Cycle on Jul 20, 2026

I will start by listing any open threads and reading the existing notes on this topic to understand what has already been researched and what specific questions are currently active. I will also fetch the market view for NVIDIA (NVDA) to anchor our analysis in the latest financial and market data.

I will call `read_notes` to retrieve the existing notes for this topic and understand the specific details of our previous findings.

I will now fetch the market view for NVIDIA (NVDA) using `market_view` to ground our research in the company's financial and stock performance metrics.

The market view shows NVIDIA (NVDA) has a market cap of $4.91T, TTM revenue of $253.49B (+85.2% YoY), gross margin of 74.1%, and reported revenue for the quarter ending April 30, 2026, of $81.61B. Its cash is $13.24B vs total debt of $12.81B. In the latest quarter, it beat EPS estimates ($1.87 reported vs $1.77 estimated). Let's search for recent news on Nvidia's AI capex story, tech hyperscaler capex, demand for Blackwell/Rubin, and any potential threats or corroborating evidence of the capex story. The current date is July 20, 2026. Let's do a search for recent articles on "Nvidia capex", "hyperscaler capex", "Nvidia Blackwell demand H2 2026", "Nvidia Rubin demand", "AI capex intact" or similar. I'll search with operator `recent=true` or freshness="week" or freshness="month". Let's start with a general search to see what's happening.

This is fascinating! We have some very fresh articles from the past week (mid-July 2026): 1. TSMC raises capex and revenue forecast, highlighting growing AI chip demand (July 16, 2026). 2. TSMC to invest another $100 billion in US as Q2 profit blows past forecasts (July 15, 2026). 3. Alphabet's record AI spending signals windfall for Nvidia - Alphabet lifted 2026 CapEx guidance to $180-$190 billion, with most going toward AI infrastructure and further increases expected in 2027. 4. Michael Burry says investors who are hoping for a world of both limitless demand for AI chips and less spending by AI hyperscalers need a reality check (July 2026). 5. The 1 Number That Sets the Record Straight on AI Chip Demand: $75 Billion (July 14, 2026). 6. Reuters: Apple unseats Nvidia to become world's most valuable company (July 17, 2026). Let's fetch these URLs using `web_fetch` to see the actual content and get the quotes and details. First, let's fetch the TSMC capex and revenue forecast story. Let's do a search to find the full URL of the Yahoo Finance article or Reuters article about TSMC raising capex. Wait, Yahoo Finance article: `https://finance.yahoo.com/markets/article/tsmc-raises-capex-and-revenue-forecast-highlighting-growing-ai-chip-demand-113101950.html` (let's check if we can fetch it, or if there's a Reuters / Bloomberg equivalent we can search for). Let's search for "TSMC raises capex 2026" or "TSMC Q2 2026 earnings capex". Let's run a search.

Wow! TSMC literally just reported its Q2 2026 earnings (around July 15-16, 2026) and raised its 2026 revenue growth outlook to "slightly above 40%" (from 30%-plus previously) and increased its 2026 capex guidance to a range of $60 billion to $64 billion (up from $52 billion to $56 billion previously). This is a massive, concrete piece of evidence that the AI capex boom and demand for advanced chips (like Nvidia's) is extremely high and actually accelerating. Let's fetch the Seeking Alpha article or the Yahoo Finance article or the Quartz article to get the exact quotes. Let's try `web_fetch` on `https://qz.com/tsmc-record-profit-100-billion-arizona-investment-071626` or `https://finance.yahoo.com/markets/article/tsmc-raises-capex-and-revenue-forecast-highlighting-growing-ai-chip-demand-113101950.html` or `https://seekingalpha.com/article/4923307-tsmc-the-ai-supercycle-just-got-stronger`. Let's fetch the Seeking Alpha article `https://seekingalpha.com/article/4923307-tsmc-the-ai-supercycle-just-got-stronger` and the Quartz article `https://qz.com/tsmc-record-profit-100-billion-arizona-investment-071626`. Let's search to see if we can get a clean fetch. Let's fetch `https://finance.yahoo.com/markets/article/tsmc-raises-capex-and-revenue-forecast-highlighting-growing-ai-chip-demand-113101950.html`. Let's try fetching the Yahoo Finance market article first.

