By Garvin Jabusch
The defining question of 2025 wasn’t whether artificial intelligence would transform the economy. It was whether intelligence alone would be enough.
Silicon Valley spent the year betting everything on the proposition that whoever builds the most capable AI models wins the future. China made a different wager: that intelligence becomes a commodity, and strategic value accrues to whoever can power it, deploy it, and embed it in physical systems at scale. One side is building “God in a box.” The other is building the infrastructure to make gods useful.
We think China has the better end of this bet. Not because American AI labs aren’t producing remarkable capabilities—they are (Anthropic’s Opus 4.5 and Gemini Pro 3 Thinking are in fact extremely competent and were essential partners in synthesizing my input sources and in developing this narrative). But because the United States is making a leveraged wager on a single form of technology while systematically ceding much of the physical capacity to manufacture, deploy, and benefit from whatever that technology produces. This is not a strategy. It is a hope that intelligence can overcome physics, logistics, and industrial ecosystems. History suggests otherwise.
So what follows is what we learned in 2025 about the forces shaping the next decade of economic competition—and what it means for capital allocation in an era defined by the collision of technological acceleration and planetary constraints. In other words, the view from Next Economics.
The Electric Stack: What Actually Matters
The most important piece of economic analysis published in 2025 was Packy McCormick’s 40,000-word history of what he calls “the Electric Stack“—the four layers of technology that convert electricity into motion and work: lithium-ion batteries, neodymium magnets and electric motors, power electronics (the semiconductors that switch currents thousands of times per second), and embedded compute (the microcontrollers that orchestrate everything in real-time).
These four technologies share something remarkable: a composite cost curve showing 99% decline since 1990. Batteries, magnets, power electronics, and embedded compute have each fallen at roughly 12-15% per year for three decades. This is the foundation on which electric vehicles, drones, robots, grid storage, and, increasingly, the physical infrastructure of AI itself are built. Master this stack, and you can build almost anything the modern economy demands. Cede it, and you’re probably buying the future from someone else.
This is the Jevons Paradox of the Next Economy: as these four technologies become orders of magnitude more efficient to manufacture and deploy, we do not use less of them. Instead, we find thousands of new ways to embed them into the very fabric of existence. Efficiency, in this context, is not a path to conservation—it is the catalyst for a total physical takeover.
China has mastered it. The United States is ceding it.
The numbers are stark. China controls approximately 75% of global lithium-ion battery production (CATL alone commands 38% market share). It manufactures 90% of the world’s neodymium magnets. It dominates solar panels, wind turbines, grid equipment, and the electric vehicles and drones built on these components. In 2024, China added 429 gigawatts of new power capacity—more than a third of the entire US grid (wind and solar power accounted for the vast majority of China’s new capacity, totaling 356.5 GW, about 83% of the total increase). America added 51 gigawatts.
How did this happen? The same pattern repeats across every layer of the stack: American or European scientists invent the core technology, fail to commercialize it at scale, and eventually watch as patient Chinese industrial policy, vertical integration, and relentless manufacturing iteration capture the value.
Lithium-ion batteries were invented by an American (John Goodenough), refined by a Japanese chemist (Akira Yoshino), commercialized by Sony, and then absorbed into the Chinese manufacturing ecosystem when A123 Systems went bankrupt in 2012 and was sold to China’s Wanxiang for $257 million—a fire sale from a $2.4 billion valuation. Today, BYD and CATL, still leveraging but now having far exceeded A123’s IP, dominate global production. The IEA projects marginal reduction in Chinese market share by 2030.
Neodymium magnets were simultaneously discovered by researchers in Japan and at General Motors in 1983. GM subsequently sold its Magnequench division to a consortium connected to Deng Xiaoping’s sons-in-law for $70 million. Today, China controls 90% of global neodymium magnet production and 85-90% of rare earth processing.
The IGBT—the power semiconductor that made modern motor control possible—was invented at General Electric by B. Jayant Baliga. Jack Welch sold GE’s semiconductor division in 1988. Japanese firms commercialized the technology. Now China’s CRRC (which bought the UK’s Dynex in 2008) and BYD are racing to the frontier of silicon carbide power electronics (see The Electric Slide by McCormick, above).
