The new space race is seizing the means of intelligence production
In space there is no place to hide. From space, masters of the earth would have the power to control the world.
My biggest takeaway after working on Capitol Hill was that our government runs primarily on fear. Politicians only overcome the friction of taking action when both parties share a common enemy; over the past decade this has often been when a foreign adversary (China) threatens American dominance.
Back in my day, the fear mongering centered around 5G networks. Lobbyists spread anxiety that Huawei would embed surveillance backdoors in every cell tower and the Chinese Communist Party would have access to American communications infrastructure. Before 5G, it was Sputnik. A beeping Soviet satellite compelled a terrified nation into funding NASA, DARPA, and the interstate highway system.
The space race was about space to the extent that it was a national defense priority; ensuring defensability was certainly a priority but It was also about demonstrating technological supremacy and proving that American systems could out-innovate Soviet systems. Fear of falling behind drove a decade of bipartisan investment that reshaped the American economy.
The release of DeepSeek was this era's Sputnik moment.
On the day of President Trump's second inauguration, January 20, 2025, a Chinese AI startup released its R1 model to the world. The timing was intentional. Within a week, DeepSeek had overtaken ChatGPT as the most downloaded free app on Apple's US App Store. Nvidia lost $600b in market value in a single day. Here was a private Chinese company demonstrating that it could match or exceed the capabilities of OpenAI and companies backed by tens of billions in American capital at a reported cost of $5.6m and training costs under $300k (though my friends who work in AI labs now reassure me that DeepSeek is really a bit of a fork and innovative ideas still remain in America's domain). A friend who works for a China consulting shop run by former Trump administration officials insists that the timing was coordinated by the CCP.
What made DeepSeek a Sputnik moment was not merely technical achievement, it was the psychological shock of discovering the gap that Americans convinced themselves existed was much closer than assumed. AI's potential in everything is so all-encompassing; applications in military, energy, agriculture, mining etc that the nation that owns the production stack becomes (without exaggeration) master of the universe.
However, comparisons of this moment to Sputnik and the space race fall short. The space race had a finish line. AI does not. It is a not a race towards a destination, it is a race to control the infrastructure upon which all future races will be run. Humanity stretches its fingertips towards owning the means of intelligence production.
Every industry, every ounce of economic activity, every instrument of state will run on AI infrastructure. Whichever nation controls the means of intelligence production controls the terms on which everyone else competes.
I wrote this essay so that I myself could better understand where the chokepoints are so that I could allocate my own dollars towards owning some amount of intelligence production. The companies and countries that control the AI infrastructure stack will extract rent from everyone forced to pass through.
The rest will pay tribute or be cut off entirely.
The cybernetic manifesto and our hybrid existence
The Cybernetic Manifesto predicted that by the late twentieth century, humanity would become cyborgs, fused with machines into seamless informational systems. The authors were directionally correct but temporally incorrect; this fusion is happening now, faster than the authors imagined.
In America, artificial intelligence adoption has been driven almost entirely by private enterprise. When Americans use AI, they treat it as a product; something to subscribe to, something to optimize their workflow, something to generate content for social media. The integration is voluntary, commercial, piecemeal. Federal agencies still run on legacy systems. AI adoption in America is happening in spite of the state, not because of it.
China has taken a different path. In contrast, Beijing treats AI as strategic infrastructure; a capability to be deployed across the entire apparatus of state and economy, from surveillance to agriculture to military logistics. The integration is mandatory, centralized, total. There is no distinction between commercial application and national strategy. They are the same thing.
On the outskirts of Shanghai, in Tinglin Township's Diantian Farm, over seventy engineers now 'herd' AI-enabled robots across the fields. These machines, iron-and-steel cattle equipped with crawler tracks and controlled via WeChat mini-programs, plough, plant, weed, and harvest with mechanical precision.
A single weeding robot operates for eight hours on one hour of charge, covering 33 hectares per day. It distinguishes crops from weeds using computer vision trained on millions of images, snipping invasive plants while leaving rice seedlings untouched. In Xinjiang, a 3,000-mu unmanned farm integrates aerial drones, ground robots, IoT sensors, and an intelligent farm management system. Seventy-five percent of operations are fully automated. Cotton yields have reached 529 kilograms per mu. These numbers were unimaginable just a decade ago.
