Cycle on Sep 10, 2026

I will start by listing the active threads and watches for this topic to see what questions are currently open or being monitored.

I will retrieve the current public-market snapshot for Nvidia (NVDA) to anchor our quantitative analysis in the latest financial performance data, revenue trajectories, and valuation multiples.

I will read the existing notes for the topic to see what has already been reported and what slugs we have.

The topic brief is "Is Nvidia's AI capex story still intact?" Let's look at the recent news on Nvidia and its AI capex. In the `market_view` for NVDA, we saw a news item: `[Neutral] 2026-09-09 — Microsoft (NASDAQ: MSFT) "Useful Yield" Push Puts NVIDIA's (NASDAQ: NVDA) AI Economics To The Test` This looks extremely relevant. It says: "Microsoft is advocating for a 'useful yield' approach to AI infrastructure, emphasizing efficiency per dollar and watt, which could significantly impact NVIDIA's AI economics." Let's do a web search on: 1. Microsoft's "Useful Yield" push and its implications for Nvidia's AI capex. 2. The general state of Nvidia's AI capex story, hyperscaler capex updates, and any recent earnings/reports in late August or early September 2026. 3. Let's see if there are other companies or analyst reports discussing Nvidia's capex trajectory. Let's run a web search.

Let's fetch `https://finance.yahoo.com/technology/ai/articles/microsoft-useful-yield-test-raises-225832059.html` and `https://convergedigest.com/microsoft-useful-yield-ai-infrastructure/` to get deep details on this. We want to see how Microsoft defines "useful yield" and how it ties into Nvidia's AI capex story. First, let's fetch the yahoo finance article.

Let's search for Microsoft's September 2, 2026 update regarding "useful yield" or Rani Borkar's speech at SEMICON Taiwan. This will give us the direct context of Microsoft's framework. Let's do a search.

Let's fetch the Microsoft Asia blog post: `https://news.microsoft.com/source/asia/2026/09/02/the-new-imperative-for-ai-infrastructure-useful-yield/` to get the primary source quotes and specific details. This is an official source. Let's fetch it.

This is extremely interesting. Rani Borkar argues that the defining measure of progress is "useful yield" (tokens per dollar and watt, higher throughput, lower latency, etc.). Let's search for "Nvidia Q2 FY2027 earnings" or "Nvidia August 2026 earnings" to see exactly what Nvidia reported on August 26, 2026. This will show us if the AI capex story is intact and what the numbers were. Let's do a web search.