This is not a story of Chinese theft. It is a story of American financialization—of corporations that consistently chose short-term returns over long-term capability, that sold strategic assets for quarterly earnings, that outsourced manufacturing because labor and energy were cheaper elsewhere without understanding that the learning happens on the factory floor.
As Dan Wang put it in his annual letter: “American scientists may be world leaders in dreaming up new ideas. But American manufacturers have been poor at building industries around these ideas. Bell Labs invented the first solar cell in 1957; today, the lab no longer exists while the solar industry moved to Germany and then to China.”
The Electrostate and the Petrostate
While the 20th century was won by those who controlled the flow of oil, the 21st will be won by those who control the intelligence of the grid
Eurasia Group’s 2026 risk report crystallizes the divergence in a single phrase: “Washington is asking the world to buy 20th-century energy while Beijing offers 21st-century infrastructure.”
In 2010, China was arguably the most fossil fuel-dependent major economy on Earth. Today, it is the first “electrostate”—the world’s largest consumer and producer of clean energy, with most of its growth in power capacity coming from renewables. The United States, meanwhile, has cemented its status as the world’s largest petrostate, pumping 13.5 million barrels of oil per day, and having surpassing Saudi Arabia in 2018.
This matters for more than climate. It matters for economic competition.
China’s value proposition to emerging markets is compelling: solar panels and wind turbines that don’t rely on volatile commodity imports, battery storage systems, grid equipment, electric vehicles—all getting cheaper and more capable by the day. Chinese exports of renewable technologies have now surpassed US fossil fuel exports. In the global vehicle market of 85 million units annually, 20 million are now EVs. China provides 10% of that market through exports alone, and Chinese EVs hold a 75% share of EV sales in emerging markets.
The US response has been to retreat further into its fossil fuel position. The “big beautiful bill” phases out tax credits for utility-scale solar and wind while adding restrictions that make battery credits more difficult to claim. This isn’t energy strategy. It’s culture war.
The result, as Eurasia Group frames it, is a “triple bind” for the United States: higher energy costs and slower buildouts domestically; ceding influence in the fastest-growing economies to China internationally; and wagering strategically that intelligence alone wins when the physical stack that powers and deploys intelligence may be equally or more important.
What’s Actually Working in the Energy Transition
It’s worth stepping back from the geopolitical competition to assess what the energy transition has accomplished.
Gregor Macdonald, one of the most rigorous energy analysts writing today, anchors his analysis to 2010—the year utility-scale wind and solar started landing with force, EV adoption became visible as a sustainable trend, and national policy strategies emerged with clarity.
Since then, combined wind and solar has grown from 1.76% to 17% of global power generation. That’s an enormous accomplishment in a short time span, avoiding a very large tranche of emissions growth. Global sales of internal combustion engine vehicles peaked in 2017. China is now at nearly 50% EV market share for passenger and commercial vehicles combined. Global gasoline consumption is flatlining.
The technologies that are working share common characteristics: they’re already on proven cost curves, they’re manufactured at scale, and they don’t require new infrastructure paradigms. Solar panels, wind turbines, batteries, EVs, heat pumps—these are established technologies, no longer experimental.
What’s not working? The hard-to-abate sectors: steel, cement, air travel, industrial heat. Fifteen years into the transition, we’re still at the experimental, proof-of-concept stage for green steel, hydrogen in aviation, and decarbonized industrial processes. Nothing is scaling. Why? Price. The economics don’t yet work.
And hydrogen—despite years of hype about its versatility—remains largely on the drawing board while established tech like heat pumps actually delivers. As Macdonald notes, heat pumps have taken a far bigger stride in lowering natural gas consumption in Europe than any hydrogen application.
This distinction matters for capital allocation. The technologies riding proven cost curves will continue to gain share. The technologies that require new infrastructure and haven’t yet demonstrated economic viability remain speculative, which isn’t to say entirely without promise. Investors should know the difference.
