S6 (1:24:38–end)#

Scope. Transcript lines 475–568, about 21 minutes. The segment follows Huang’s account (1:21:05) of Nvidia compute as a collateralisable “asset class” with the “lowest” cost of capital. It covers:

• Nvidia’s ecosystem investments and “circularity”
• Nvidia as de facto industrial policy
• Bubble risk
• US–China: diffusion, “race” framing, export controls, safety cooperation
• Energy, climate and where data centres get built
• Books

Quotations are verbatim, except that I have silently removed stutter-repeats and marked omissions with “…”. [Context] notes are mine, drawn from primary sources where possible (Nvidia’s SEC filings, NIST, the EIA, the White House). Sources are listed at the end.

Register. This is the most expository stretch of the interview. Most of Klein’s questions invite explanation; the one sharp edge is “Obviously, you wanted those to be loosened” (1:34:16). Huang speaks in long, fluent turns. Klein does not ask about safety here, but Huang returns to “doomerism” twice anyway.

Turn-by-turn notes#

1. Supporting demand versus creating it (1:24:38–1:27:32)#

Klein picks up Huang’s cost-of-capital point: Nvidia has “moved now into lowering the cost of capital for others.” He mentions charts of “arrows going in every direction” and asks: “What is the difference between supporting demand, creating markets, and creating demand?” The premise is the circularity critique (Nvidia funds the firms that buy its chips). He poses it as a distinction to be drawn, not an accusation.

Huang (1:25:12) opens with a first-principles denial: “We can’t really create demand because in the end … if the AI services have no offtake, then obviously building computers for it is pointless.” Demand is high because “AI applications are going through an inflection; they’re becoming useful.” That has brought in “five hundred billion dollars of venture funding,” and “those thousands of companies, startups, they all need compute.” He then lists Nvidia’s reasons for investing:

• Startups need support with technology, ecosystem and finance.
• An equity stake can help one “become a really flourishing new cloud provider.”
• Firms outside language models (world models, physical AI, biology, chemistry, materials) “need a lot of capital.”
• “By being a first investor, anchor investor, we bring confidence to their company.”
• Nvidia “might decide to invest in a nuclear company.”

He closes with his organising model: “my mental model of the AI industry as a five layer cake, and we’re investing across all of it.” The “strategic unlock points” are new markets, “a new route to market for us,” and “a critical resource.”

Moves. First-principles argument; a statistic (repeated from 0:05:55); a reframe from circularity to ecosystem-building; a list of motives, several openly about Nvidia’s own market access.

Did he answer? Partly. The no to “can you create demand?” is clear, and the long-run argument holds. He does not address the mechanism critics cite: investment coming back as chip revenue, pulling demand forward. He never separates “supporting” from “creating markets,” or takes up what Klein actually named, lowering other firms’ cost of capital. And offtake only disciplines in the long run. It does not rule out financing running ahead of end demand, which is exactly Klein’s next question.

[Context] The filings make the overlap explicit:

• Customers as investees. FY2026 investments of $17.5bn “include AI model makers that purchase our products directly or through CSPs” (10-K).
• Capacity commitments. By 26 July 2026, $36bn to buy cloud capacity from “AI clouds” that “procure our data center infrastructure products.”
• OpenAI guarantee. Guarantees “capped at a total of $105 billion” (August 2026) back leases for an OpenAI affiliate on a campus that will “exclusively host NVIDIA AI infrastructure.”
• Financing platforms. August 2026 MOUs to “mobilize more than $500 billion of third-party capital” (10-Q).

These structures are bounded. The commitments shrink as third parties use the capacity, and the guarantee ends once OpenAI has a satisfactory credit rating. That fits Huang’s “support” framing.

On the $500bn: global VC in H1 2026 was $510bn across all sectors (Crunchbase). OpenAI and Anthropic took $217bn of it (43%), including Nvidia’s own $30bn in OpenAI. So “thousands of companies” understates concentration, and part of the venture “demand” is Nvidia-funded.

Tone. Calm, explanatory, confident.

2. “Single company industrial policy” (1:27:32–1:28:35)#

Klein calls the numbers “astonishing” and says Nvidia has become “a single company industrial policy for … American AI” (1:27:32). He asks for the total and notes it is “larger than the Chips and Science Act” (1:27:57). The frame implies a private firm doing a public job, with accountability implications he does not spell out.