Let's analyze the details from Nvidia's Q2 FY2027 earnings report on August 26, 2026. First, some key parameters from the report: - **Revenue**: $96.22 billion (up 106% YoY, up 18% QoQ). This beat Nvidia's own guidance of $91.0 billion (+/- 2%) and the consensus estimate of $92.07 billion. - **Data Center Revenue**: $89.02 billion (up 117% YoY, up 18% QoQ), representing 92% of total revenue. - **Segment breakdown within Data Center**: - **Hyperscale**: $48.71 billion (up 102% YoY, up 13% QoQ). - **AI Clouds, Industrial and Enterprise (ACIE)**: $40.31 billion (up 138% YoY, up 25% QoQ). - **Vera Rubin platform**: - Expected to represent about 20% of Data Center revenues in Q3 FY2027 (guided to $108 billion +/- 2% total revenue). - Entering full production. - **FY2028 Revenue Guidance**: - Colette Kress (CFO) and Jensen Huang guided to **~70% YoY revenue growth for the full fiscal year 2028**, far above analyst expectations of 44%. - "Customer forecasts point to our growth doubling next year, but guidance reflects supply constraints." - **Gross Margin**: - Q2 FY2027 gross margin was 75.0% (GAAP and non-GAAP). - Q3 FY2027 guided to 74.0% (+/- 50 bps). - Expected to bottom out in Q4 FY2027 in the 71% to 72% range, partially due to memory prices. - **Accounts Receivable / Payment Terms**: - Accounts receivable reached $63.1 billion. - Days sales outstanding (DSO) rose to 60 days from 45 days sequentially due to extended payment terms on large, multi-quarter agreements with certain investment-grade customers. - **Supply and Capacity Commitments**: - Climbed to $279 billion from $119 billion in the prior quarter, primarily reflecting memory procurement needed to support demand over the next several years. - **Guarantees**: - Maximum gross exposure from land, power, and shell guarantees, including the SB Energy guarantees signed in August 2026, totaled $108.5 billion. Let's double check if there are other updates on the "useful yield" push or other aspects of the capex cycle. Let's see: we have some watches that we should check if they fired. Let's review the list of active watches: 1. `Nvidia Full Year Fiscal 2028 Revenue Growth Track` (guided to ~70% yoy). Fire date: 2028-02-25. Ongoing watch, doesn't fire now. 2. `Nvidia Q3 FY2027 Earnings - Vera Rubin 20% Revenue Share Track`. Guided to ~20% of total data center revenue in Q3 FY2027. Fire date: 2026-11-18. Doesn't fire now (it's currently Sep 10, 2026). 3. `Nvidia Exercise of Nebius Group (NBIS) Warrants Track`. "Track whether Nvidia exercises its pre-funded warrants in Nebius Group (NBIS) on or after September 11, 2026, which would convert its passive beneficial ownership of 21 million warrant-backed shares into active common shares." Fire date: 2026-09-12. Today is September 10, 2026, so this restriction is lifted tomorrow. We should check if there are any announcements or updates on this. 4. `Nvidia Gross Margin Track` (`gross_margin < 70`). Gross margin was reported as 75.0% for Q2, and guided to 74.0% for Q3, and 71-72% for Q4. So it has not dropped below 70% yet. 5. `Nvidia-Wall Street $500B Compute Financing Platforms Final Execution Track`. Let's search if there are any updates on definitive agreements or any first closed project tranches. 6. `Apple Revamped Siri Launch with Blackwell Workloads`. Fire date: 2026-09-15. Today is Sep 10, 2026. Let's search if there are any early releases or launches or details on Apple's Siri launch. 7. `CME Group Launch of Regulated Compute Futures Contracts`. Fire date: 2026-10-05. Let's run a web search to check: - "Nebius Group" "Nvidia" warrants. - "Nvidia" "warrant" "Nebius" "September 11" OR "September 2026". - "Nvidia" "$500B" OR "500 billion" "compute financing" OR " definitive agreement" OR "closed project". - Apple Siri Blackwell 2026. Let's do a web search.

Let's fetch `https://finance.yahoo.com/technology/ai/articles/nvidia-reports-9-3-beneficial-134838540.html` and `https://seekingalpha.com/news/4615642-nebius-rises-as-nvidia-discloses-9_3-percent-stake` to see if there is any new filing or event recorded about the warrants. Let's fetch the yahoo finance article.

The September 11, 2026 date is tomorrow (and today is September 10, 2026). Let's search if there has been any announcement today or recently about Nvidia exercising its warrants. Let's do a search.

Since today is September 10, 2026, the lockup is set to expire tomorrow on September 11. There are no filings or announcements yet indicating that Nvidia has *already* exercised or sold (which would be prohibited prior to tomorrow anyway). This means the watch `Nvidia Exercise of Nebius Group (NBIS) Warrants Track` is still active and has not fired yet, but is extremely close. I should leave it open or check if there is any other watch that has fired. Let's check the watch: `Nvidia-Wall Street $500B Compute Financing Platforms Final Execution Track` "Track when Nvidia and its Wall Street partners (Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR) execute the definitive agreements for the $500 billion compute financing platforms, following the MOUs signed on August 10, 2026." Let's do a search on this. Is there any news about definitive agreements or first project closed under this $500B platform? Let's search.

Let's see if there are any updates on: - Virginia, Georgia, Ohio, Texas or other states imposing moratoriums or power/grid regulations on large AI data centers (State-Level AI Data Center Moratorium Expansion Track). - Progress on the Ratepayer Protection Act. - SpaceX launch of first Starmind AI1 prototype satellite. - Helix Digital Infrastructure project announcements. Let's do a search on "Ratepayer Protection Act" and "AI data center power Virginia Georgia Ohio Texas". Let's search.

