes`: - `limit` (default 50, max 200) - `topic_id` (required) No other parameters are declared. This means `read_notes` only returns the list of notes with slugs, titles, previews, updated_at, and source_counts. Wait, if we write a note with the same slug, it will overwrite/update it. Since we want to update the findings in place, we will write a comprehensive note with the same slug. Let's make sure we write a highly detailed note that incorporates the prior findings (which we can infer from the previews and our understanding of the topic) and adds the new Q2 2026 earnings details and Nvidia's roadmap updates. Let's look at the previews of the relevant notes: 1. `nvidia-record-q1-results-and-trillion-dollar-order-book`: "Nvidia's artificial intelligence capital expenditure story remains remarkably robust in late July 2026, anchored by blowout financial results, unprecedented hyperscaler capex commitments, and a..." 2. `big-tech-capex-reaches-710b-in-2026`: "The capital expenditure boom supporting artificial intelligence infrastructure has reached unprecedented heights in late July 2026. The world's largest technology companies—Alphabet, Microsoft,..." 3. `custom-silicon-and-datacenter-hardware-competition`: "While Nvidia's near-to-mid-term artificial intelligence infrastructure demand remains exceptionally robust (see [[nvidia-record-q1-results-and-trillion-dollar-order-book]]), the long-term competitive..." Let's do some more search on "Nvidia Q1 2026 results" or "Nvidia Q1 FY2027 results" (wait, Nvidia's fiscal year is ahead, so Q1 FY2027 was reported in May 2026, and Q2 FY2027 will be reported in August 2026). Let's verify Nvidia's latest quarter numbers from the `market_view` we pulled: - Quarter ending April 30, 2026 (this is Q1 FY2027): Revenue was $81.61B, Net income was $58.32B, Operating income was $53.54B, EBITDA was $71.00B. - Gross margin was 74.1%. - TTM Revenue is $253.49B. - Market capitalization is $5.01T. Let's search for "Nvidia Q1 FY2027 earnings" or "Nvidia Q1 2026 earnings" to confirm the details of their latest earnings report so we can carry them forward accurately in our note.

This is very helpful! Let's review the core facts we've gathered: **1. Nvidia's Financials (from Q1 FY2027 reported May 20, 2026, and TTM / Market Cap from `market_view`):** - **Q1 FY2027 Revenue:** $81.6B (+85% YoY), beating Wall Street estimates of ~$78.9B. - **Data Center Revenue:** $75.2B (+92% YoY, representing 92.1% of total revenue), driven by Blackwell system shipments. - **Adjusted EPS:** $1.87 (vs. $1.76/1.78 expected). - **Free Cash Flow:** $48.59B in Q1 FY2027. - **Gross Margin:** 74.1%. - **Nvidia's Cumulative Revenue Guidance:** CEO Jensen Huang stated Nvidia expects "at least" $1 trillion in cumulative revenue from its Blackwell and Vera Rubin chips through the end of 2027. He added: "I am certain computing demand will be much higher than that... In fact, we are going to be short." - **Q2 FY2027 Guidance:** Guided to $91 billion (+/- 2%). **2. Alphabet's Live Q2 2026 Earnings (reported July 22, 2026):** - **Revenue:** $119.8B (+24% YoY). - **Google Cloud Revenue:** $24.8B (+82% YoY), with operating income of $8.8B (margin 35.6%) and a backlog of $514B (up $50B sequentially). - **Q2 Capex:** $44.9B, up 100% YoY (from $22.4B in Q2 2025) and up 26% sequentially (from $35.7B in Q1 2026). - **Free Cash Flow:** Slashed to negative $5.9B (compared to positive $10.4B in Q2 2025) due to the capex sprint. - **Full-Year 2026 Capex Guidance:** Raised to $195B–$205B (up from $180B–$190B previously). - **2027 Capex Guidance:** Management guided that 2027 capex will "increase significantly." - **Nvidia Integration:** Sundar Pichai explicitly mentioned