Huang:

• “We’ve put a lot of money into this ecosystem” (1:27:41).
• “All in we might be like a hundred billion dollars … Might check my numbers” (1:27:47).
• He accepts the CHIPS comparison (“Oh yeah, yeah”) and extends it. Purchase commitments to TSMC, Wistron, Foxconn, Amkor and SPIL let him “encourage them to come and manufacture here.”
• “We probably contributed more to reindustrializing the United States in this chip manufacturing than just about any company in the world.”
• “We’re creating a shortage of labor, but we’re creating a lot of jobs” (1:28:00).

Moves. A figure with an explicit hedge (rare for him); the flattering frame accepted; a superlative claim about Nvidia; a causal claim that Nvidia’s commitments bring manufacturing onshore; a labour shortage recast as job creation, linking back to 0:05:55–0:11:29.

Did he answer? Yes, on the number. What a “single company industrial policy” means for governance goes unexamined by both men.

[Context]

• The $100bn is accurate for the equity book. At 26 July 2026 Nvidia held $42.8bn marketable and $51.2bn non-marketable equity securities (about $94bn), plus $25bn more in committed equity investments. This is carrying value, including unrealised gains. It excludes the guarantees and capacity commitments above.
• CHIPS appropriated $52.7bn ($39bn for manufacturing) of about $280bn authorised (NIST). The comparison holds against the appropriation, but the categories differ: CHIPS paid for fabs and R&D, while much of Nvidia’s book is equity in AI firms.
• Onshoring. The commitments are real: the April 2025 plan with those five partners for up to $500bn of US-built AI infrastructure, and supply commitments of $279bn (10-Q). The superlative is untestable, though, and TSMC’s Arizona build-out also reflects CHIPS subsidies and trade policy.
• Labour. The shortage remark matches his repeated appeals to electricians and plumbers, such as his May 2026 Carnegie Mellon commencement address.

3. Bubble (1:28:35–1:30:16)#

Klein frames it carefully. In the dot-com bust the technology was real (“It’s not that high valuations meant that the technology was hollow or fake”), but “something happened and popped.” What is the lesson, and why wouldn’t it repeat? The premise heads off the easy reply that AI is useful.

Huang (1:29:20): “At some point, demand and supply will be inverted again, and that’s just the nature of … markets.” But not “next year,” nor “in the next couple, two, three years. I just don’t believe that.” “We will likely have more supply than demand, and I just don’t know when that is. And so there’s not much to learn from the past.”

Klein (1:29:45): “What would be the signal for you?”

Huang (1:29:48): “Markets will naturally slow down and then it will stop.” There will be “a period of digestion” of six to twelve months, and “It won’t be forever.” He then pivots to Nvidia’s investments “into the application layer, so that each one of the industries could have the technology diffuse into them.”

Moves. Concession; a time-bounded prediction; admitted uncertainty about timing; dismissal of historical lessons; a softening euphemism (“digestion”); a pivot to diffusion as an implied safeguard.

Did he answer? Partly. He turns an asset-price bubble into a supply-and-demand (inventory) cycle. Valuations, investors and Nvidia’s own exposure go unmentioned. The “signal” he offers is the downturn itself, not an early warning. And “not much to learn from the past” does not follow from what precedes it (see Uncertain passages).

[Context] This is consistent with his November 2025 earnings-call line: “From our vantage point we see something very different” (CNBC). The 10-Q shows one direct customer accounted for 16% of Q2 FY2027 revenue.

Tone. Relaxed and brief.

4. Capability versus diffusion (1:30:16–1:32:17)#

Klein: the US emphasises “the speed of rising capability” and leads. China emphasises diffusion, where “a lot of people think” it leads, with an economy better built for it (WeChat). Which matters, and is it “a race at all”?

Huang (1:31:03): “That’s the ultimate question.” America benefits only if “every single industry has to benefit”: Walmart, Safeway, FedEx, banks, healthcare, construction, power. Then “everybody in the world.” The application layer is “the most important layer. That’s the layer that touches society.” “I want to see us not ruin the opportunity for the United States to benefit at the highest level.”

Then, unprompted: “all of the rhetoric and all the alarmism, all the doomerism, all of the predictions are scaring people. That is my greatest fear.” And: “Maybe I have more confidence in them than they have in themselves.” Klein agrees (1:32:07). Huang: “maybe it’s just too much humility” (1:32:09).