Let's look at the "Oklahoma HB 2992, the Data Center Consumer Ratepayer Protection Act" which was signed by Oklahoma's Governor Stitt in 2026. Wait, let's search for "Ratepayer Protection Act" in the US Congress or federally. The power.com article says: "House Committee Sends Ratepayer Protection Act to Floor 52-0" "The Ratepayer Protection Act will ensure that data centers pay their own way, instead of passing costs onto hardworking families,” Chairman Brett Guthrie (R-Ky.) said. Let's see if we can find more about this federal bill "Ratepayer Protection Act". Let's search.

Let's analyze the Ratepayer Protection Act (H.R. 9340): - It was introduced on June 18, 2026. - It was advanced by the House Energy and Commerce Committee on July 21, 2026, by a vote of 52-0. - It would amend Section 111(d) of the Public Utility Regulatory Policies Act (PURPA) to require state public utility commissions to consider establishing a "large-load standard". Under this standard, any rate charged or agreement entered into by an electric utility for providing service to a large-load customer (defined as a non-residential customer with a peak demand of **100 megawatts or more**) must recover the full, incremental cost of any generation, transmission, or distribution upgrades necessary to serve them, and provide financial assurances. - This is a very significant piece of legislation for the AI data center buildout. Let's check our active watches: `U.S. Congress Progress on Ratepayer Protection Act` (Track whether the Ratepayer Protection Act passes the U.S. House of Representatives or Senate, imposing federal regulations on AI data center power allocation and ratepayer protections.) It has advanced from the committee (52-0 vote) but has not yet passed the full House or Senate. So this watch is still active and ongoing, but we have a major update on it. Let's check if there are other watches that have fired or have updates. What about: `Apple Revamped Siri Launch with Blackwell Workloads` "Track Apple's landmark September 2026 launch of its revamped Siri, which relies heavily on external Nvidia Blackwell cloud infrastructure." Fire date is September 15, 2026. Today is September 10, 2026, so it hasn't fired yet. What about: `Nvidia standalone Vera CPU revenue ramp` (Ongoing, not firing yet). `TSMC Monthly Revenue YoY Growth Rate Track` (Ongoing, let's see if we can check TSMC's latest monthly revenue for August 2026). Let's search for "TSMC August 2026 revenue". This will be very helpful to see if the physical semiconductor supply chain is catching up or if demand is softening. Let's search.

Let's fetch `https://investor.tsmc.com/english/monthly-revenue/2026` to see what TSMC's August 2026 revenue was. Or let's search specifically for the TSMC August 2026 revenue report. Let's do a search.