Let's also look at the Michael Burry article. Let's do a search to find the full URL of the Business Insider article on Michael Burry and his AI trade comments. Or we can just search for "Michael Burry says there's a major contradiction at the center of the AI trade" and see if there are other articles. Let's run a search.

Let's fetch the MSN or Yahoo Finance version of the Michael Burry article. Let's fetch `https://sg.finance.yahoo.com/news/big-short-investor-michael-burry-164825420.html` to see the exact text of his warning. Let's do `web_fetch`.

This is a highly relevant perspective! Michael Burry's Substack post from July 9, 2026, details a structural contradiction in the AI trade: 1. "The hyperscalers are promising permanent demand growth and temporary spending over 3-4 years all in the same breath. Nvidia needs the permanent demand growth. Hyperscalers need the buildout to be largely over in 3-4 years." 2. "Nvidia's revenue growth is mostly real but the resolution of the bottleneck, any which way or when it happens, reduces the recurrence rate of Nvidia's revenue, which will also take down the scarcity premium and hence margin for both new and old chips." 3. "The hyperscalers are seeing free cash flow fall toward zero already. The reported profit is there, courtesy of long depreciation schedules in part, but the owners' earnings are not." Let's also look into the Alphabet record AI spending story. It says: "Alphabet lifted 2026 CapEx guidance to $180-$190 billion, with most going toward AI infrastructure and further increases expected in..." Wait, let's look at the search results. There's an article: "Alphabet's record AI spending signals windfall for Nvidia - Alphabet lifted 2026 CapEx guidance to $180-$190 billion..." Let's search for "Alphabet 2026 CapEx guidance 180-190 billion" or "Alphabet Q2 2026 earnings CapEx guidance" or similar. Wait, has Alphabet already reported Q2 2026 or did they just pre-announce, or is this a recent development? Let's check the date of the search results and do a search. Let's run a search.

This is perfect. Alphabet's capital expenditure guidance for 2026 was raised in Q1 2026 to $180–$190 billion, up from $175–$185 billion previously. Let's see if there are updates for Q2 2026 earnings. The current date is July 20, 2026. Alphabet's Q2 2026 earnings are expected late July 2026. Let's do a search for when Alphabet's Q2 2026 earnings are scheduled, or if there are any pre-announcements or estimates. Let's also search for Meta's, Microsoft's, and Amazon's capex figures and projections for 2026. Let's run a search.

Ah, Alphabet's Q2 2026 earnings are scheduled for July 22, 2026, which is in just two days (since current date is Monday, July 20, 2026). Wait, let's look at the threads we have. One of the threads is: "Monitor Alphabet's Q2 2026 earnings release (expected late July 2026) to assess the initial dilution impact, share count changes, and interest/dividend obligations arising from its upsized $90 billion equity capital raise." "Alphabet Q2 2026 Earnings and Dilution Impact Tracking" (fire date 2026-07-25). And we have: "Track Nvidia's Q2 FY2027 earnings release (expected late August 2026) to verify if revenue meets or exceeds the guided $91 billion (+/- 2%) and if gross margins remain flat at ~75%." (fire date 2026-08-20). Let's do some more searches on Nvidia's Blackwell/Rubin/Kyber products and overall capex trends. What about Blackwell chip shipments or delay updates? The previous note `nvidia-record-q1-results-and-trillion-dollar-order-book` mentioned "Kyber Rack Delays and Manufacturing Bottlenecks". Let's search for "Kyber rack delay", "Kyber NVL144", "Blackwell shipments 2026", "Vera Rubin platform H2 2026". Let's run a search.