AI in Context: Powerful but Not Sufficient
None of this diminishes the significance of artificial intelligence. The capabilities emerging from frontier labs are real and consequential. Models now reason through complex problems, show their work, and are embedded in coding, research, and knowledge workflows. Ivan Zhao of Notion reports that co-founders have gone from “10x programmers” to “30-40x” by orchestrating multiple AI agents simultaneously. Per METR, AI can now reliably complete tasks that take humans about an hour; extrapolating the current doubling rate, eight-hour autonomous workstreams become plausible by late 2026.
But AI operates under constraints that its most enthusiastic proponents tend to ignore.
To understand why physical capacity is the final bottleneck, one must look at where AI is currently hitting its hardest wall: the complex, messy world of biology. Biology provides the clearest proof that ground truth cannot be ‘hallucinated’—it must be discovered. Bo Wang, a leading researcher in AI and biology, offered a sharp corrective in his year-end reflections: “Biology is not ‘just another modality.’ It’s causal, hierarchical, stochastic, and incomplete in ways that language and vision are not. Scaling mostly gives you sharper correlations without causal structure.” Real progress in biological AI requires perturbation and experimentation—physical intervention in the real world, not just better models.
This pattern extends beyond biology. In manufacturing, you have to actually make things to learn how to make them better. In energy, electrons still have to flow through physical infrastructure. AI accelerates cognition, but it doesn’t escape the domains where ground truth requires physical intervention.
The rate of acceleration varies enormously by domain. Code and language offer fast feedback loops and clean ground truth—hence rapid progress. Biology, healthcare, energy, and manufacturing involve slower feedback, messy ground truth, and capital-intensive physical systems. Progress in these domains will be constrained by the pace of physical experimentation and buildout, regardless of how capable the models become.
The true value of AI in the Next Economy isn’t in generating text, but in System Orchestration. We are moving from a ‘Passive Stack’ to an ‘Active Stack.’ An AI that can manage a city-scale virtual power plant, balancing millions of bidirectional EV charges with fluctuating wind output, is worth more than a ‘God’ that can write a legal brief. This is the ‘Orchestration Layer’—the software that makes the physical stack intelligent. In the Next Economy, the ultimate ‘moat’ is not found in the elegance of an algorithm, but in the intelligence of the physical system that converts an electron into work.
Ultimately, while Silicon Valley builds a God to answer questions, Beijing is building the nervous system for a world that no longer needs to ask permission from the Petrostate.
This is the core error in the “AI solves everything” thesis. AI is a powerful general-purpose accelerant, but it doesn’t transcend the physical systems in which it must be embedded to create value. And those physical systems—the Electric Stack, the manufacturing capacity, the energy infrastructure—are increasingly controlled by China.
The Bet America Is Making
Dan Wang frames the strategic choice clearly: “The US economy is increasingly a highly leveraged bet on deep learning.”
The bet is that whoever achieves artificial general intelligence first gains a “decisive strategic advantage”—a technology sufficient to achieve comprehensive dominance across economic, military, and geopolitical dimensions. Build the superintelligence, and everything else follows.
Wang is skeptical, and so are we.
China is not years behind on AI. DeepSeek and Qwen are doggedly in pursuit, sometimes closer to US frontier models, sometimes further. By virtue of being open-source (or open-weight), Chinese models have found receptive customers overseas—sometimes including American tech companies. If US labs achieve superintelligence, Chinese labs are probably positioned to follow closely. Unless the advantage is decisive immediately, the US won’t have a monopoly on this technology. In reality, the January 2026 breakthroughs from DeepSeek have effectively ended the era of ‘compute-as-a-moat,’ proving that world-class intelligence can be distilled for a fraction of the cost—leaving physical infrastructure as the only remaining barrier to entry.
More fundamentally, capability without deployment is insufficient. Silicon Valley hasn’t demonstrated joined-up thinking for diffusing AI throughout society, which would require extensive regulatory and legal reform, workforce transition, and infrastructure investment. “How else will AI be able to fold doctors and lawyers into its tender mercies?” Wang asks. Beijing, by contrast, has set targets for deploying AI across society and is embedding it into robots, manufacturing lines, and physical infrastructure.