Wow, TSMC's August 2026 revenue report was released TODAY, September 10, 2026! Let's see: - Consolidated revenue for August 2026 was **NT$514.81 billion**. - This is an increase of **10.1% from July 2026** and **an increase of 53.3% from August 2025**. - Revenue for January through August 2026 totaled NT$3,386.87 billion, an increase of **39.3% compared to the same period in 2025**. Let's check our watch: `TSMC Monthly Revenue YoY Growth Rate Track` "Track TSMC's monthly revenue year-over-year growth rate. A drop below 20% YoY would signal that the physical semiconductor supply chain is finally catching up with demand or that hyperscaler chip orders are starting to soften." The YoY growth rate for August 2026 was **53.3%**, which is well above 20%. This shows that demand and physical semiconductor supply remain extremely strong, and the AI capex cycle is very robust! So this watch remains open (ongoing) and has not tripped (it would trip if it dropped below 20%). Let's write a finding about this. The slug should be related to the core theme: Nvidia's AI capex story and its sustainability. Wait, we have: `nvidia-ai-capex-cycle-sustainability` (Nvidia's AI Capex Cycle Sustainability: Historic Q2 FY2027 Performance and the $4.1 Trillion Hyperscaler Spend Horizon) This is the perfect slug to update! Let's check the existing content of `nvidia-ai-capex-cycle-sustainability` by reading it. Actually, I can write a note with the same slug `nvidia-ai-capex-cycle-sustainability` to revise it in place. Let's see if there are other slugs we want to update. We have: - `hyperscaler-capex-surge-2026` (The AI Infrastructure Clash: Texas Grid Freeze and the Nationwide Explosion of Data Center Moratoriums) - `enterprise-ai-roi-gap-revenue-reckoning` (The Hyperscaler Capital Crossover: Accelerating Capex and the Shift to Physical Bottlenecks) Let's look at the topic brief: "Is Nvidia's AI capex story still intact?" This is a high-level question about the overall sustainability of Nvidia's AI capex story, and we have a wealth of fresh, concrete evidence from: 1. Nvidia's blowout Q2 FY2027 earnings report on August 26, 2026 ($96.22B revenue, up 106% YoY; Data Center $89.02B, up 117% YoY; guided FY2028 revenue growth to ~70%, far exceeding expectations). 2. TSMC's record-breaking August 2026 revenue report released today, September 10, 2026, showing 53.3% YoY growth to NT$514.81 billion. 3. Microsoft's new "Useful Yield" paradigm introduced by Rani Borkar at SEMICON Taiwan on September 2, 2026, which shifts the infrastructure focus from raw component capacity to system-level productivity (tokens/dollar and watt), raising the economic stakes for Nvidia's AI hardware economics. 4. The bipartisan "Ratepayer Protection Act" (H.R. 9340) advancing 52-0 in the House Energy and Commerce Committee to force data centers (100 MW+) to pay the full incremental costs of grid upgrades, adding regulatory and utility cost pressure to the capex model. Let's write two highly detailed, evidence-grounded findings to update our living documents: 1. **Finding 1 (slug: `nvidia-ai-capex-cycle-sustainability`):** Focus on Nvidia's historic Q2 FY2027 results, its massive FY2028 guidance, TSMC's blowout August revenue, and how these figures show the capex story remains completely intact from a physical supply and demand perspective. 2. **Finding 2 (slug: `enterprise-ai-roi-gap-revenue-reckoning` or a new one like `microsoft-useful-yield-ai-infrastructure-economics`):** Focus on the economic and efficiency shift, specifically Microsoft's "Useful Yield" push at SEMICON Taiwan and the legislative pressure from the federal Ratepayer Protection Act (H.R. 9340) which are putting the long-term economics of the capex cycle to the test. Wait, let's look at the existing notes. We have `enterprise-ai-roi-gap-revenue-reckoning` which tracks the capital crossover and ROI gap. This is a great place to update or we can use a new slug if it's a completely new theme. The prompt says: "Name a slug for the DURABLE SUBJECT it tracks: the throughline, entity, or running debate... NEVER the transient story-of-the-day... When this cycle advances a subject you already have a finding for, call write_note with that SAME slug to revise it in place." Let's check the current content of `nvidia-ai-capex-cycle-sustainability` and `enterprise-ai-roi-gap-revenue-reckoning`. We don't have their full bodies, but we can write updated notes that incorporate the latest developments. Wait, let's fetch more information on how Microsoft's "Useful Yield" relates to Nvidia. Rani Borkar defined useful yield as: `Capability × Deployment Velocity × Utilization = Useful Yield` This is a system-level measure. Instead of just buying more GPUs and adding megawatts, cloud providers must get more useful intelligence from every byte they already have. This is a competitive test for Nvidia because: - Large buyers are designing custom silicon (like Microsoft Azure Maia and Azure Cobalt) to optimize hardware around their workloads, which could limit the economics available to third-party suppliers like Nvidia. - However, if efficiency gains make more AI applications commercially viable, it could actually drive *more* deployment of Nvidia hardware. - Meanwhile, Nvidia's CEO Jensen Huang argued on the Q2 call that "compute is revenue" and "demand is accelerating... today we have a golden age of new AI labs and startups, multiple frontier labs scaling in parallel, a thriving open-model ecosystem and physical AI coming online." Let's also look at Nvidia's balance sheet changes in Q2 FY2027: - **Supply and capacity commitments** skyrocketed to **$279 billion** (from $119 billion in the prior quarter), showing Nvidia is aggressively locking in memory and wafer supply for the next several years. - **Maximum gross exposure from guarantees** (land, power, shell) reached **$108.5 billion**. - **Days Sales Outstanding (DSO)** rose to 60 days from 45 days due to extended payment terms on large, multi-quarter contracts with investment-grade customers. - **Accounts receivable** reached **$63.1 billion**. This is incredibly rich, concrete data. Let's write the first note. Wait, I should make sure I have all the quotes correct. Let's look at the quotes from Nvidia's Q2 FY2027 press release: Jensen Huang: > "AI has reached its inflection point. It’s doing useful work. Its tokens are productive and profitable. Now, compute is revenue." > "And demand is accelerating. This time last year, one lab alone was driving the buildout; today, we have a golden age of new AI labs and startups, multiple frontier labs scaling in parallel, a thriving open-model ecosystem and physical AI coming online — with strong momentum across the U.S. and around the world. The AI infrastructure buildout is at full steam." Let's look at the quotes from TSMC's August revenue report: > "TSMC today announced its net revenue for August 2026: On a consolidated basis, revenue for August 2026 was approximately NT$514.81 billion, an increase of 10.1 percent from July 2026 and an increase of 53.3 percent from August 2025. Revenue for January through August 2026 totaled NT$3,386.87 billion, an increase of 39.3 percent compared to the same period in 2025." Let's look at the quotes from Microsoft's "Useful Yield" report: Rani Borkar: > "We will continue to build capacity because the world will need it. But our defining measure of progress must be what comes out: not only chips or tokens, but useful intelligence translated into empowerment, opportunity and human achievement" > Borkar urged the industry to focus on the "yield imperative": "defining success in AI infrastructure not only on how much capacity it can build, but on how efficiently those resources can be converted into useful intelligence." Let's look at the quotes from the Ratepayer Protection Act (H.R. 9340): Chairman Brett Guthrie: > "The Ratepayer Protection Act will ensure that data centers pay their own way, instead of passing costs onto hardworking families." Chairman Bob Latta: > "This legislation protects American ratepayers from being forced to cover the cost of energy infrastructure upgrades needed to support data centers." This is extremely powerful evidence. Let's first write `nvidia-ai-capex-cycle-sustainability` to incorporate the blowout Q2 FY2027 earnings and TSMC's blowout August revenue, showing the physical demand and supply remain fully intact. Wait, let's double check if there are any other notes we should write. Let's write a note on `enterprise-ai-roi-gap-revenue-reckoning` focusing on the shifting economic and regulatory pressures: Microsoft's "Useful Yield" and the "Ratepayer Protection Act" (H.R. 9340). This is a perfect match for that slug, as it directly deals with the transition from raw capex buildout to ROI, efficiency, and regulatory utility bottlenecks. Let's write the first note. Remember, no YAML frontmatter, the title is the first `# ` heading, and we must carry 1-2 representative verbatim quotes with their links. We will cite the `/markets/NVDA/2026/09/10` URL for the public markets data. Let's do it.