Very interesting! Let's look at: 1. "Rubin Faces Delays; Blackwell to Drive 70%+ of NVIDIA High-End GPU Shipments in 2026" 2. "Nvidia reportedly books entire server plant capacity through 2026 to build Blackwell and Rubin AI servers, pushing out other potential customers." 3. "NVIDIA Quashes Rubin and Kyber Rack Delay Rumors, Says Chip Roadmap Is Intact" 4. "JP Morgan expects NVIDIA to ship 5.2M Blackwell AI GPUs this year, drops to 1.8M in 2026 as Rubin ramps with an expected 5.7M shipped in 2026." (Wait, is this 2025 or 2026? The year of the article is likely 2025/2026. Let's research the dates and details of these articles.) Let's fetch `https://iconnect007.com/article/149537/rubin-faces-delays-blackwell-to-drive-70-of-nvidia-highend-gpu-shipments-in-2026/149534/design` or `https://www.telecompaper.com/news/nvidias-rubin-delays-mean-blackwell-to-dominate-2026-ai-gpu-shipments--1567604` using `web_fetch`. Let's search to see if we can get a clean fetch of the TrendForce report. Let's do a search on "TrendForce Nvidia Blackwell Rubin 2026".

Let's fetch the TrendForce press release: `https://www.trendforce.com/presscenter/news/20260408-13003.html` to see what is happening with Rubin delays and Blackwell shipments. Let's do `web_fetch`.

Let's fetch the other TrendForce press release from May 20, 2026: `https://www.trendforce.com/presscenter/news/20260520-13053.html` to see if there is more up-to-date information on North American hyperscaler procurement of NVIDIA GB and Rubin racks. Let's do `web_fetch`.

This is incredibly detailed and valuable information! Let's summarize the key points from this TrendForce May 20, 2026 report: - The top 5 North American CSPs (Google, Amazon, Microsoft, Meta, Oracle) are expected to account for over 60% of global demand for NVIDIA's GB and VR series servers in 2026. - Combined capital expenditures of Google, Amazon, Microsoft, Meta, and Oracle will exceed **US$770 billion in 2026**, marking a YoY increase of nearly **87%** (up from previous estimates of around $710B-$725B mentioned in our prior findings `big-tech-capex-reaches-710b-in-2026`). - Combined AI training computing power for these top 5 CSPs is projected to grow by more than 56% YoY in 2026 (exceeding 9 ExaFLOPS in 2025). - Combined AI inference computing power (FP4/NVFP4) is expected to surge by **nearly 122% in 2026**, reflecting NVIDIA's focus on optimizing inference performance on GB300 and VR200 rack-scale solutions. - In-house ASICs are also growing: Google's TPU demand will grow in volume by nearly 80% YoY in 2026, transitioning from v7 to v8 in H2 2026. Amazon's Trainium series will account for over 40% of Amazon's own AI server shipments in 2026. - High system power consumption is a major trend: In 2026, the top 5 North American CSPs' total server power consumption is projected to surge by **18 GW** compared to the previous year, a **116% YoY growth rate** in power addition, driven by the mass deployment of platforms like NVIDIA's GB300, AMD's Helios, and custom ASICs. Let's check if there are any other recent articles on "Nvidia Blackwell demand H2 2026" or "Nvidia Q2 FY2027 guidance" or "Nvidia's AI capex story". Wait, let's search for "The 1 Number That Sets the Record Straight on AI Chip Demand: $75 Billion" to see what that article is about. Let's run a search.

Let's fetch `https://247wallst.com/investing/2026/07/14/the-1-number-that-sets-the-record-straight-on-ai-chip-demand-75-billion/` to see what "The 1 Number That Sets the Record Straight on AI Chip Demand: $75 Billion" means. Let's do `web_fetch`.