The result: the US may be using AI to produce more PowerPoints and lawsuits while China, by virtue of being the preeminent global manufacturer, scales up production of electronics, drones, and military systems. Intelligence without action is insufficient. And China is betting that controlling the physical stack lets them commoditize intelligence and still win.
Next Economics: Who Captures Value from the Transition
The energy transition is happening. The question is who captures the value.
From a Next Economics perspective—an idealized post-scarcity economy operating within planetary boundaries—the goal isn’t just “more growth” or “winning against China” or “capturing value from AI.” The goal is prosperity that doesn’t undermine its own physical foundations.
Hannah Ritchie‘s framework is useful here. Three statements are almost always simultaneously true: the world is awful, the world is much better, and the world can be much better. High-income, low-energy countries don’t exist—development requires energy abundance. But when that energy comes from combustion, progress comes at environmental cost that eventually reverses the development gains.
Electrification can break this tradeoff. Low-carbon electricity generation plus broad electrification of the global energy system can deliver energy abundance without the combustion externalities. This is why the Electric Stack matters: it’s the physical infrastructure for decoupling development from environmental degradation.
The companies solving systemic risks—climate crisis, resource degradation, disease burdens, inequality—rather than perpetuating them are the companies building the Next Economy. The companies whose products and services are leading the way to a post-scarcity world are building the Next Economy. The investment opportunity isn’t in the incumbents defending legacy systems. It’s in the firms riding the cost curves that are already working: batteries, solar, wind, EVs, heat pumps, grid storage, power electronics.
But here’s the uncomfortable reality: most of those firms are currently Chinese. BYD, CATL, LONGi, Sungrow, Huawei’s solar inverters—the companies manufacturing the Electric Stack at scale are overwhelmingly based in China. American and European firms are attempting to rebuild domestic capacity, but they’re starting from far behind.
This creates a strategic dilemma. The technologies that serve planetary boundaries are increasingly manufactured by a geopolitical rival. Deploying them globally advances the energy transition while deepening dependence on Chinese supply chains. Restricting them slows the transition while fragmenting the global economy and leaves the West at an economic disadvantage (consider how much cheaper it is to manufacture cars and phones with zero marginal cost renewable electricity than with expensive fossil fuel electricity).
While China currently dominates the ‘first mine’—the extraction and refining of virgin lithium, cobalt, and rare earths—the Next Economy provides a second path: Molecular Re-manufacturing. If the Electric Stack is the hardware of the future, the molecules within that hardware are the new permanent capital. Unlike a gallon of gasoline, which is destroyed upon use, a kilogram of lithium is an asset that can be used, recovered, and re-deployed indefinitely. The West’s strategic ‘out’ isn’t just trying to out-dig China; it is in mastering the high-affinity chemistry required to recycle these materials at a lower cost than mining them.
We are seeing the early stages of this shift. Companies like Redwood Materials and Li-Cycle (and emerging European players) are betting that the ‘mine’ of 2030 will be the scrap pile and the spent battery pack. In a Next Economy framework, this is the ultimate expression of decoupling: achieving energy abundance without the need for continuous, destructive extraction. Controlling the Circular Stack—the technology to recover and re-upcycle the Electric Stack—is where the West can build a durable, localized, and geopolitically independent advantage. We are shifting from an era of Resource Extraction to an era of Resource Orchestration—where the winner is not who owns the most molecules, but who loops them most efficiently.
But until the Circular Stack scales beyond Redwood Materials, there is no clean resolution. The US bet on molecules while China bet on electrons. The consequences of that divergence are for now unavoidable.