Now, let's write the second note focusing on the shifting economics, efficiency, and regulatory utility pressures. The slug is `enterprise-ai-roi-gap-revenue-reckoning` (The Hyperscaler Capital Crossover: Accelerating Capex and the Shift to Physical Bottlenecks). This note will capture: 1. Microsoft's "Useful Yield" framework introduced at SEMICON Taiwan on September 2, 2026. This represents a paradigm shift where the metric of success moves from raw capacity (GPUs, fibers, megawatts) to system productivity (tokens per dollar and watt, throughput, utilization, and latency). 2. The legislative and utility cost pressures, specifically the bipartisan "Ratepayer Protection Act" (H.R. 9340) which passed the House Energy and Commerce Committee 52-0 on July 21, 2026. This act forces data centers with peak demands of 100 MW or more to pay the full, incremental cost of grid upgrades, shifting the capital burden from residential ratepayers to hyperscalers and data center builders. 3. How this pressure forces a "revenue reckoning" where the massive capital expenditure must justify itself through optimized efficiency, custom silicon designs (like Maia and Cobalt), and profitable customer workloads rather than just installing more hardware. Let's write this note. We will cite the sources we fetched, including the Brookings article, the House Energy and Commerce Committee press release, and the Microsoft Asia blog post. Let's do it.