Embedded tweet: 2014140962876825737In January 2025 (the same month DeepSeek was released), China's State Council issued its annual No. 1 Central Document, the highest-priority policy directive for agriculture. For the first time, the document identified new quality productive forces in agriculture as a top national priority, explicitly calling for AI integration across the entire food production chain: crop cultivation, animal husbandry, pest prevention, yield optimization. The National Smart Farming Plan mandates AI deployment from seed to harvest.
China must feed 1.4b people with less than 10% of the world's arable land and even less of its freshwater. Food self-sufficiency has declined from 93.6% in 2000 to 65.8% in 2020, with projections showing further erosion to 58.8% by 2030. The average age of Chinese farmers exceeds fifty in most provinces. China has deployed over 5,000 agricultural drones powered by its BeiDou satellite system. The country's agricultural robotics market is projected to grow from $750 million in 2022 to nearly $3 billion by 2030. Meanwhile, America's Farm Bill debates center on subsidy allocation, not technological transformation.
Embedded tweet: 2014444207679701451AI is not a productivity enhancement for China. It is an imperative survival strategy for the nation. By 2030, there will be almost no major economic activity untouched by AI.
This asymmetry explains the sudden urgency in Washington. It explains why the Trump administration has moved aggressively to secure critical mineral supplies, why Greenland has become a geopolitical flashpoint, why rare earth processing capacity is now discussed in the same breath as national defense. The administration understands that the AI race is not a software competition. It is a resource competition, and resources are physical. Resources exist in specific places, controlled by specific nations, subject to specific chokepoints.
You cannot build advanced semiconductors without rare earth elements. You cannot process rare earth elements without massive energy inputs. You cannot generate that energy without uranium, natural gas, or grid infrastructure. You cannot train frontier AI models without advanced chips. Every layer of the AI stack rests on physical foundations that America does not fully control.
China controls 60% of rare earth mining and 90% of rare earth processing. China dominates the refining of cobalt, lithium, and graphite, the materials that power the batteries that power the data centers that power the AI. China has spent two decades securing supply chains while America spent two decades offshoring them.
The Trump administration's resource nationalism is not random belligerence. It is a recognition (however terribly articulated as seen this past wekend) that sovereignty in the AI age requires control over the physical inputs to intelligence production. Greenland's rare earths, Arctic shipping routes, domestic uranium production, critical mineral stockpiles are not distractions from the AI competition, they are the AI competition.
The question is whether America can rebuild the industrial base it dismantled over the past decades of peace, secure the resources it neglected, and deploy AI at the scale and speed that strategic competition demands. China has a twenty-year head start on supply chain integration. It has a political system capable of mandating adoption. It has a population accustomed to state-directed technological transformation.
The cybernetic fusion is coming either way. The only question is who controls the mines, refineries, and foundries that make the code possible.
Seeking value in chokepoints
The AI supply chain can be understood as a vertical stack, with each layer dependent on the layers beneath it:
- Raw Materials: Rare earths, copper, silicon, uranium
- Semiconductor Equipment: Lithography, deposition, etching
- Foundries: Chip manufacturing
- Memory & Storage: DRAM, NAND, HBM
- Processors: GPUs, TPUs, AI accelerators
- Networking: Data center interconnects
- Energy Infrastructure: Power generation and transmission
- Data Centers: cloud compute facilities
- Software and Models: AI frameworks, foundation models, applications, agents
Bottlenecks at lower layers propagate upward: a shortage of high-bandwidth memory constrains GPU production -> a shortage of EUV lithography machines constrains advanced chip manufacturing -> a shortage of rare earths constrains everything.
Layer 1: Raw Materials
China controls 61% of global rare earth mining and over 90% of refining capacity. More critically, it controls 94% of permanent magnet production. These components are essential for electric vehicles, wind turbines, and the motors that position hard drive heads with nanometer precision.
When China imposed export controls on seven medium and heavy rare earth elements in April 2025 (samarium, gadolinium, terbium, dysprosium, lutetium, scandium, and yttrium) it demonstrated the ability to choke the AI supply chain at its root. The October 2025 escalation went further: Beijing asserted jurisdiction over foreign-made products containing Chinese-origin rare earth materials.