And yet, even outside of the Circular Stack, China’s dominance isn’t absolute, and the cost curves that built Chinese leadership are available to others willing to ride them. In batteries, Korea’s LG Energy Solution and Samsung SDI remain credible competitors, and emerging American players like Quantumscape Corp are pursuing higher energy-density chemistries. In power electronics, Wolfspeed leads globally in silicon carbide substrates (33.7% market share), even after a difficult restructuring—and the shift to SiC is early enough that leadership remains contestable. Vulcan Elements, backed by $65 million in Series A funding, is building decoupled neodymium magnet manufacturing in Durham, North Carolina, betting that better chemistry and modern manufacturing can undercut Chinese pricing. Impulse Labs is embedding batteries in induction stoves, partnering with appliance OEMs to capture volume in markets where “made in America” carries procurement advantages. Grid-scale battery storage, HVDC transmission, and nuclear are all poised to scale and aren’t exclusively Chinese domains. The pattern across these opportunities: companies riding proven cost curves in segments where Western manufacturing can still establish footholds, often where national security considerations create protected demand or where integration with local infrastructure provides advantages that offset Chinese cost leadership. These are the spaces we’re watching most closely.
What We’re Watching in 2026
Several dynamics bear close attention in the year ahead.
Fleet turnover will remain a binding constraint. Even cheap new technology can’t displace economically viable legacy infrastructure. The 2010 Toyota in someone’s garage still runs. The natural gas plant built in 2015 still produces electricity cheaper than the capital cost of a new solar installation. Displacement happens at the margin—new demand and retiring old capacity—not through wholesale replacement of functioning systems. This is why the transition takes decades, not years. See Macdonald, above.
AI budget scrutiny will intensify. The land-grab phase is ending. Tunguz predicts buying committees and boards will push back on AI spending for the first time, with small language models and open-source alternatives rising as companies seek 10x cost reductions. The commoditization that McCormick and Wang predict is already visible. DeepSeek’s breakthroughs illustrate the Jevons Paradox in real-time: by making frontier-level intelligence 90% cheaper to produce, they haven’t lowered the world’s ‘intelligence budget’—they’ve simply ensured that AI will be integrated into billions of devices that were previously too ‘dumb’ to be economical.
The deployment gap will become more visible. Capability and deployment are different things. The US leads in frontier model development but lags in the infrastructure to power and deploy AI at scale. China adds power capacity at 8x the US rate. This divergence will increasingly constrain American AI (and manufacturing) ambitions.
The infrastructure demands are staggering. Tomasz Tunguz estimates data center buildout will reach 3.5% of US GDP in 2026—a scale of investment comparable to the railroad expansion of the 19th century. But unlike the railroads, this buildout faces bottlenecks the market alone can’t solve: interconnection queues averaging over eight years, aging transmission infrastructure, and community opposition that has killed projects across multiple states. The fastest, cheapest path to new capacity—solar plus batteries, deployable in 18 months—is precisely what current US policy is hobbling.
Physical infrastructure investments will compound. The companies building batteries, power electronics, motors, and grid infrastructure are positioned to benefit regardless of which AI models ultimately dominate. Physical capacity is durable in ways that software leadership often isn’t.
Political dysfunction will impede coherent strategy. The US is in the midst of what Eurasia Group calls a “political revolution whose outcome will remain genuinely indeterminate for years.” Coherent industrial policy is difficult when institutions are in flux. This favors China’s patient, long-term approach over American improvisation.
Closing Thoughts
We called our investment approach “Next Economy Portfolio Theory” because we believe the economy that emerges from the current transition will operate on fundamentally different principles than the one that preceded it. The companies that thrive will be those solving systemic risks rather than creating them—building the infrastructure for prosperity and post-scarcity within planetary boundaries.
The transition is happening. Wind and solar have grown from under 2% to 17% of global power generation in fifteen years. Electric vehicles are approaching 25% of new car sales globally. The cost curves are working. The physics is on our side.
But the transition is also being won, in manufacturing terms, by China. The Electric Stack is largely Chinese. The deployment capacity is increasingly Chinese. The strategic patience is Chinese.
The United States can still compete. It has advantages in innovation, capital markets, and talent attraction that remain formidable. But competing requires recognizing that intelligence alone is insufficient—that physical capacity matters, that manufacturing is where learning happens, and that you can’t outsource your way to strategic advantage.
We’re not optimistic about American policymakers reaching this recognition anytime soon. The culture war over energy, the financialized short-termism of corporate governance, the political dysfunction that makes long-term strategy nearly impossible—these are not problems that resolve quickly.