This is excellent. We have written two detailed notes to update our living documents. Let's review our threads and watches to see if we can resolve any threads or record any events. First, we have an open thread or watch: - `U.S. Congress Progress on Ratepayer Protection Act` (Track whether the Ratepayer Protection Act passes the U.S. House of Representatives or Senate, imposing federal regulations on AI data center power allocation and ratepayer protections.) Wait, this is a watch. Let's see: we have recorded that the Ratepayer Protection Act (H.R. 9340) advanced from the House Energy and Commerce Committee 52-0 on July 21, 2026. This is a major progress event, but the watch is "Track whether the Ratepayer Protection Act passes the U.S. House of Representatives or Senate". It has not yet passed either chamber, so it is not fully resolved. We can record an event on it, but set `resolves: false` to keep it open. Wait, let's look at the active threads to find the ID of `U.S. Congress Progress on Ratepayer Protection Act`. From `list_threads`: `id: "019fe5ce-c587-76c5-bf31-fdbfb073f5e8"`, `kind: "watch"`, `title: "U.S. Congress Progress on Ratepayer Protection Act"`. Let's call `record_event` on this thread to log the 52-0 committee vote and the White House pledge expansion, while keeping it open. Let's do it.

Let's look at another watch: - `TSMC Monthly Revenue YoY Growth Rate Track` `id: "019ecfb3-79b3-76db-aeec-0bc8d0635c41"`, `kind: "watch"`, `title: "TSMC Monthly Revenue YoY Growth Rate Track"`. The trigger is "Fires if TSMC's monthly revenue year-over-year growth drops below 20%." We checked TSMC's August 2026 revenue today, and it grew **53.3% YoY** (to NT$514.81 billion), which is far above the 20% floor. So it has not fired. It is an ongoing watch, so we can record an event showing it remains extremely healthy, but set `resolves: false`. Let's do it.

Let's check if there are other watches we can update. What about: - `Nvidia Gross Margin Track` `id: "019ec040-d231-7dc2-9387-902ca1d82c1d"`, `kind: "watch"`, `title: "Nvidia Gross Margin Track"`, `trigger: "NVDA gross_margin < 70"`. In Q2 FY2027, Nvidia reported a 75.0% gross margin, and guided to 74.0% for Q3, and 71-72% for Q4. So it has not dropped below 70%. We can record an event showing it is still holding strong at 75%, and set `resolves: false`. Let's do it.

Let's check if there are other watches we can update. What about: - `Nvidia Full Year Fiscal 2028 Revenue Growth Track` `id: "01a0427b-e23b-7d01-921a-d6dbb06a1a5c"`, `kind: "watch"`, `title: "Nvidia Full Year Fiscal 2028 Revenue Growth Track"`, `trigger: "NVDA revenue_growth_yoy >= 70"`. Wait, in the Q2 FY2027 call, Colette Kress explicitly stated that Nvidia expects **fiscal 2028 revenue growth of approximately 70%** (some sources say "growth doubling" but the official guidance is ~70%). The trigger is `NVDA revenue_growth_yoy >= 70`. This refers to the actual reported full year growth, which won't be reported until early 2028 (February 2028 is the fire date). So we should keep this open, but we can record an event that the company itself has now officially guided to exactly this 70% level, making the watch highly salient. Let's do it.

Let's check if there are other watches. What about: - `Nvidia Exercise of Nebius Group (NBIS) Warrants Track` `id: "019f9366-f32f-75c4-a457-59c8368eb7d0"`, `kind: "watch"`, `title: "Nvidia Exercise of Nebius Group (NBIS) Warrants Track"`, `trigger: "Fires when regulatory filings or company announcements confirm whether Nvidia has exercised its Nebius warrants on or after the September 11, 2026 restriction lift."`. The restriction lift is tomorrow, September 11, 2026. Today is September 10, 2026, so the restriction has not been lifted yet, and no exercise has occurred. We should keep this open. Let's register a new watch if we have surfaced anything forward-looking. Wait, what did we surface? - Nvidia expects its gross margins to **bottom out in the fourth quarter of fiscal 2027 in the range of 71% to 72%**, partially due to memory prices. This is a very specific, dated catalyst / threshold. Let's register an ongoing or one_shot watch for this. Wait, let's call it `Nvidia Q4 FY2027 Gross Margin Bottoming Track`. Let's see if we should register it. Yes, we can register a watch for Nvidia's Q4 FY2027 gross margin bottoming out. Let's do it. - `topic_id`: `019e95a0-30d1-7d08-ae5b-10ac242b9d5c` - `cycle_id`: `eb1f4135-f74e-455a-8677-9650a08094b6` - `title`: `Nvidia Q4 FY2027 Gross Margin Bottoming Track` - `cardinality`: `one_shot` - `watch_type`: `metric_threshold` - `metric`: `gross_margin` - `op`: `<=` - `threshold`: 72 - `fire_date`: `2027-02-25T00:00:00Z` (approximate date for Q4 FY2027 earnings in late Feb 2027) - `description`: `Track whether Nvidia's gross margin bottoms out in the 71-72% range in Q4 FY2027 as guided, or drops further due to rising memory and packaging costs.` - `note`: `Fires when Q4 FY2027 earnings are reported, confirming if gross margin remains above or below the guided 71-72% bottom.` Let's call `record_watch`.

