> ## Content Index > Fetch the complete content index at: https://tokyobrainie.ghost.io/llms.txt > Use this file to discover other available public pages before exploring further. # NVIDIA、1GWで400億ドルは本当か——Vera Rubinの売上機会、PER25.9倍と日本への波及を検証｜Opportunity Is Not Cash - URL: https://tokyobrainie.ghost.io/news-cutting-room-411bc361-d73b7e1d-en/ - Published: 2026-11-07T03:00:00.000Z - Updated: 2026-11-07T03:00:00.000Z - Description: NVIDIA’s USD 40 Billion per Gigawatt: Opportunity Is Not Cash Collected TokyoBrainie · News Analysis · 7 November 2026 JST. The valuation claims refer to the U.S. cl - Author: TokyoBrainie - Tags: News Cutting Room, Signal Room, English, #Import 2026-11-11 08:54 [English](https://tokyobrainie.ghost.io/news-cutting-room-411bc361-d73b7e1d-en/) · [简体中文](https://tokyobrainie.ghost.io/news-cutting-room-411bc361-d73b7e1d-zh-hans/) · [日本語](https://tokyobrainie.ghost.io/news-cutting-room-411bc361-d73b7e1d-ja/) **NVIDIA’s USD 40 Billion per Gigawatt: Opportunity Is Not Cash Collected** TokyoBrainie · News Analysis · 7 November 2026 JST. The valuation claims refer to the U.S. close on November 3, not an updated November 6 valuation. The casino image is fictional satire; AWS tokens do not depict actual payment or settlement. ### Three findings The USD 40 billion/GW figure is management’s opportunity estimate, not backlog, recurring annual revenue, profit or collected cash. Trailing revenue growth of 83.4% can be reproduced from filings. A P/E of 25.9 cannot uniquely reveal the growth rate investors expect. Japan has concrete deployment and development connections, but announcements still need to become paid contracts, operating assets and economic returns. ### 1\. Audit the status of each number The starting source is [Forbes Japan](https://forbesjapan.com/articles/detail/107752?ref=tokyobrainie.ghost.io). This analysis checks the article’s figures and claims against corporate disclosures. | Claim | Finding | | ------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------- | | FY2027 Q2 revenue USD 96.2bn, up 105.9% | USD 96,221m divided by USD 46,743m minus one is approximately 105.85%, rounding to 105.9% | | Data Center revenue about USD 89bn, up 18% sequentially | Consistent with reported results; a market-platform classification, not the same thing as a reportable segment | | FY2028 growth about 70%; Rubin about 20% of Q3 Data Center revenue | Management forecasts, not realized results. FY2028 roughly spans February 2027–January 2028 | | Q3 guidance USD 108bn ±2% | A forecast range of USD 105.84bn–110.16bn | | Rubin USD 40bn/GW versus Blackwell USD 25bn/GW | Confirmed as management’s estimate on transcript p.4; detailed independent reconstruction remains unavailable | | Trailing revenue growth 83.4% | Reproduced below; revenue growth is not EPS growth | | Drawdown 13.6%, one-year return 2.3% versus S&P 500 11.9%, P/E 25.9 versus 22.8 | Not independently reproduced in this investigation; retained only as source-article claims, not evidence of cheapness | | FY2026 Compute & Networking share 89.6% | Primary-source recalculation unfinished; do not substitute it for Data Center’s revenue share | [Results](https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-Financial-Results-for-Second-Quarter-Fiscal-2027/?ref=tokyobrainie.ghost.io) / [Call transcript](https://investor.nvidia.com/files/content%5Ffiles/TRANSCRIPT%5F-NVIDIA-Corp-NVDA-US-Q2-2027-Earnings-Call-26-August-2026-5%5F00-PM-ET.pdf?ref=tokyobrainie.ghost.io) ### 2\. What does USD 40 billion/GW represent? Vera Rubin is a platform spanning CPU, GPU and networking, with complementary Groq LPU inference racks. Every installation need not buy the same product mix. The commercial logic is broader NVIDIA content per project, not simply a higher GPU price. USD 25bn to USD 40bn is a 60% increase in estimated opportunity—not automatically a 60% increase in like-for-like prices or profit. [Platform announcement](https://nvidianews.nvidia.com/news/nvidia-vera-rubin-platform?ref=tokyobrainie.ghost.io) | Illustration | Arithmetic and limitation | | -------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------- | | 1GW = 1 million kW | USD 40,000/kW for Rubin versus USD 25,000/kW for Blackwell; installation-linked opportunity, not an annual electricity charge | | Hypothetical 0.5GW | USD 20bn, assuming the reference mix and pricing; not a forecast | | Hypothetical 1GW delivered evenly over three years | Roughly USD 13.3bn annually; actual recognition depends on delivery, acceptance and performance obligations | | Hypothetical facility power 1GW and PUE 1.2 | IT load about 0.833GW; applicable only if compatible with management’s denominator | PUE is total facility power divided by IT power. Utility capacity and IT load are not interchangeable. GW measures power; GWh measures energy. A constant 1GW for a year uses 8.76TWh, but multiplying that energy figure by USD 40bn would be dimensionally wrong. The article’s yen equivalents assume JPY 158/USD: USD 40bn becomes JPY 6.32tn and USD 25bn becomes JPY 3.95tn. This