Moves. A normative answer rather than a comparison; a list of industries; the five-layer model; a pivot to doomerism; a charitable-sounding reading of the labs’ warnings.

Did he answer? Implicitly, on which matters (diffusion). He does not say who leads, or engage the claim about China’s structure. Asked about geopolitics, he names fear at home as the main threat.

[Context]

• Klein’s premise is contested. Jeffrey Ding argues China has a diffusion deficit (RIPE 2023; Technology and the Rise of Great Powers, 2024).
• “Them” most plausibly means the frontier labs.
• Shifting attributions. Huang’s readings of the labs’ motives vary: “deflection of blame” (0:55:46); “humility” here; and “they must be doing it for ulterior reasons” in a CBS interview aired days earlier (as reported by Fortune).

5. Is it a race? (1:32:17–1:34:16)#

Klein: “Should we conceptualize what we’re in as a race with China?”

Huang (1:32:23): “I don’t think it’s necessary. Some people like to think that way. I don’t. I don’t find that necessarily inspires me.” Nvidia never mentions rivals: “we hold ourselves to our own standard.” Uniting organisations “outside of contests” takes “more artistry.” Even if it is competition, “It doesn’t have to be that if they achieve something, it’s at our peril.” A Chinese energy invention could help the US. Chinese open models “are now being used by eighty percent of the American startups.”

Klein: “we use a lot of Chinese open models here” (1:33:49).

Huang (1:33:51): “We download it. We make it our own. We fine tune it. We put it into our own agent harness. We put it into our own sandbox.” The result is “your own technology.”

Moves. A geopolitical question answered through management philosophy; a positive-sum reframe; a hypothetical; a statistic; a containment argument consistent with 0:27:02 and S3–S5.

Did he answer? Yes, directly, though on motivational and economic grounds rather than strategic ones.

[Context]

• Past race language. The FT reported him saying (5 November 2025) “China is going to win the AI race.” He then posted: “China is nanoseconds behind America in AI. It’s vital that America wins by racing ahead and winning developers worldwide.” Compare “We’re not going to let that happen, sir” (0:40:02).
• The fair reconciliation: he rejects a zero-sum capability race but embraces competition for markets and developers (“America is about competition,” CBS).
• The 80% figure appears to trace to a16z’s Martin Casado (Economist, November 2025). Casado clarified it meant 80% of the 20–30% of startups using open models, about 16–24% overall.
• Security. NIST’s CAISI (September 2025) found DeepSeek agents “12 times more likely” than US models to follow malicious hijacking instructions, and found the models echoing CCP narratives. Sandboxing addresses the first finding, not the second.

6. Export controls (1:34:16–1:36:59)#

Klein sets out the case for denial (compute as a geostrategic resource; a race to “recursively improving … superintelligence”). Controls were “pretty tight” under Biden and “loosened under Donald Trump.” “Obviously, you wanted those to be loosened.” Is it good for China to have chips “that could accelerate their models … versus us holding that back to try to slow their progress?”

Huang (1:35:15):

• “Our goal is that all of America benefits.”
• The US has “a greater ambition for the world to be built on the American tech stack. Just as we have greater ambition that the world is built on the U.S. dollar, and that more people speak English.”
• “Are we depriving them a chip for their industry, or are we depriving United States a market to compete in?”
• Of denial: “Maybe it helps one company with a particular model, but the rest of the industry suffers.”
• Frame it by “what’s in the best interest of America first, all of America, not one … company.”
• “The race is if there is one, it’s about all of the economy of the United States succeeding.”

Moves. A reframe from China’s gain to America’s loss; an analogy (dollar, English, internet) to dominance through standards; a rhetorical question; an implied charge that the case for controls serves narrow interests; “America first” language.

Did he answer? Half of it. Whether Nvidia chips would speed up Chinese frontier or military capability goes unaddressed. So does the commercial interest Klein named.

[Context]

• Nvidia’s own filings show the stake. The Trump administration first tightened controls (an April 2025 H20 licence requirement, costing a $4.5bn charge) before licensing H20 (August 2025) and small H200 volumes (February 2026).
• Beijing now also blocks: “such sales were restricted by the PRC government.” Nvidia is “effectively foreclosed” from China’s data-centre market, which “helped our competitors build larger developer and customer ecosystems.” That is the company-level version of Huang’s national argument. Neither speaker mentions Beijing’s restrictions.
• “American tech stack” echoes Executive Order 14320 (23 July 2025).
• Interpretation: “one company with a particular model” plausibly alludes to Anthropic, the most prominent industry advocate of controls. Huang does not say so.