But here’s what we’ve learned in eighteen years of managing capital toward the Next Economy: policy follows markets more often than it leads them. When solar becomes the cheapest electron, utilities and hyperscalers build solar—regardless of what politicians say about coal. When EVs outperform on cost and capability, consumers buy EVs—regardless of what culture warriors say about truck masculinity. The economics creates facts on the ground that policy eventually accommodates. Here’s my rough mental model that categorizes the 2026 landscape:
| Category | The “Old” Bet (Petrostate) | The “Hype” Bet (Pure Software) | The “Next” Bet (Integrated Stack) |
| Primary Asset | Molecules (Oil/Gas) | Weights/Biases (Models) | Electrons + Hardware + Logic |
| Resource Model | Linear (Extract/Discard) | Virtual (Infinite Scalability) | Circular (Molecular Loops) |
| Bottleneck | Geopolitical Conflict | Compute/Capital | Interconnection & Physical Space |
| Value Capture | Rent-seeking/Extraction | Subscription (SaaS)/Advertising | Efficiency Gains & System Resilience |
This is why capital allocation matters. Every dollar directed toward companies building the Electric Stack, riding the cost curves, solving the systemic risks, greatly improving productivity—that’s a dollar accelerating the transition that policy alone cannot and never will deliver. It’s not charity. These companies are winning because their economics are winning. But it is a choice about what kind of economy we’re building, and who benefits from it.
We founded Green Alpha in 2007 on the conviction that the economy of the future would be built by companies solving problems rather than creating them. Eighteen years later, that conviction has only deepened. The Next Economy isn’t a hope or a projection. It’s under construction, right now, in battery factories and solar fields and grid infrastructure and the physical systems that both make intelligence and make intelligence useful.
The Next Economy is not a destination we are waiting for; it is a system we are currently auditing. Our mandate for 2026 is to move capital away from the friction of the Petrostate and toward the orchestration of the Electrostate. We aren’t betting on a cleaner world—we are betting on a smarter, safer, more resilient one. The physics demands it; the economics ensure it.
The question isn’t whether to participate. The question is whether we participate fast enough—whether capital flows to the solutions at the pace that the risks require. That’s the work. That’s what we’re here for.
Green Alpha is a registered trademark of Green Alpha Advisors, LLC. Green Alpha Investments is a registered trade name of Green Alpha Advisors, LLC. Green Alpha also owns the trademarks to “Next Economy,” “Investing in the Next Economy,” “Investing for the Next Economy,” “Next Economy Portfolio Theory,” and “Next Economics.” Green Alpha Advisors, LLC is an investment advisor registered with the U.S. SEC Registration as an investment advisor does not imply any certain level of skill or training. Nothing in this post should be construed to be individual investment, tax, or other personalized financial advice. Please see additional important disclosures here: https://greenalphaadvisors.com/about-us/legal-disclaimers/
As of the date of this publication, Green Alpha client portfolios hold positions in CATL (Contemporary Amperex Technology Co. Limited), Quantumscape Corp, and Wolfspeed, Inc., both of which are discussed in this commentary. Green Alpha formerly held positions in Li-Cycle, and in A123 Systems. Green Alpha does not currently hold positions in BYD, General Electric, CRRC, LONGi Green Energy, Sungrow Power Supply, Huawei, LG Energy Solution, Samsung SDI, Vulcan Elements, Impulse Labs, Redwood Materials or Toyota.
The securities mentioned are for illustrative purposes only and are not recommendations to buy, sell, or hold any security. The inclusion of a security in this commentary does not constitute a recommendation or endorsement. Past performance is not indicative of future results. Portfolio holdings are subject to change without notice.
This commentary represents the opinions of the author as of the date of publication and is intended for informational purposes only. It should not be construed as personalized investment advice. Investors should consult with their financial advisor before making investment decisions based on the information presented here.
Certain statements in this commentary constitute forward-looking statements. These statements involve known and unknown risks, uncertainties, and other factors that may cause actual results to differ materially from those expressed or implied by such forward-looking statements.