that Google Cloud is deploying "the new NVIDIA Vera Rubin platform" alongside its own TPU 8t and 8i. - This demonstrates that the AI capex story is not only intact but *accelerating*, even as Wall Street voices near-term concerns about free cash flow depletion and depreciation risks. **3. Nvidia's Roadmap Denial & Kyber Rack Rumors (July 6, 2026):** - **The Delay Rumors:** Research firm SemiAnalysis published a report claiming that Nvidia's advanced Kyber NVL144 server architecture (designed to scale up to 144 Rubin Ultra GPUs using copper-based NVLink 7) faced "major setbacks" and was delayed by over a year, pushing its release from late 2027 to 2028. They also claimed the quad-die Rubin Ultra was scaled down to a dual-die solution. - **Nvidia's Response:** On July 6, 2026, Nvidia officially denied these rumors, disputing the SemiAnalysis report and stating that its AI chip roadmap remains intact. CEO Jensen Huang affirmed that the company is on track to produce "giant amounts" of Vera Rubin-based products. Wccftech noted that historical delay rumors (like thermal or HBM issues with Blackwell) were resolved, and Blackwell entered production ahead of schedule. Let's now update our findings using the correct slugs. The relevant slugs are: - `nvidia-record-q1-results-and-trillion-dollar-order-book` (for Nvidia's Q1 FY2027 and the overall capex/demand story) - `big-tech-capex-reaches-710b-in-2026` (for Big Tech's capex scale, updating it with Alphabet's Q2 earnings and the raised 2026/2027 guidance) - `custom-silicon-and-datacenter-hardware-competition` (for the competition between TPU and GPU, updating it with Alphabet's direct TPU sales starting in Q2 2026, and the hybrid deployment model) Wait, let's look at the rules for writing findings: - Write ONE note per theme, never a combined write-up. - Name and link the actual thing (the story, article, paper, or page) — not just "the discussion". - Carry 1-2 REPRESENTATIVE VERBATIM QUOTES, copied exactly, each kept with the link it came from. - Say what it means. - Use stable slugs. Use `write_note` with the SAME slug to update. - Do NOT use YAML frontmatter. The title is the first `# ` heading. Let's carefully construct each note. ### Note 1: `nvidia-record-q1-results-and-trillion-dollar-order-book` We will update this note to reflect that Nvidia's AI capex story remains exceptionally strong, backed by Alphabet's massive capex guidance upgrade in late July 2026 and Nvidia's official denial of Kyber NVL144 platform delays on July 6, 2026. Let's write this note first. Wait, let's make sure we have the exact quotes and links. From the Wccftech article on Nvidia denying Kyber delays: Link: `https://wccftech.com/nvidia-quashes-rubin-kyber-rack-delay-rumors-says-chip-roadmap-is-intact/` Quote: "NVIDIA has denied reports of a delay regarding its upcoming Rubin & Kyber products, stating that its chip roadmap remains intact." Quote: "The rumor was initiated by SemiAnalysis, which said that Kyber racks featuring the NVIDIA Rubin Ultra chips were delayed to 2028." From the Tech-Insider article on Nvidia's Q1 FY2027 results: Link: `https://tech-insider.org/nvidia-earnings-81-billion-quarter-2026/` Quote: "The chipmaker posted $81.6 billion in revenue for its first quarter of fiscal 2027, up 85% year over year, with data center revenue alone reaching $75.2 billion—a 92% surge that underscores how completely the AI buildout now revolves around one company’s silicon." Quote: "He said the company expects “at least” $1 trillion in cumulative revenue from its Blackwell and Vera Rubin chips through the end of 2027—and then went further." Let's call `write_note` for `nvidia-record-q1-results-and-trillion-dollar-order-book`.