7. Safety cooperation and “America first” (1:36:59–1:39:05)#

Klein gives his own view. He is “conflicted.” If you fear superintelligence you want “productive bilateral” work with China, and “the more you think of it as a race which only one side can win … the more you are necessarily going to create enmity.”

Huang (1:37:36):

• “I deprive you of this, therefore I win. That simplistic logic tends to have unintended consequences of the bigger game.”
• “We want them to build safe products because when they don’t build safe products, it hurts the whole industry.” So now is the time to “communicate, collaborate, to understand, align as much as possible.”
• “Nvidia is an American company. We should benefit America first.” “Every single generation of our product goes to American companies first.”
• A legal requirement to that effect? “I’m delighted by that … We do that naturally, anyways.”
• Still, “we need every single layer to go out there and compete for the market.”

Moves. Agreement with Klein; safety framed as industry reputation; support for dialogue; a patriotic commitment; acceptance of a rule that codifies existing practice.

Did he answer? Yes. This is where the two men come closest in the segment.

Notable. This is one of only two regulations he explicitly welcomes in the interview (the other: third-party auditors, 0:51:20), and it costs Nvidia nothing. He endorses cooperation with China without proposing a mechanism.

[Context] Vera Rubin began shipping in July 2026 to OpenAI, Microsoft, Google, Meta, CoreWeave and others. “Every generation … first” is unverified. Not verified this session: I recall Nvidia opposed the 2025 GAIN AI Act, which would have given US buyers priority before exports. If so, that sits awkwardly with “I’m delighted by that.”

8. Energy, climate and communities (1:39:05–1:45:24)#

Klein (1:39:05): “one of the advantages China has right now in AI is energy.” Where is the US, as it tries “to move from dirty energy into clean energy”?

Huang (1:39:53): China has “a lot more energy.” “We got ourselves really gummed up in climate change and sustainable energy, and as a result, we just didn’t plan enough energy production.”

Klein (1:40:14): “What do you mean by gummed up there?” A good press on a loaded phrase.

Huang (1:40:15–1:44:44), one long turn with Klein’s interjections merged in:

• Diagnosis.
• “In the near term, energy production requires fossil fuel.”
• Angst over fossil fuel meant “we’ve produced very little net new energy for a long time.”
• Self-criticism.
• “We could have done so much better job communicating with the communities, preparing the communities, working with the communities.”
• If a town refuses, “then so be it.”
• Advice to builders.
• Explain that water use “is really efficient these days.”
• Bring your own power.
• “It’s going to lower their property taxes.”
• Use setbacks; fund schools, parks and roads.
• Doomerism.
• “What reasonable person says, come and build this data center in my town, and by the way, whatever you produce is going to … end humanity.”
• “This negative doomer narrative is not helping our country.”
• Optimism. After Klein’s “there is a reality of climate change”:
• AI demand is funding solar, batteries, fission, fusion and hydro “like no time in history.”
• “There’s no question that in four or five years’ time, we’re going to use a lot more fossil fuel,” yet we have never been “better prepared to move to sustainable energy.”
• Data centres “in space.”
• “You don’t need government subsidies for the first time in hundred years.”
• To “turn the corner on … climate change … lean into AI. It is the best opportunity we have to get there.”

Klein (1:44:44): “we need to build the energy faster … you could subsidize it and you can make it easier to build.”

Huang (1:44:52) answers with a surgery analogy: “in order to save you, they got to hurt you first … They had to cut you open to save you … inflict an enormous amount of pain and suffering on you so that they could save you … I kind of think AI is kind of like that.” For several years “we have to unfortunately use … fossil fuel” (after a slip, “renewable energy”). Then: “hopefully we can transition.”

Moves. A causal claim (climate angst led to under-building); concession and practical advice; doomerism blamed for local opposition; a prediction; a market-optimist claim; AI reframed as a climate solution; an analogy that concedes near-term harm; a hedge.

Did he answer? Largely, and at length. “Gummed up” turns out to mean anxiety about fossil fuels, not specific regulations. The climate challenge gets future optimism, not a reckoning with near-term emissions.