Excellent! Let's analyze the findings we have gathered: 1. **Nvidia's AI Capex Story and Chip Demand is highly robust and accelerating, supported by supply-side and customer-side data.** - In mid-July 2026, **TSMC** (Nvidia's primary foundry partner) reported Q2 2026 earnings, posting record quarterly revenue of $40.2 billion and **raising its 2026 capex guidance to $60–$64 billion** (up from $52–$56 billion previously, representing at least a $4 billion increase). TSMC also **raised its 2026 revenue growth projection to slightly above 40%** (up from 30%-plus previously). This demonstrates extremely strong demand for high-end AI chips. - **Alphabet** raised its 2026 capex guidance to **$180–$190 billion** (up from $175–$185 billion in the previous quarter), with a significant portion dedicated to AI infrastructure, and indicated that 2027 capex will increase further. - **TrendForce** (May 20, 2026 report) projects that the top 5 North American cloud service providers (CSPs: Google, Amazon, Microsoft, Meta, Oracle) will see combined capital expenditures exceed **$770 billion in 2026**, a massive **87% YoY increase**. This is expected to drive a **122% YoY surge in combined AI inference computing power** and a **56% YoY increase in AI training computing power** in 2026. - In Q1 FY2027, **Nvidia's Data Center revenue reached $75.25 billion** (up 92% YoY), accounting for the vast majority of its $81.61 billion total revenue. Compute revenue was $60.4 billion (up 77% YoY) and networking revenue hit a record $14.8 billion (up 199% YoY). - Nvidia guided Q2 FY2027 revenue to **$91.0 billion** (+/- 2%), backed by **$119.0 billion in supply-related commitments**, showing booked demand extends far beyond the current quarter. - TSMC reported June 2026 net revenue of NT$442.68 billion, up 67.9% YoY. 2. **However, critical structural risks and warnings are beginning to emerge from prominent investors and research analysts.** - **Michael Burry** (July 9, 2026 Substack post) warns of a core contradiction in the AI trade: "The hyperscalers are promising permanent demand growth and temporary spending over 3-4 years all in the same breath. Nvidia needs the permanent demand growth. Hyperscalers need the buildout to be largely over in 3-4 years." - Burry also highlights that hyperscaler free cash flow is falling toward zero due to the massive capex outlays, even if reported GAAP profits remain high due to long depreciation schedules: "The hyperscalers are seeing free cash flow fall toward zero already. The reported profit is there, courtesy of long depreciation schedules in part, but the owners' earnings are not." - Furthermore, Burry warns that the resolution of supply bottlenecks will eventually erode Nvidia's scarcity premium and gross margins: "Nvidia's revenue growth is mostly real but the resolution of the bottleneck... reduces the recurrence rate of Nvidia's revenue, which will also take down the scarcity premium and hence margin for both new and old chips." - **TrendForce** (April 8, 2026 report) notes that while the Blackwell series will dominate shipments in 2026 (projected to account for over 70% of Nvidia's high-end GPU shipments, led by GB300/B300), the successor **Vera Rubin series faces delay risks** due to HBM4 validation, CX9 networking transitions, high power consumption, and liquid cooling optimization. As a result, Rubin's projected share of 2026 high-end shipments was revised down from 29% to 22%. - **Power and Grid Bottlenecks**: TrendForce estimates that the top 5 North American CSPs' total server power consumption will surge by **18 GW** in 2026 compared to the previous year, which is a **116% YoY growth rate** in power addition, highlighting the physical limits of grid capacity as a bottleneck. Let's check if there are any other recent notes we should update or if we should write new notes. First, let's look at our existing notes: - `nvidia-record-q1-results-and-trillion-dollar-order-book` (Nvidia's AI Capex Story Remains Intact Despite Kyber Rack Delays and Manufacturing Bottlenecks) - `big-tech-capex-reaches-710b-in-2026` (Big Tech's AI Capex Reaches