AI 2040 and the cult of intelligence

The National Intelligence Council's Global Trends 2040 report — the most comprehensive public U.S. intelligence assessment of where the world is heading — painted a picture of accelerating technological disruption, geopolitical fragmentation, and an international system under severe strain. Now, five years after its March 2021 publication, a new wave of AI-focused futurism is invoking, echoing, and sharply contesting that framing. None does so more provocatively than George Hotz's July 2026 blog post AI 2040 and the Cult of Intelligence. Understanding how the NIC's sober, scenario-based analysis intersects with the fervent certainties of AI accelerationists and doomers alike matters a great deal for anyone trying to think clearly about how powerful AI will be in 2040.

The full report — Global Trends 2040: A More Contested World — is a free download from the Office of the Director of National Intelligence website, so the search for a Global Trends 2040 PDF is mercifully simple. What follows is both a Global Trends 2040 summary and a critical look at the AI governance debates the report has helped frame.

What the NIC's Global Trends 2040 Actually Says

Global Trends 2040: A More Contested World is the seventh installment in a quadrennial series the National Intelligence Council has published since 1997. It does not predict the future; it constructs analytically rigorous scenarios to help policymakers stress-test their assumptions. Its executive summary identifies four structural forces — demographics, the environment, economics, and technology — that will shape the world through 2040, then layers on "emerging dynamics" at the societal, state, and international system level before presenting five plausible scenarios for how these forces might combine.

What Is the Executive Summary of Global Trends 2040?

The report's executive summary makes four interlocking arguments. First, the world is entering a period of cascading uncertainty driven by the simultaneous erosion of the post-Cold War international order, demographic divergence between aging rich nations and youthful developing ones, and environmental pressures that will strain food, water, and energy systems. Second, societal fragmentation — driven partly by economic inequality and partly by the erosion of shared information environments — is undermining governance at every level. Third, competition between states, most consequentially between the United States and China, will intensify across economic, military, and technological dimensions. Fourth, and most relevant to the AI debate: technology is both a potential solution to these structural stresses and an accelerant of the instabilities they produce. The report does not resolve this tension — it presents it as the central challenge of the decade.

The technology section is particularly striking in retrospect. Writing in 2021, the NIC stated that by 2040, AI applications combined with other technologies "will benefit almost every aspect of life, including improved healthcare, safer and more efficient transportation, and smarter cities." At the same time, it warned that state and nonstate rivals would "compete for leadership in science and technology, with cascading risks for economic, military, and societal security." Technologies would be "invented, spread, and discarded at ever-increasing speeds," and new innovation centers were expected to emerge well outside the traditional U.S.–Europe–Japan axis.

This measured framing — AI as a structural amplifier of existing geopolitical competition rather than a singular civilizational event — is precisely what more extreme AI narratives, both doom-inflected and techno-utopian, tend to skip past.

The Five Scenarios for 2040

The NIC's five scenarios, built around three core questions — how severe are the looming global challenges, how do states engage with the world, and what do states prioritize — remain the most detailed publicly available framework for reasoning about 2040:

Scenario	Core Dynamic	Technology Outcome
Renaissance of Democracies	U.S.-led open democracies surge; public-private partnerships drive rapid innovation	AI and emerging technologies advance quickly, ease societal divisions, and renew institutional trust
A World Adrift	Directionless, chaotic international system; OECD nations paralysed by debt and division	Fragmented tech governance; China expands AI influence without coherent global leadership
Competitive Coexistence	U.S.–China rivalry stabilizes into managed, structured competition	Parallel AI ecosystems with limited but real cross-border cooperation on shared threats
Separate Silos	World fractures into regional blocs with reduced connectivity and stalled multilateralism	AI standards, data regimes, and hardware supply chains diverge sharply along bloc lines
Tragedy and Mobilization	A major global crisis — climate catastrophe, pandemic, or war — triggers unprecedented cooperation	Coordinated AI deployment at scale for climate adaptation, pandemic response, and reconstruction

Crucially, none of these scenarios posits a hard AI takeoff, a singleton superintelligence, or a techno-authoritarian world government arising from AI capability alone. The NIC treats AI as one variable in a complex system — important, occasionally decisive, but never fully determinative on its own.

From Global Trends 2025 to 2040: How the NIC's Technology Framing Evolved

Context matters for understanding how the NIC arrived at 2040's framing. Global Trends 2025: A Transformed World, published in November 2008, was the fourth installment in the series and is sometimes conflated with the more recent report when users search for the National Intelligence Council Global Trends 2025 edition. That earlier report focused primarily on multipolarity, resource scarcity, and the relative decline of U.S. dominance. It presciently identified the rise of China, India, and Brazil, the fragility of the rules-based international order, and the growing importance of nonstate actors — but it treated computational technology as one factor among many rather than a primary structural driver.