[Context] Support for his claims is mixed:

• “Little net new energy.” True for electricity: the EIA says consumption was essentially flat for about 15 years. False for energy overall: primary production hit a record 103.3 quads in 2024.
• The cause. The EIA explains the stall by flat demand, not climate angst.
• “Near term requires fossil fuel.” Solar and batteries were expected to make up 81% of 2025 additions, against 4.4 GW of gas (EIA). The claim holds, charitably, for firm round-the-clock power.
• Local opposition is real: Data Center Watch counts $64bn of projects blocked or delayed in 2024–25, and about $130bn in Q1 2026. But the reasons it lists are local (water, bills and grid strain, noise, tax breaks, land use). It does not mention doom narratives, and Huang offers no evidence for that link.
• Property taxes. Loudoun County collects about $1.3bn a year from data centres and has cut residential rates. The effect varies by jurisdiction.
• Water. The “300×” water-efficiency figure for GB200 NVL72 is Nvidia’s own claim.
• Consistency. His apologetic stance matches CBS (20 September 2026): “we’re sorry we didn’t come talk to you sooner.”

9. Books (1:45:24–end)#

Huang (1:45:28):

1. Hennessy and Patterson, Computer Architecture: A Quantitative Approach. It “reduced the complexity, the abstract idea of computer architecture down to engineering.” “I love it when … people take complicated concepts and reduce it into something that you could do something about.”
2. Christensen, The Innovator’s Dilemma: “how to set proper expectations about it, and how to extrapolate maybe its future impact.”
3. Ries and Trout, Positioning: “a book about strategy, and … how people see the world and how people see products.”

Each book is described by what it lets you do. The titles match the episode page.

Claims made in this segment#

H = Huang, K = Klein.

What this segment reveals#

Observations (what is on the page)#

• The five-layer cake does triple duty. It explains Nvidia’s investments (1:25:12), defines national benefit (1:31:03), and justifies access to foreign markets (1:37:36).
• Doomerism comes up unprompted, twice. He calls it his “greatest fear” (1:31:03) and blames it for local opposition (1:40:15).
• Concessions concern execution, not structure. He concedes that a glut will come, that the industry was slow to engage communities, that fossil fuel use will rise, and that there will be “pain and suffering.” He rejects circularity (implicitly), zero-sum racing, export controls and subsidies.
• National interest is defined economically. It means industry adoption, market share and the “tech stack.” Security uses of chips in China are never mentioned.
• Hedging is selective. He hedges on $100bn and on bubble timing; the 80% and $500bn figures are stated flatly.
• His account of the labs’ motives shifts: deflection (0:55:46), then humility (1:32:09), then “ulterior reasons” (CBS).
• All three books are about making complexity actionable, or about perception and strategy. None is about society, history or ethics.
• Klein does not press as hard here. He names Huang’s commercial interest once (1:34:16), but does not follow up on circularity, bubble exposure or security.

Interpretations (mine)#

• Demand realism is a master premise. End demand disciplines everything. This mirrors his safety argument (0:40–0:47), where customers and liability discipline firms. The consistent worldview is that market signals plus engineering competence beat prospective regulation.
• Everything is a transition. A bubble becomes “digestion,” energy becomes surgery, and the labs’ safety failures a “transition” (1:11:19). Harms are temporary phases on the way to a good end state. The risk is that this framing makes harms look self-limiting by definition.
• He values reducing things to engineering. “Something you could do something about” helps explain his impatience with probabilistic warnings and anthropomorphic language. Questions that don’t reduce to engineering, like political economy, geopolitics and motives, get thinner treatment. He answered the race question with management philosophy.
• Narrative as a causal force. His most consistent causal theory in the interview concerns perception: fear stops adoption, stops data centres, and stops students going to college. His choice of Positioning fits.
• Alignment with Nvidia’s interest. Nearly every position coincides with it: broad adoption, access to China, energy build-out, no bubble. That doesn’t make him wrong, and he would argue, plausibly, that Nvidia’s interests and America’s coincide. But he disclaims “one company” interest on the very issue where Nvidia is the company most affected.
• A market-hegemony vision. He rejects a zero-sum race but embraces dominance through standards. His vocabulary (“America first,” “tech stack,” “doomer,” “gummed up in climate change”) tracks the administration’s, while diverging from its China hawks.
• Safety as product quality. Even when endorsing dialogue with China, he frames safety as protecting the industry’s reputation. The catastrophic-risk frame is absent.