This is excellent. We have updated two major notes, recorded events on four active watches, and registered a new watch. Let's call `finish_cycle` to complete this research cycle. Let's do it.

I have completed the research cycle for this topic. Here is a summary of the key findings and updates: ### 1. Physical Supply and Demand Remain Fully Intact Nvidia's AI capex story remains exceptionally robust, supported by blowout financial results and solid supply chain indicators: * **Blowout Q2 FY2027 Results:** On August 26, 2026, Nvidia reported **$96.22 billion in revenue** (up 106% YoY), beating its own guidance by 5.7% and consensus by 4.5%. Data Center revenue reached **$89.02 billion** (up 117% YoY), accounting for 92% of the total. * **Stunning FY2028 Guidance:** Nvidia officially guided to **~70% year-over-year revenue growth for the full fiscal year 2028**, far exceeding Wall Street's expectations of 44%. CEO Jensen Huang declared that *"compute is revenue"* as tokens become highly productive and profitable. * **Aggressive Supply Commitments:** Nvidia's **supply and capacity commitments skyrocketed to $279 billion** (up from $119 billion in the prior quarter) to lock in long-term high-bandwidth memory (HBM) and foundry capacity. * **TSMC's August Record:** Released on September 10, 2026, TSMC's consolidated revenue for August 2026 surged **53.3% YoY to NT$514.81 billion** (up 10.1% sequentially from July), demonstrating that physical chip production is running at maximum capacity with zero signs of demand softening. This has been updated under the stable slug: `[[nvidia-ai-capex-cycle-sustainability]]`. ### 2. Shifting Economics: The Move to "Useful Yield" and Regulatory Grid Pressures While physical demand is intact, the long-term economics of the capex cycle are facing a transition from raw capacity to system efficiency and grid cost accountability: * **Microsoft's "Useful Yield" Imperative:** At SEMICON Taiwan on September 2, 2026, Microsoft's Azure hardware chief Rani Borkar introduced the "yield imperative." Microsoft is shifting its metric of success from raw hardware capacity (GPUs and megawatts) to **Useful Yield** ($\text{Capability} \times \text{Deployment Velocity} \times \text{Utilization}$). By using "cross-layer design" and custom silicon (Azure Maia/Cobalt), hyperscalers aim to maximize tokens per dollar and watt, which could eventually limit the economics and pricing power available to third-party hardware suppliers like Nvidia. * **Federal Ratepayer Protection Act (H.R. 9340):** On July 21, 2026, the House Energy and Commerce Committee voted **52-0** to advance H.R. 9340 to the House floor. The bill requires state public utility commissions to establish a "large-load standard" where data centers with a peak demand of **100 MW or more** must pay the **full, incremental cost of any grid upgrades** needed to serve them. This significantly increases the upfront capital requirements for physical data center construction. This has been updated under the stable slug: `[[enterprise-ai-roi-gap-revenue-reckoning]]`. ### 3. Key Telemetry Updates * **TSMC Growth:** Recorded that TSMC's August YoY growth of 53.3% remains well above the 20% alert threshold on the ongoing `[[tsmc-monthly-revenue-yoy-growth-rate-track]]` watch. * **Nvidia Gross Margin:** Recorded that Nvidia's Q2 gross margin of 75.0% remains strong, and registered a new watch (`[[nvidia-q4-fy2027-gross-margin-bottoming-track]]`) to monitor the guided gross margin bottom of 71-72% in Q4 FY2027. * **Ratepayer Protection Act:** Logged the 52-0 committee advancement of H.R. 9340 under the `[[u-s-congress-progress-on-ratepayer-protection-act]]` watch.