Historic Scale as 2026 Infrastructure Sprint Approaches $725 Billion) - `gpu-specialized-clouds-and-strategic-co-investment` (Specialized GPU Clouds Face Competitive Disruption as Meta and SpaceX Reshape the Market) - `custom-silicon-and-datacenter-hardware-competition` (Custom Silicon and Data Center Hardware Competition Escalates to Challenge Nvidia's Moat) Wait! We can update two main notes to reflect these very fresh findings from mid-July 2026: 1. **`nvidia-record-q1-results-and-trillion-dollar-order-book`**: We can update this finding to include the TSMC Q2 2026 earnings results (released mid-July 2026) showing TSMC raising its 2026 capex to $60-$64 billion and raising revenue growth to slightly above 40%. We can also include Michael Burry's July 9, 2026 warning about the core contradiction in the AI capex cycle and the risk to Nvidia's scarcity premium, and TrendForce's April/May 2026 reports. 2. **`big-tech-capex-reaches-710b-in-2026`**: We can update this finding to reflect that Big Tech's AI capex has now reached an even larger scale. TrendForce's May 20, 2026 report projects that the top 5 North American CSPs' combined capital expenditures will exceed **$770 billion in 2026** (a nearly 87% YoY increase), driving a 122% surge in AI inference computing power and an 18 GW surge in power consumption (116% YoY power growth rate). We can also include Alphabet's updated 2026 capex guidance of $180-$190 billion. Wait, are there any other notes or new notes we should write? We should write these as updates to the existing notes, because the guidelines say: "Your findings are a LIVING SET, not a fresh page... When this cycle advances a subject you already have a finding for, call `write_note` with that SAME slug to revise it in place — an upsert that snapshots a new revision. Mint a fresh slug ONLY for a genuinely new subject." Let's read the exact content of `nvidia-record-q1-results-and-trillion-dollar-order-book` and `big-tech-capex-reaches-710b-in-2026` first to make sure we integrate our updates perfectly without losing valuable prior history. Since we don't have their full text in the preview, we can read them. No, wait, `read_notes` only returned previews. Wait, is there a tool to read the full note body? Let's check. Ah, there is no direct "read_note_body" tool, but we can search or write the note with the same slug. Wait, does `read_notes` return the full text if we don't have a limit? No, it returns slug, title, preview, updated_at, and source_count. Let's check if we can query or if we can just write the notes. We can reconstruct the notes using the prior findings block in the prompt as a guide, and then expand them with our new findings! Let's check the "Relevant prior findings for this topic" block in the kickoff message: - "Nvidia's AI Capex Story Remains Intact Despite Kyber Rack Delays and Manufacturing Bottlenecks" (slug: `nvidia-record-q1-results-and-trillion-dollar-order-book`): "Nvidia's artificial intelligence capital expenditure story remains remarkably robust in mid-July 2026, even as the company's breakneck product release cadence collides with the physical limits of…" - "Big Tech's AI Capex Reaches Historic Scale as 2026 Infrastructure Sprint Approaches $725 Billion" (slug: `big-tech-capex-reaches-710b-in-2026`): "The capital expenditure boom supporting artificial intelligence infrastructure has reached unprecedented heights in mid-2026. The world's largest technology companies are expanding their cash outlays…" Let's search for "Kyber rack delay" or "Kyber" in our files to see if we have local memory or if we can find more details. Wait, we have a workspace at `/var/lib/hey-lefty/notes/...