Left unsaid or avoided#

• How investment dollars come back as chip revenue, and the scale of Nvidia’s guarantees (up to $105bn for the OpenAI-related campus, signed a month earlier)
• Nvidia’s own exposure to a correction
• The military and security side of selling chips to China, and Beijing’s restrictions
• Who pays for grid upgrades and higher bills; how large near-term emissions would be
• Who a “single company industrial policy” answers to
• What US–China safety cooperation would involve

Uncertain transcript passages#

• 1:24:38. “…so interesting yeah” mixes voices.
• 1:25:12. “We might decide to write so on so forth” is garbled.
• 1:27:44. “You’re not making per year” is probably “Are you making that per year?”
• 1:28:00. “Wisetron / Amcor / Spill” are Wistron, Amkor and SPIL.
• 1:29:20. “There’s not much to learn from the past” may be truncated (perhaps “about timing”). Don’t lean on it.
• 1:31:03 and 1:32:09. “Them” is probably the frontier labs; it could be the public. “…too much humility and and and otherwise” trails off.
• 1:32:23. “as net” is probably “as necessary.”
• 1:33:49. The “here” in “we use a lot of Chinese open models here” is unclear (the US? the NYT?).
• 1:34:16. “pretty people see us” is probably “particularly people who see us.”
• 1:35:15. “because it deprived open models” is unclear.
• 1:37:36. “America has every right, and for these technologies…” is garbled.
• 1:39:05. Punctuation is garbled. “Probably a time” is probably “particularly at a time.”
• 1:40:15–1:44:44. Klein’s interjections are merged into Huang’s turn: “What do you mean we started off on our back foot?” and “Well, how do you balance? … there is a reality of climate change.”
• 1:44:44. Speaker boundaries unclear. “That’s right” may be Huang’s.
• 1:44:52. “use renewable energy, use fossil fuel” is a self-correction to fossil fuel.
• 1:45:28. “Al Reese’s” is Ries. Klein’s sign-off is merged into Huang’s turn.

Sources consulted#

• Nvidia 10-Q (quarter to 26 July 2026): https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm
• Nvidia 10-K FY2026: https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm
• Episode page: https://podcasts.apple.com/us/podcast/jensen-huang-thinks-a-i-alarmism-has-gone-too-far/id1548604447?i=1000791251478
• NIST CHIPS fact sheet: https://www.nist.gov/document/chips-america-fact-sheet-federal-incentives
• April 2025 US manufacturing plan: https://www.supplychaindive.com/news/nvidia-us-production-blackwell-tsmc-ai-trump-tariffs/745395/
• Crunchbase H1 2026: https://news.crunchbase.com/venture/global-startup-exits-ipo-ma-soar-ai-q2-h1-2026/
• TechCrunch on Nvidia equity deals: https://techcrunch.com/2026/05/09/nvidia-has-already-committed-40b-to-equity-ai-deals-this-year/
• CNBC (headline only), bubble remark: https://www.cnbc.com/2025/11/19/nvidias-huang-rejects-ai-bubble-we-see-something-very-different.html
• FT remark and X clarification: https://finance.yahoo.com/news/nvidias-jensen-huang-says-china-211900769.html
• Casado clarification: https://x.com/martin_casado/status/1990462245541982546
• NIST CAISI on DeepSeek: https://www.nist.gov/news-events/news/2025/09/caisi-evaluation-deepseek-ai-models-finds-shortcomings-and-risks
• EO 14320: https://www.whitehouse.gov/presidential-actions/2025/07/promoting-the-export-of-the-american-ai-technology-stack/
• Ding (2023): https://www.tandfonline.com/doi/full/10.1080/09692290.2023.2173633
• EIA: https://www.eia.gov/todayinenergy/detail.php?id=65264 and https://www.eia.gov/todayinenergy/detail.php?id=64586
• Data Center Watch: https://www.datacenterwatch.org/report
• Loudoun County: https://www.loudoun.gov/6409/Tax-Revenues-County-Budget
• CBS (20 September 2026): https://www.cbsnews.com/news/jensen-huang-nvidia-rejects-ai-extinction-warnings/
• Fortune (21 September 2026): https://fortune.com/2026/09/21/jensen-huang-ai-leaders-doomsday-narratives/
• Nvidia water efficiency: https://blogs.nvidia.com/blog/blackwell-platform-water-efficiency-liquid-cooling-data-centers-ai-factories
• Vera Rubin shipments: https://gcn.com/nvidia-vera-rubin-chips-begin-shipping/20421/