Excellent! Now let's update the second note: `big-tech-capex-reaches-710b-in-2026`. We want to update this note with the details of Alphabet's Q2 2026 earnings report on July 22, 2026. Let's review the details: - Alphabet's Q2 capex was $44.9 billion, up 100% year-over-year. - Alphabet raised its full-year 2026 capex guidance to $195B–$205B (up from $180B–$190B guided in Q1, and $175B–$185B in Q4 2025). - CFO Anat Ashkenazi guided that 2027 capex is expected to "increase significantly." - Alphabet's aggressive spending caused its Q2 free cash flow to turn negative ($5.9 billion), down from positive $10.4 billion in Q2 2025. - This has triggered Wall Street concerns about depreciation risk and FCF depletion, but the hyperscalers are ignoring the market's frosty reaction because they are supply-constrained and see massive demand (Google Cloud grew 82% to $24.8 billion). - Combined 2026 capex for the four largest hyperscalers (Alphabet, Microsoft, Meta, and Amazon) is on pace to reach $725 billion. Let's find the exact quotes and links for Alphabet's earnings. From the Alphabet Q2 2026 earnings transcript (Yahoo Finance / Investing.com): Link: `https://finance.yahoo.com/quote/GOOG/earnings/GOOG-Q2-2026-earnings_call-657320.html/` OR `https://www.investing.com/news/transcripts/earnings-call-transcript-alphabet-beats-q2-2026-estimates-shares-fall-on-capex-surge-93CH-4807140` Quote: "CapEx was $44.9 billion in the second quarter, with the vast majority of this spent in technical infrastructure to support our investments in AI." Quote: "We had negative free cash flow of $5.9 billion in the second quarter, driven by our investments in CapEx." Quote: "Moving to investments, we are updating our full-year 2026 CapEx guidance range to $195 billion-$205 billion, up from our previous estimate of $180 billion-$190 billion. As we previously shared, we continue to expect our CapEx to increase significantly in 2027..." From the Startup Fortune article: Link: `https://startupfortune.com/meta-and-microsoft-walk-into-earnings week-with-145-billion-question-marks-hanging-over-them/` Quote: "Together with Alphabet and Amazon, the four hyperscalers are on pace to spend $725 billion on AI infrastructure in 2026..." Let's call `write_note` for `big-tech-capex-reaches-710b-in-2026`.

Fantastic! Now let's update the third note: `custom-silicon-and-datacenter-hardware-competition`. We want to update this note with findings from Alphabet's Q2 earnings call regarding Google's TPU strategy, its direct TPU sales, and the hybrid deployment model. Let's review the details: - Alphabet is selling its custom TPU chips directly to external customers and neoclouds (e.g., trying to sell custom TPUs directly to neoclouds like Nscale to compete with Nvidia). - Google recognized revenue from direct sales of its TPU systems to customer data centers for the first time in Q2 2026. - The vast majority of TPU system sales revenue is expected to be recognized in 2027, with a smaller portion recognized in 2026. - Google employs a hybrid deployment model using both custom TPUs (TPU 8t for training and TPU 8i for inference) and Nvidia GPUs (including the new Vera Rubin platform). - A Deutsche Bank analysis points out that this hybrid deployment allows Google to keep its unit construction costs lower than market concerns suggest. - Google's TPU advantage is significant: for every Nvidia GPU Google deploys internally, it deploys approximately nine custom TPUs (a 1:9 ratio). - However, Google remains supply-constrained and is renting Nvidia GPUs from SpaceX for $920 million/month to keep up with demand, showing that custom silicon is still not enough to bypass Nvidia entirely. Let's find the exact quotes and links. From the Stocktwits/TradingView article on Google expanding TPU sales to battle Nvidia: Link: `https://www.tradingview.com/news/stocktwits:354636929094b:0-googl-reportedly-expands-tpu-sales-to-battle-nvidia-amid-wall-street-price-target-cuts-on-big-tech-ahead-of-q2-earnings/` Quote: "According to The Information, Google's tensor processing units (TPUs) have so far been housed almost entirely in Google's own facilities and rented out only through Google Cloud. Now the company is trying to sell its custom AI chips directly to 'neoclouds,' young companies built around renting out Nvidia GPUs." Quote: "It said that one of Google’s targets was Nscale, a two-year-old neocloud in which Nvidia is a major investor and preferred shareholder." From the Tech Times article on Alphabet Q2 earnings: Link: `https://www.techtimes.com/articles/321312/20260722/alphabet-q2-earnings-cloud-revenue-not-80-billion-paper-eps-tells-story.htm` Quote: "Unlike AWS and Microsoft, which primarily rely on NVIDIA's GPUs for AI inference, Alphabet has built its compute stack around custom ASICs it designed entirely in-house..." Quote: "One analysis found that for every NVIDIA GPU Google deploys internally, it deploys approximately nine custom TPUs — a 1:9 ratio that illustrates just how deeply the proprietary silicon advantage is embedded in Alphabet's infrastructure economics." Let's call `write_note` for `custom-silicon-and-datacenter-hardware-competition`.