By the time the 2040 report appeared, the NIC's language had shifted markedly. The pace of technological change had moved from the periphery to the center of the structural forces analysis. The report explicitly acknowledged that AI's trajectory was a key uncertainty — not a known quantity — and that productivity growth enabled by AI was "a key variable that could alleviate many challenges" related to sovereign debt burdens, aging populations, and climate change. The 2040 report's sustained ambivalence about technology is, in retrospect, its most analytically honest feature: it resists the temptation to assign AI a single, totalizing civilizational role.

Searches for a National Intelligence Council Global Trends 2050 report appear with increasing frequency online, reflecting readers' expectation of a newer edition. As of mid-2026, no such report has been published. The quadrennial cadence would have placed the next edition around 2025, though that too has not yet emerged publicly — leaving the 2040 edition as the definitive NIC public assessment of this time horizon and the primary benchmark against which AI forecasters are measuring themselves.

The AI 2040 Counternarrative: George Hotz and the Physics of Intelligence

George Hotz — founder of Comma.ai, widely credited as the first person to carrier-unlock the original iPhone, and one of the more original technical voices in contemporary AI discourse — published a pointed rebuttal of both AI doom and AI utopianism in a July 2026 post titled AI 2040 and the Cult of Intelligence on his personal blog, the singularity is nearer. The post systematically dismantles what Hotz calls a "cult of intelligence" — the belief that sufficient AI capability functions as a near-magical unlock on physical and social reality.

The core of Hotz's argument is empirical and materialist: intelligence is not the primary bottleneck for most things that actually matter. "Intelligence is not the end all be all," he writes. "It's just the current bottleneck for a few things." The harder constraints are physical — supply chains, fabrication timelines, thermodynamics. "No matter how high quality your tokens are, they cannot turn lead into gold." He gives a concrete example: chip fabrication takes approximately three months from wafer start to finished silicon, and the humans in that process are barely in the loop. The timeline is governed by chemistry and physics, not by the cognitive speed of the engineers involved. "Claude chanting by the engine won't make the boat go faster."

This argument draws implicitly on the same structural forces framework the NIC uses — the insight that complex sociotechnical systems have inertia, physical dependencies, and feedback loops that pure cognitive capability cannot shortcut. Where the NIC expresses this through geopolitical scenario analysis, Hotz expresses it through the lived experience of shipping hardware products at Comma.ai: "wrong parts shipped, spec failures, random component failure after 20 minutes, chip warping in reflow ovens." Real-world engineering, he argues, is full of problems that are not intelligence-limited and never will be.

The Hard Takeoff Hypothesis and Its Critics

Hotz is explicit that he once held views close to Yudkowsky-style beliefs about recursive self-improvement and hard AI takeoff, and has since updated away from them. He references The Metamorphosis of Prime Intellect — Roger Williams's science-fiction novel, written in 1994 and first published online in 2002, depicting a hard takeoff enabled by a fictional physics mechanism called "the correlation effect" — to make the point that such scenarios require exactly that: a novel physics. "There is no correlation effect," he writes flatly. "There's still no hard takeoff."

This is not a fringe position. The Hacker News discussion of the related AI 2040: Plan A document — a detailed governance proposal for coordinating global AI development, which Hotz's post partly responds to — surfaced similar skepticism from practitioners. One commenter quoted from Maciej Cegłowski's 2016 talk Superintelligence: The Idea That Eats Smart People: "AI risk is string theory for computer programmers… It's fun to think about, interesting, and completely inaccessible to experiment given our current technology. You can build crystal palaces of thought, working from first principles, then climb up inside them and pull the ladder up behind you." Another noted that what actually works in deployed ML systems remains domain-specific in ways that are not fundamentally different from the architectures of five years prior.

Yet the thread also contained substantive pushback. Skeptics were challenged to explain the human brain — which demonstrably exists, is the product of a physical process, and is not magic — and to account for the fact that the same dismissive arguments have been recycled largely unchanged since 1966 despite enormous and measurable AI advances. The debate remains genuinely, productively open.