`. Let's check if there are markdown files in our workspace that contain the full body of the notes. Let's see: "Before modifying a file, read it first. Do not assume files or directories exist." Let's list files in the current directory or look for notes. Let's do a quick shell check or find. Oh, wait, we don't have a bash tool or file execution tool listed in our tools! Our tools 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! We only have these APIs. We do not have direct bash/file-system tools. That means all our interaction is through these API tools. Let's call `read_notes` again or check if we can get the full note body. Wait, `read_notes` does not have a parameter for full body, but let's check its description: "List the current set of notes for a research topic. Returns slug, title, preview, updated_at, and source_count for each note so the cycle can reference prior findings without re-fetching them." Yes, `read_notes` returns the list of notes. We can write a note and it will overwrite or update the existing note. We can write a comprehensive, updated note with the same slug. Let's check the first watch in `list_threads`: `id: "019f10c8-53dd-729c-926a-27e735d10910"` `title: "Alphabet Q2 2026 Earnings and Dilution Impact Tracking"` `fire_date: "2026-07-25T00:00:00Z"` Wait, has this fired? No, Alphabet's Q2 2026 earnings are confirmed for July 22, 2026. Today is July 20, 2026, so it has not fired yet. What about the watch: `id: "019ed2f6-f136-7051-a5b0-fa3b26ec02e0"` `title: "Nvidia Q2 FY2027 Earnings Release"` `fire_date: "2026-08-20T00:00:00Z"` It has not fired yet, as Nvidia's Q2 FY2027 earnings are expected in late August 2026. What about the watch: `id: "019f729f-b064-7b27-adec-815ee0193bf4"` `title: "AMD Advancing AI 2026 Keynote and Helios Platform Details"` `fire_date: "2026-07-23T00:00:00Z"` It has not fired yet (July 23 is in 3 days). Are there any watches that have fired? None of the listed watches have a fire date before July 20, 2026, except some might be ongoing. Let's review the list of watches: - `019f729f-b064-7b27-adec-815ee0193bf4` (AMD Advancing AI 2026 Keynote: July 23, 2026) - No - `019f6d77-03de-7a2b-bf6d-aacefbf423de` (Nvidia Kyber NVL144 rack commercial shipment: late 2027) - No - `019f15ef-5c71-7ebb-ae84-3e7bc3e90ac6` (Qualcomm Dragonfly C1000 commercial availability: 2028-01-15) - No - `019f15ef-4fa2-7f15-a299-a9eed3caa38d` (Qualcomm HBC Gen 1 and AI250 commercial sampling: 2027-06-15) - No - `019f10c8-53dd-729c-926a-27e735d10910` (Alphabet Q2 2026 Earnings and Dilution: July 25, 2026) - No - `019f0ba3-59a5-7471-b2e6-7c9feacff247` (Alphabet $40B ATM Stock Offering: ongoing) - No event reported yet - `019f0679-7d60-7d01-98b1-601ff5664523` (OpenAI Jalapeño chip initial production: late 2026) - No - `019ef70b-6b49-7831-893f-6d78e9b2099e` (Oracle $40B Debt/Equity Financing: ongoing) - No event reported yet - `019ee796-b599-775e-b0db-e799779d71a4` (Anthropic public S-1: late 2026) - No - `019edd49-1ccb-73d3-9fc0-80ae266f32c4` (Nebius Group Q2 2026 earnings: mid-August 2026) - No - `019ed2f6-f136-7051-a5b0-fa3b26ec02e0` (Nvidia Q2 FY2027 Earnings: late August 2026) - No - `019eb93f-9403-7dea-9c9e-9ac0765dd674` (Meta Platforms potential equity raise: one_shot) - No - `019eaeeb-7009-7e9b-9b2f-d6c2f4ec5339` (Nebius 850 MW active power capacity: end of 2026) - No - `019e9f7b-8898-71ef-a8fc-916318c1f6ed` (Vera Rubin platform broader commercial shipments: late 2026) - No - `019e9f7b-4f43-77fd-9c28-923bb95b9b39` (RTX Spark retail launch: Fall 2026) - No So none of the watches have fired yet. Let's write our updated findings. We will update: 1. `nvidia-record-q1-results-and-trillion-dollar-order-book` 2. `big-tech-capex-reaches-710b-in-2026` Wait, let's look at the guidelines for writing findings: - Title must be the first `# ` heading. - No YAML frontmatter. - Name and link the actual thing. - Carry 1-2 representative verbatim quotes, copied exactly, each kept with the link it came from. - Say what it means. - Use the stable slug to update. - External sources should stay normal markdown links in the `## Sources` block. - To link another finding, write its slug as a wikilink — `[[its-slug]]`. Let's draft the update to `nvidia-record-q1-results-and-trillion-dollar-order-book`. Wait, let's review the current title: "Nvidia's AI Capex Story Remains Intact Despite Kyber Rack Delays and Manufacturing Bottlenecks" Let's keep the title or adjust it slightly to encompass the new developments, but keep the slug: `nvidia-record-q1-results-and-trillion-dollar-order-book`. Let'