is a conversion assumption, not a verified current FX rate. The public explanation does not supply a sufficiently detailed bill of quantities, realized prices, configuration and power denominator to reproduce USD 40bn independently. Its strategic logic is plausible; it is not a verified order book. Applying it to all global data-center construction would ignore custom chips, competitors, upgrades, discounts, product adoption and timing. ### 3\. NVIDIA’s revenue engines | Business mechanism | Products and caveat | | -------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------ | | Compute hardware | GPUs, CPUs, boards and systems such as DGX, sold directly or through partners. Customer total capex is not all NVIDIA revenue | | Networking and data movement | NVLink, InfiniBand, Spectrum Ethernet, ConnectX and BlueField extend content beyond the GPU | | Paid software and support | AI Enterprise, operations software and vGPU use contractual subscription or consumption models | | Development platforms and models | CUDA-related SDKs, NIM, Nemotron, Omniverse and Isaac combine differing free and paid rights. Not every CUDA execution generates a toll | | Edge and industrial applications | GeForce, RTX PRO, Jetson and DRIVE serve different applications; applications are not accounting divisions. GeForce NOW is also a service offering | | Cloud distribution | DGX Cloud/Lepton provide routes to compute; partners offer NVIDIA technology. AWS GPU-hour receipts cannot simply be booked as NVIDIA GPU-hour revenue | Depending on the contract, software fees may be based on the number of GPUs, the number of users or actual usage. Model rights and NIM runtime rights are distinct. [Licensing FAQ](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-software-licensing-overview-and-faq/?ref=tokyobrainie.ghost.io) Compute & Networking and Graphics are reportable segments. FY2027 market-platform reporting changed to Data Center and Edge Computing; Data Center includes Hyperscale and ACIE. Use recast comparisons, rather than adding legacy Gaming or Automotive categories to the new totals. Software-only profitability cannot be inferred from a mixed hardware/software segment. [10-Q, Note 13](https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm?ref=tokyobrainie.ghost.io) ### 4\. From purchase order to operating capacity Amazon corroborates the additional two-million-GPU plan, but the schedules differ. | Document, both dated August 26 | Schedule | | ------------------------------ | ------------------------------------------------------------------- | | NVIDIA earnings call | FY2027 Q3 through FY2029 Q2, approximately August 2026–July 2028 | | Amazon joint release | Calendar 2027–2028; includes Blackwell Ultra, Rubin and Rubin Ultra | The quantity is supported as a plan; exact starting and completion months remain unreconciled. These are not two million Rubin GPUs alone. Neither GPU count nor GPU power alone establishes GW-level revenue. CPU, networking, cooling, configuration and pricing matter. [Amazon release](https://press.aboutamazon.com/aws/2026/8/aws-and-nvidia-to-deliver-2-million-additional-gpus-and-next-generation-infrastructure-for-agentic-and-physical-ai?ref=tokyobrainie.ghost.io) Execution requires land and permits, deliverable power and grid access, manufacturing/HBM/packaging, cooling and connectivity, commissioning, financing and paying AI customers. Chip shipments can precede productive operation and customer payback. Efficiency can lower unit costs and expand demand; it can also reduce the hardware needed for a fixed workload or encourage custom-chip substitution. Performance gains cannot be translated directly into sales growth. Management’s 71–72% gross-margin trough and subsequent 72–73% expectation remain forecasts requiring later confirmation. Supply constraints do not prove customer solvency or ROI. ### 5\. Reproducing 83.4%, without inventing a valuation conclusion All revenue inputs below are USD millions. | Calculation | Result | | ----------------------------------------------------------------- | --------------------------- | | Latest TTM: FY2026 215,938 − FY2026 H1 90,805 + FY2027 H1 177,837 | 302,970 | | Prior TTM: FY2025 130,497 − FY2025 H1 56,084 + FY2026 H1 90,805 | 165,218 | | 302,970 / 165,218 − 1 | 83.3759%, rounding to 83.4% | [Annual results](https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-Financial-Results-for-Fourth-Quarter-and-Fiscal-2026/?ref=tokyobrainie.ghost.io) / [Prior interim comparison](https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Announces-Financial-Results-for-Second-Quarter-Fiscal-2026/default.aspx?ref=tokyobrainie.ghost.io) / [Latest interim results](https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-Financial-Results-for-Second-Quarter-Fiscal-2027/?ref=tokyobrainie.ghost.io) The [historical-price page](https://investor.nvidia.com/stock-info/historical-price-lookup/default.aspx?ref=tokyobrainie.ghost.io) did not expose the required price rows. S&P 500 earnings definitions and aggregation were also not