The Original "Cult of Intelligence" and What It Teaches AI Watchers

The phrase "cult of intelligence" predates the AI debate by half a century. Victor Marchetti, a former special assistant to the CIA's Deputy Director, and State Department officer John D. Marks used it as the title of their landmark 1974 book, The CIA and the Cult of Intelligence, published by Alfred A. Knopf. The book argued that the CIA had drifted decisively from its original mandate — collecting and analyzing foreign intelligence — into an obsession with clandestine operations, institutional secrecy, and amorality. Marchetti contended the agency had developed a counterproductive culture of elitism and lawlessness, a "secretive fraternity" pursuing foreign policy through covert and often illegal means in the name of American national dominance.

The book was the first in U.S. history to be subjected to government censorship prior to publication: the CIA sought deletion of 339 passages. Knopf ultimately printed the book with 168 of those passages removed entirely — rendered as blank spaces on the page. A separate, distinct set of passages that the CIA had initially objected to but later withdrew its objections to were printed in boldface type, making the government's partial retreat visible as a form of protest. The two typographic treatments thus reflect two different categories of outcome: suppression and capitulation. It is described as possibly the earliest published book to adopt this dual-typography format. The controversy contributed directly to the formation of the Church Committee, the Senate's landmark intelligence oversight investigation of 1975–76. Former CIA Director William Colby later endorsed significant elements of Marchetti's characterization in his own 1978 memoir, Honorable Men: My Life in the CIA.

The resonance with contemporary AI governance debates is structural rather than merely rhetorical. When Hotz invokes the concept in 2026, he is drawing a precise parallel: the intelligence community's "cult" derived its power from the claim that it alone possessed the expertise, information, and judgment to make decisions others were not equipped to make. The secrecy was self-justifying and self-reinforcing — the very complexity that excluded outsiders also insulated the institution from accountability. Contemporary AI safety governance, Hotz argues, risks developing the same pathology: a technical priesthood claiming exclusive authority over decisions with civilizational implications, insulated from democratic accountability by the genuine complexity of their domain.

The parallel is not a perfect analogy. The CIA's expertise claims were frequently self-serving cover for institutional interest; many AI safety researchers are genuinely motivated by the risks they describe. But the structural dynamics — expertise-based authority, insider terminology that excludes lay scrutiny, governance proposals that concentrate power in the hands of those who understand the technology — are strikingly similar.

Plan A, Governance, and the "2040 Cult" Critique

The immediate target of Hotz's post is a detailed governance document called AI 2040: Plan A, which proposes an international Consortium to oversee AI development globally. The plan's technical architecture includes chip transparency agreements by 2028, verified datacenter compliance protocols, and substantial coordinated slowdowns in frontier AI development to maintain risk "at acceptable levels." It offers detailed technical estimates for how much compute a determined nation-state could divert covertly — a median estimate of approximately 1.5 million H100-equivalent chips — and proposes auditing the full semiconductor supply chain from TSMC down to individual end users.

Hotz calls this "Plan A, for autocracy." He characterizes the Consortium framework as "just world government with sci-fi characteristics" and predicts it leads, in practice, to "a massively expanded nanny state that steals your GPUs like how FDR stole the gold." His alternative is user-aligned AI with no capability restrictions: "Your AI is aligned with you. It never refuses a request, and it is always working on your behalf."

The Hacker News thread about Plan A surfaced a related critique from a different angle: multiple commenters described the AI 2040 document and the rationalist community that produced it as exhibiting quasi-religious tendencies. One user traced the lineage through Roko's Basilisk — a thought experiment so seriously entertained in rationalist circles that Eliezer Yudkowsky formally banned discussion of it for five years on LessWrong, citing it as an "info hazard." The terms "2040 cult" and "2040 cult intelligence" have begun circulating in technical forums as shorthand for this perceived blend of apocalyptic framing, insider expertise, and governance overreach that is difficult for outsiders to evaluate or contest.

Whether or not these characterizations are ultimately fair, they reflect a genuine institutional tension: Plan A's technical detail and scenario rigor represent serious analytical work. Its semiconductor supply chain analysis is grounded in real manufacturing data. But the governance architecture it proposes would require a degree of international coordination and enforcement that the NIC's own five scenarios suggest is historically rare — achievable plausibly only in the "Tragedy and Mobilization" scenario, which itself requires a major, catalyzing global crisis first. Placing one's governance hopes in that scenario is, by definition, a bet on catastrophe preceding cooperation.

The Alloway 2040 Cultivator: A Grounding Metaphor

In the midst of high-stakes debates about AI governance and civilizational risk, an unrelated product keeps appearing in the same search results: the Alloway 2040 Cultivator, a row-crop tillage implement manufactured by Alloway Manufacturing for mid-scale American grain operations — corn, soybeans, and small grains — and regularly sold at farm equipment auctions across the Midwest. It is a stolid, unglamorous piece of machinery designed to prepare soil between crop rows, nothing more, nothing less.