Excellent! We have successfully updated the three key notes: 1. `nvidia-record-q1-results-and-trillion-dollar-order-book` 2. `big-tech-capex-reaches-710b-in-2026` 3. `custom-silicon-and-datacenter-hardware-competition` Let's check if there are any other notes we should update or if we should record any threads or watches. Wait, let's review the active threads and watches: - `Big Tech Q2 2026 Earnings Capex Revisions` (id: `019f96aa-dbf3-77a5-b4d5-41accd80bf93`): This is set to fire when all three of Microsoft, Amazon, and Meta have reported Q2 2026 earnings. Since they report on July 29 and July 30, 2026, this watch has *not* fired yet. It remains open. - `Nvidia Kyber NVL144 rack commercial shipment verification` (id: `019f6d77-03de-7a2b-bf6d-aacefbf423de`): This watch is to verify if Nvidia begins commercial shipments of the Kyber NVL144 rack-scale systems by late 2027 or if SemiAnalysis's predicted delay to 2028 is borne out. This watch is for late 2027, so it remains open. - `Nvidia Q2 FY2027 Earnings Release` (id: `019ed2f6-f136-7051-a5b0-fa3b26ec02e0`): Expected late August 2026. Remains open. Wait, are there any new watches we should register? Let's see: - Alphabet's direct TPU sales ramp and 2027 revenue recognition: This is already covered by the general custom silicon competition, but we can register a watch for Google's Q3 2026 and Q4 2026 earnings to track the exact revenue contribution and margin impact of direct TPU system sales. Wait, let's check if we have any other forward-looking catalysts. Alphabet's use of third-party capacity (the "bridging strategy") in Q3 2026 could put pressure on Google Cloud operating margins. We can register a watch for Google Cloud's Q3 2026 operating margin to see if it dips from the 35.6% reported in Q2 2026 due to third-party capacity costs. Let's register this watch! Let's check the parameters of `record_watch`: - `topic_id` (required) - `cycle_id` (required) - `title` (required) - `cardinality` (required, enum: `one_shot`, `ongoing`) - `watch_type` (required, enum: `dated_catalyst`, `scheduled_release`, `metric_threshold`, `live_process`, `tripwire`, `thesis_breaker`) - `description` - `entity` - `fire_date` (RFC3339 date) - `metric` - `op` - `threshold` - `note` Let's register a watch for Google Cloud's Q3 2026 operating margin compression. - Title: `Google Cloud Q3 2026 Operating Margin and Third-Party Capacity Impact` - Cardinality: `one_shot` - Watch Type: `scheduled_release` - Entity: `GOOG` - Fire Date: `2026-10-25T00:00:00Z` (expected release date for Q3 2026 earnings is late October 2026) - Description: `Track Google Cloud's Q3 2026 operating margin to verify if the bridging strategy of using third-party capacity puts downward pressure on the 35.6% operating margin achieved in Q2 2026.` - Metric: `operating_margin` - Op: `<` - Threshold: `35.6` - Note: `Fires when Alphabet reports Q3 2026 earnings. Check Google Cloud operating margin vs Q2's 35.6% and assess the margin impact of third-party capacity rentals.` Let's call `record_watch` for this.

Perfect! We have successfully registered the watch. Let's call `finish_cycle` to complete this research cycle, as we have updated all three key notes, grounded them in concrete evidence, and registered a forward-looking watch. Wait, let's verify if we need to call `resolve_thread` or `record_thread`. None of the listed threads have resolved yet, as Microsoft, Meta, and Amazon have not reported Q2 earnings (they report on July 29 and July 30, and the watch fire date is August 5, 2026). So we do not resolve any of them. Let's call `finish_cycle` now.