established. The article’s drawdown, returns and P/Es are therefore unreproduced, not disproved. Replication needs both comparison closes, the October 6 high and full-year high check, split adjustments, price versus total returns, GAAP versus adjusted earnings, diluted EPS, and the index’s P/E methodology. Trailing and forward multiples must not be mixed. Earnings quality also matters. Q2 diluted EPS was USD 2.46 GAAP versus USD 2.22 adjusted. Adjustments remove items including investment gains: adjusted profit need not exceed GAAP profit. A low trailing multiple can reflect non-core gains, so it does not automatically demonstrate cheap operating earnings. [Reconciliation](https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-Financial-Results-for-Second-Quarter-Fiscal-2027/?ref=tokyobrainie.ghost.io) A P/E is compatible with many combinations of margins, reinvestment, discount rates, growth duration and terminal value. Sales growth is not EPS growth. As arithmetic only, take an initial earnings level of 1 and the article’s hypothetical price of 25.9\. Earnings rising 30% with a 20x multiple produce 26, almost unchanged. Earnings rising 10% with a 30x multiple produce 33, about 27% higher. These are sensitivities, not forecasts or revealed market expectations. The supported conclusion is that historical revenue growth is reproducible; undervaluation and a specific “priced-in” growth rate are unproven. A reverse DCF must disclose its assumptions. ### 6\. Investigate financial circulation without assuming fraud The 10-Q reports USD 31.005bn in net additions to non-marketable equity holdings during the half year and USD 7.504bn in associated unrealized gains. These investments cannot all be labelled customer financing or GPU-purchase funding. [Notes 5–6](https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm?ref=tokyobrainie.ghost.io) Equity investment, loans, purchase guarantees and cloud-service commitments are different transactions. Customer-investor overlap alone does not prove fictitious sales; management’s assertion of real demand does not eliminate credit risk. Track contract payment terms, third-party customer receipts, receivable ageing and losses, earnings excluding investment revaluations, and investment outflows alongside operating cash flow. When financing precedes collection, identify the duration and who bears a loss. ### 7\. Japan: documented connections versus profit hypotheses | Route | Evidence and remaining condition | | ---------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Local infrastructure | July 16 NVIDIA announcement with Noetra: planned 140MW, 13,750 Vera CPUs and 27,500 Rubin GPUs. Announcement is not completion, acceptance or paid operation | | Factory and robotics development | July 15 announcement describes Fujitsu exploring a control platform and technology integration by FANUC, Yaskawa and Kawasaki Heavy Industries. Commercial contracts and operating benefits still matter | | Telecom and industrial AI | SoftBank, NEC, Hitachi and Sony are among named prospective participants. Participation is not booked profit | | Materials, equipment, testing, optics, cooling and power systems | Plausible transmission channels; this investigation has not verified individual Rubin design wins, volumes or margins | [Japan infrastructure](https://nvidianews.nvidia.com/news/japan-government-industrial-leaders-and-nvidia-launch-the-worlds-first-national-ai-infrastructure?ref=tokyobrainie.ghost.io) / [Industrial ecosystem](https://nvidianews.nvidia.com/news/japans-robotics-and-manufacturing-leaders-build-on-nvidia-cosmos-to-advance-physical-ai-frontier?ref=tokyobrainie.ghost.io) Multiplying 0.14GW by USD 40bn yields USD 5.6bn mathematically, but this is not a disclosed NVIDIA contract value for the Japanese project and is not adopted as one. Users may gain compute and automation while bearing imported equipment costs, FX exposure, power costs, replacement cycles and platform dependence. Suppliers can face expansion costs and input inflation. NVIDIA growth does not establish uniform profit growth across Japanese “related stocks.” ### 8\. What would change the assessment? The next results call is scheduled for November 17 at 2pm PT/5pm ET—November 18 at 7am JST—with results expected earlier. [Schedule](https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Sets-Conference-Call-for-Third-Quarter-Financial-Results/default.aspx?ref=tokyobrainie.ghost.io) Check Q3 against USD 108bn ±2%, realized versus forecast Rubin mix, margins and operating cash conversion, the AWS schedule discrepancy, a reproducible GW/configuration model, and Japanese construction, energization, commercial operation and supplier profitability. **TokyoBrainie’s Cut:** The jackpot is easy to illustrate. Cash collected is harder to establish. Read the power denominator, product content, delivery, profit and collection—in that order—before arriving at “cheap.” --- [Original Notion archive](https://app.notion.com/p/411bc361ccd281e89798c43fd73b7e1d?pvs=204&ref=tokyobrainie.ghost.io)