Copper hit a record $11,705 per tonne in December 2025, up 31% year-to-date. The IEA warns of a potential 30% supply shortfall by 2035. AI data centers are emerging as significant new demand: Bloomberg New Energy Finance estimates data centers could consume over 500,000 tonnes annually by 2030.
The supply response is structurally constrained. New copper mines take 29 years on average to permit and build in the US. Ore grades have fallen 40% since 1991.
Microsoft signed a 20-year PPA with Constellation Energy to restart Three Mile Island. Meta signed for 1.1 GW from the Clinton plant. Amazon secured 2 GW from Susquehanna. Google partnered with Kairos Power for SMRs.
Embedded tweet: 2012416124093407740The driver is AI's insatiable appetite for power. Deloitte projects US data center power capacity will grow from 33 GW in 2024 to 176 GW by 2035—a more than fivefold increase. Nuclear is the only carbon-free baseload source that can deliver the 24/7 reliability AI workloads require.
Enriched uranium prices have surged to $190 per SWU, up from $56 three years ago, reflecting Russia's 40% share of global enrichment capacity.
Embedded tweet: 2013377245931311521Layers 2 & 3: Equipment and Foundries
If there is a single company that embodies the fragility of the AI supply chain, it is ASML Holding. The Dutch firm maintains a 100% monopoly on extreme ultraviolet lithography machines, the only equipment capable of printing the sub-7nm features required for cutting-edge AI chips. Each machine costs $150-200m, takes 18 months to build, and requires 250 crates to ship.
EUV lithography requires 13.5nm light generated by vaporizing tin droplets with a 50,000-watt CO2 laser, hitting each droplet twice, 50,000 times per second. The light must then be reflected through mirrors polished to atomic smoothness. No other company has mastered this. Canon and Nikon abandoned EUV development decades ago.
Under pressure from the US, the Netherlands has restricted EUV exports to China. This is arguably the most consequential export control in modern history. Without EUV, China cannot manufacture chips below approximately 7nm, a ceiling that will increasingly constrain its AI ambitions.
Taiwan Semiconductor Manufacturing Company fabricates an estimated 92% of the world's most advanced chips. Every NVIDIA GPU, every AMD data center processor, every Apple chip, every Amazon Graviton is manufactured by TSMC.
This concentration creates profound geopolitical risk. Taiwan sits 100 miles from mainland China, which claims the island as its territory. A military conflict, blockade, or even a severe earthquake could trigger a global technology crisis. The CHIPS Act's $52 billion is explicitly designed to reduce this concentration.
TSMC's advanced packaging capability, particularly Chip-on-Wafer-on-Substrate (CoWoS), has emerged as an equally critical bottleneck. CoWoS is essential for stacking HBM memory directly on AI accelerators, the configuration that gives NVIDIA's chips their performance advantage. This is why there is so much emphasis in TSMC establishing a factory in Arizona and a shift in attention towards Intel for domestic production.
Embedded tweet: 2005183355990802874Embedded tweet: 2007409626615276013Layer 4: Memory
Embedded tweet: 2009175426636374479Memory has emerged as the critical enabler and constraint of AI scaling. High-Bandwidth Memory (HBM), a specialized DRAM architecture that stacks memory dies vertically and places them adjacent to the processor, provides the bandwidth necessary to feed data to AI accelerators at the rates they require.
The numbers are staggering. NVIDIA's GB200 NVL72 rack contains 13.4 terabytes of HBM compared to 640 gigabytes in the previous-generation DGX H100. The upcoming GB300 increases this to 21.7 TB. AMD's MI400 Helios rack will contain 31.1 TB. AI servers use 34x more HBM content than previous generations.
SK Hynix holds 62% of the HBM market; Samsung holds 17%; Micron holds approximately 21%, making it the only non-Korean supplier. All three have production sold out through the end of 2026.
Micron's position is particularly compelling from a sovereignty perspective. As the only US-headquartered HBM supplier, Micron offers geographic diversification from Korean concentration risk. The company has announced approximately $200 billion in US investments: two fabs in Idaho, up to four in New York, Virginia expansion, and domestic HBM packaging capabilities. CHIPS Act funding of $6.165 billion in grants and $7.5 billion in loans supports this expansion.