I have completed the research cycle on the status of Nvidia's AI capex story. Below is a summary of the findings and updates made during this cycle. ### Executive Summary Nvidia’s AI capex story remains **exceptionally strong and fully intact** in late July 2026. While Wall Street has voiced near-term anxieties regarding capital intensity, free cash flow compression, and depreciation risks, the actual demand from hyperscalers is accelerating. This cycle’s findings are anchored by two major developments: 1. **Alphabet’s Monumental Q2 2026 Earnings (Reported July 22, 2026):** Alphabet kicked off the tech earnings season by doubling down on its AI infrastructure bet. It reported a staggering **$44.9 billion in Q2 capex** (a 100% year-over-year surge) and raised its full-year 2026 capex guidance to **$195 billion to $205 billion** (up from $180B–$190B). Crucially, CFO Anat Ashkenazi guided that 2027 capex will "increase significantly." 2. **Nvidia’s Swift Denial of Kyber Platform Delays (July 6, 2026):** Following a report from research firm SemiAnalysis claiming that Nvidia’s advanced Kyber NVL144 server architecture (designed for 144 Rubin Ultra GPUs) was delayed by over a year to 2028, Nvidia officially denied the rumors. The company confirmed that its AI chip roadmap remains fully intact, with CEO Jensen Huang affirming they are on track to deliver "giant amounts" of Vera Rubin-based products. --- ### Key Findings & Document Revisions I have updated three core findings under stable slugs to reflect these developments: #### 1. Nvidia's AI Capex Story Remains Intact Supported by Record Earnings and Massive Supply Chain Expansion * **Slug:** `nvidia-record-q1-results-and-trillion-dollar-order-book` * **Update:** Integrated Nvidia's blowout Q1 FY2027 results (reported late May 2026) featuring **$81.6 billion in revenue** (+85% YoY) and **$75.2 billion in Data Center revenue** (+92% YoY). Carrying forward Nvidia's multi-year cumulative revenue target of **"at least" $1 trillion** from Blackwell and Vera Rubin platforms through 2027, and detailing Nvidia's official denial of the Kyber NVL144 delays. * **Source Citation:** [/markets/NVDA/2026/07/26](/markets/NVDA/2026/07/26) #### 2. Big Tech's AI Capex Reaches Historic Scale as 2026 Sprint Approaches $725 Billion * **Slug:** `big-tech-capex-reaches-710b-in-2026` * **Update:** Documented Alphabet's Q2 2026 earnings results, including its $44.9B quarterly capex, negative $5.9B free cash flow swing (its first cash-burning quarter in over two decades), and raised full-year guidance of $195B–$205B. This positions the four largest hyperscalers (Alphabet, Microsoft, Meta, and Amazon) to spend a combined **$725 billion** on AI infrastructure in 2026. * **Source Citation:** [Alphabet Inc. (GOOG) Q2 FY2026 earnings call transcript](https://finance.yahoo.com/quote/GOOG/earnings/GOOG-GOOG-Q2-2026-earnings_call-657320.html/) #### 3. Custom Silicon and Data Center Hardware Competition Escalates to Challenge Nvidia's Moat * **Slug:** `custom-silicon-and-datacenter-hardware-competition` * **Update:** Explored Google's massive TPU program (utilizing an estimated 1:9 Nvidia GPU to custom TPU deployment ratio). Documented a historic milestone: in Q2 2026, Google recognized revenue from **direct sales of its TPU systems to customer data centers** for the first time, taking the custom silicon fight directly to Nvidia's neocloud turf (such as Nscale). However, the finding highlights that custom silicon is still insufficient to bypass Nvidia, as Google remains supply-constrained and is paying SpaceX **$920 million per month** to rent Nvidia GPU capacity. * **Source Citation:** [Alphabet Q2 Earnings: Cloud Revenue, Not an $80 Billion Paper EPS, Tells the Story](https://www.techtimes.com/articles/321312/20260722/alphabet-q2-earnings-cloud-revenue-not-80-billion-paper-eps-tells-story.htm) --- ### Forward-Looking Monitoring Registered To track the near-term margin pressure resulting from hyperscaler supply deficits, I registered a new watch: * **Title:** Google Cloud Q3 2026 Operating Margin and Third-Party Capacity Impact (`one_shot`, `scheduled_release`) * **Entity:** `GOOG` * **Fire Date:** October 25, 2026 * **Description:** Track Google Cloud's Q3 2026 operating margin to verify if the bridging strategy of using third-party capacity (renting outside GPUs) puts downward pressure on the 35.6% operating margin achieved in Q2 2026. The remaining watches on Microsoft, Meta, and Amazon's Q2 earnings (reporting July 29 and July 30, 2026) remain active and will be resolved in upcoming cycles.