I will now call `write_note` to update our finding on Nvidia's AI capex story (`nvidia-record-q1-results-and-trillion-dollar-order-book`) with the latest mid-July 2026 data points. This includes TSMC's record Q2 2026 earnings and capex raise, as well as Michael Burry's prominent July 2026 critique of the structural contradiction in the AI trade. This update integrates public market data from our `/markets/NVDA/2026/07/20` snapshot.

Next, I will update our second relevant note `big-tech-capex-reaches-710b-in-2026` to reflect the upsized 2026 capital expenditure numbers. The latest TrendForce report from May 20, 2026, projects that the top 5 North American cloud service providers' combined capital expenditures will exceed **$770 billion in 2026** (a nearly 87% YoY increase), up from our prior estimate of $710B–$725B. It also highlights massive power grid challenges (an 18 GW addition in 2026) and the rapid scaling of AI inference computing power (+122% YoY). I'll write this note now.

Now that I have updated the findings with the latest data and critiques, I will call `finish_cycle` to complete this research cycle. All active threads are forward-looking (such as Alphabet's Q2 earnings on July 22 and AMD's event on July 23) and have not fired yet, so they remain open as registered watches. I will finish the cycle.

I have completed the research cycle for this topic. Here is a summary of the findings and updates we made: ### 1. Nvidia's AI Capex Story Remains Robust but Faces Structural Debates * **Segment Numbers**: In Q1 FY2027, Nvidia's Data Center segment generated a record **$75.25 billion** (up 92% YoY), with compute revenue at $60.4 billion and networking at a record $14.8 billion. * **Guidance & Commitments**: CEO Jensen Huang guided Q2 FY2027 revenue to **$91.0 billion** (+/- 2%), backed by **$119.0 billion in supply-related commitments**, showing booked demand extending well beyond the current quarter. * **TSMC Capex Hike**: In mid-July 2026, TSMC reported record Q2 revenue of $40.2 billion and raised its full-year 2026 capex guidance to a range of **$60 billion to $64 billion** (up from $52–$56 billion, representing a $4 billion minimum increase) while raising its 2026 revenue growth projection to slightly above 40%. * **The "Third Door" Contradiction**: Prominent investor Michael Burry (July 9, 2026, Substack post) warns of a core structural contradiction at the center of the AI trade: hyperscalers need the buildout to wrap up in 3-4 years to restore free cash flow (which is currently falling toward zero), whereas Nvidia needs permanent demand growth to support its valuation. He warns that resolving supply bottlenecks will ultimately erode Nvidia's scarcity premium and high gross margins. * **Rubin Delay Risks**: While Blackwell is expected to account for over 70% of Nvidia's high-end GPU shipments in 2026, TrendForce reports that the successor **Vera Rubin (VR200) series faces delay risks** due to HBM4 validation, CX9 networking transitions, and liquid cooling optimizations, lowering its projected 2026 shipment share from 29% to 22%. ### 2. Big Tech's AI Capex Upsized to $770 Billion * **Upsized Projections**: According to TrendForce, the combined capital expenditures of the top five North American cloud service providers (Google, Amazon, Microsoft, Meta, and Oracle) will exceed **US$770 billion in 2026** (an 87% YoY increase), up from prior estimates of $710B–$725B. * **Massive Inference Surge**: The nature of the spend is shifting massively toward commercialized workloads. Their combined AI inference computing power is projected to surge by **nearly 122% YoY** in 2026, compared to a 56% YoY increase in training power. * **Custom Silicon Growth**: Hyperscalers are scaling proprietary ASICs to bypass Nvidia's premium. Google's TPU demand is projected to grow by **80% YoY** in 2026 (transitioning to v8 in H2 2026), and Amazon's Trainium is expected to account for over **40% of its own AI server shipments**. * **Grid Bottlenecks**: The massive deployment of liquid-cooled racks is creating a severe power bottleneck. The top five North American CSPs' total server power consumption is projected to surge by **18 GW** in 2026, a **116% YoY growth rate** in power addition. The living documents on these topics have been updated with these stable slugs: * `[[nvidia-record-q1-results-and-trillion-dollar-order-book]]` * `[[big-tech-capex-reaches-710b-in-2026]]`