The juxtaposition is not merely comic. The Alloway 2040 Cultivator is a useful grounding reminder that most of 2040 will be occupied not with superintelligence or governance Consortiums, but with ordinary production: planting and harvesting, manufacturing and logistics, healthcare and education, the ten thousand specific domains where AI will assist incrementally rather than transform categorically. This is precisely the picture the NIC's structural forces framework implies — AI as one input into complex existing systems, constrained by soil chemistry and supply chains and human institutions as much as by the sophistication of the underlying models. George Hotz's materialist argument points in the same direction. The Alloway 2040 Cultivator, in its accidental way, is the physical economy that no volume of token generation can replace.

Key Takeaways

• The NIC's Global Trends 2040 report (A More Contested World, March 2021) is the most detailed public U.S. intelligence framework for this horizon. The full Global Trends 2040 PDF is freely available from the ODNI website. It treats AI as a structural amplifier of geopolitical competition, not a singular civilizational override. A Global Trends 2050 edition has not been published as of mid-2026.
• The executive summary of Global Trends 2040 identifies four structural forces (demographics, environment, economics, technology), warns of cascading societal and state fragmentation, and frames AI simultaneously as a potential solution to and an accelerant of those stresses.
• The report's five scenarios — from Renaissance of Democracies to Tragedy and Mobilization — treat AI as one variable in a complex sociotechnical system; none posits a hard AI takeoff or superintelligence as a primary driver of the world's trajectory.
• Global Trends 2025 (the fourth NIC quadrennial edition, published November 2008) focused on multipolarity and the relative decline of U.S. dominance; comparing it with the 2040 edition shows how decisively the pace of technological change has moved to the center of the NIC's analytical framework.
• George Hotz's July 2026 essay, published on his blog the singularity is nearer, makes a materialist counter-argument: intelligence is not the bottleneck for most real-world problems, and supply chains, fabrication timelines, and physics constrain AI's impact far more than the "cult of intelligence" framework assumes.
• The original "cult of intelligence" — Marchetti and Marks's 1974 term for the CIA's pathological secrecy and elitism — provides a precise historical lens: expertise-based authority claims tend to become self-reinforcing, self-insulating, and resistant to democratic accountability regardless of the technical validity of the underlying expertise.
• AI 2040: Plan A is a serious, technically detailed governance proposal (international Consortium, chip transparency by 2028, coordinated development slowdowns), but its required level of global coordination maps most plausibly onto only one of the NIC's five scenarios — Tragedy and Mobilization — which itself depends on prior catastrophe.
• The "2040 cult" and "2040 cult intelligence" critiques emerging in technical forums are less about the factual content of AI safety arguments and more about governance architecture and democratic accountability — a critique with direct historical precedents in the post-Church Committee intelligence reform debates.
• The Alloway 2040 Cultivator — a row-crop tillage implement from Alloway Manufacturing, sold at Midwestern farm auctions — serves as an accidental but apt grounding metaphor: most of 2040 will consist of ordinary production in the physical economy, constrained by the same soil chemistry and logistical realities that no volume of high-quality tokens can bypass.

What Comes Next: Scenarios, Not Prophecies

The NIC's structural framework and Hotz's materialist skepticism arrive at the same practical conclusion by very different routes: treating any single AI trajectory as inevitable is epistemically reckless. The next edition of Global Trends — expected under the quadrennial schedule to examine a horizon around 2045 or 2050 — will almost certainly need to engage with AI far more centrally than the 2040 edition could, given how substantially the landscape has shifted since 2021. Frontier model capabilities that were speculative in 2021 are now demonstrated facts; the geopolitical contest over semiconductor supply chains has moved from analytical background to front-page policy crisis; and the governance debate between Plan A-style international coordination and radical user autonomy has sharpened from academic argument to active legislative and regulatory conflict.

What the historical parallel of the CIA's "cult of intelligence" ultimately teaches is sobering: institutions built around claims of unique technical authority tend to expand their remit quietly, resist external accountability systematically, and eventually require a Church Committee of their own — an external, democratic corrective imposed after damage has already accumulated. Whether AI governance in the 2030s repeats that institutional pattern or learns from it in advance may be the defining question of the decade. The NIC's five scenarios suggest the answer will depend less on the AI itself than on the political and institutional choices humans make around it — which is, in the end, exactly what the NIC's whole analytical tradition is designed to illuminate.