The stock performance reflects the transformation: Micron rose 239% in 2025, SanDisk 388%, Western Digital 219%. Memory is undergoing a structural shift from commodity to strategic asset. As Micron's CEO noted:
Memory is now essential to AI's cognitive functions, fundamentally altering its role from a system component to a strategic asset.
Layer 5: Processors
NVIDIA's dominance is well documented: over 94% share of the discrete GPU market, data center revenue of $51.2 billion in fiscal Q3 2026. The moat rests not merely on hardware but on the CUDA software ecosystem, which has accumulated over two decades of developer investment.
The competitive landscape is evolving. AMD's MI350 series is its fastest-ramping product in history. The MI450, launching Q3 2026 on TSMC's 2nm process, targets direct competition with NVIDIA's Blackwell and Rubin architectures. AMD's multi-year partnership with OpenAI, including 1 gigawatt of MI450 deployment in H2 2026, validates the competitive threat.
Custom silicon from hyperscalers represents another vector. Google's TPUs, Amazon's Inferentia and Trainium, Microsoft's Azure Maia are all designed to reduce NVIDIA dependence.
Geopolitics adds complexity. China represented 26% of NVIDIA's revenue in FY2022; export restrictions have reduced this to approximately 13% in 2025. Domestic alternatives like Huawei's Ascend 910C are scaling rapidly.
Embedded tweet: 2007279533150347440Layer 7: Energy Infrastructure
US electricity consumption is growing 2.5% annually after 25 years of stagnation. Goldman Sachs projects a 165% increase in data center power demand by 2030. Interior Secretary Doug Burgum has framed it directly: "The US must win the AI arms race, linking energy security to national security."
Independent power producers with nuclear exposure (Constellation Energy, Vistra, Public Service Enterprise Group) have surged as investors price in the AI data center catalyst. The Westinghouse $80b contract for large-scale reactors signals policy direction.
So...why greenland?
The Trump administration's interest in Greenland is not just saber rattling poorly articulated posturing. The island contains the 8th largest rare earth reserves globally: 36-42 million metric tons of rare earth oxides, second only to China. More critically, Greenland has the largest rare earth reserves of any territory with zero active mines. Its Kvanefjeld deposit is the third largest land-based rare earth deposit on Earth. The territory contains 25 of 34 EU critical raw materials.
The challenge is extraction. Eighty percent of Greenland is covered in ice. Arctic mining is 5-10x more expensive than elsewhere. One expert characterized the concept as
Completely bonkers... might as well mine on the moon.
Yet, the moment is so pressing that development is proceeding. The US Export-Import Bank issued a $120 million letter of interest for Tanbreez. Prediction markets price a 40% probability that the US takes some form of control. Greenland represents the logical endpoint of the sovereignty repricing trend, a frozen landmass valued not for what it is but for what it contains.
The sovereignty trade
The world is returning to its natural state of naked competition for control of physical resources. The postwar rules-based order, in which sovereignty was subsidized by American hegemony and supply chains were optimized for efficiency rather than resilience, is being unwound. Physical possession is becoming the only law.
The sovereignty trade is a structural repricing of physical reality in a world where cognitive production is becoming the primary axis of great power competition. The companies and commodities that form the substrate of intelligence production are not merely AI beneficiaries, they are the physical foundation upon which the future of human-machine intelligence will be built.
Uranium enrichment is now priced at $190/SWU (up from $56 three years ago); copper at record highs with deficits projected through the decade; memory stocks up 200-400% in a single year; gold and silver hitting 45-year simultaneous records as central banks accumulate. Greenland has become a subject of great power competition.
These are not disconnected phenomena. They are manifestations of a single underlying trend: the repricing of sovereignty itself.
Robots harvesting rice in the Shanghai suburbs are a signal. The country that can grow food, generate power, manufacture chips, and train AI models without depending on rivals will dominate the 21st century.
The race for seizing the means of intelligence production has begun.
Image is my favorite piece of AI artwork Théâtre D'opéra Spatial. It was one of the first pieces created using MidJourney that shook artists to their core because it won an human art competition without detection.