Since August 8th, NVIDIA has underperformed the S&P 500 by nearly 10%, its sharpest relative decline since the Liberation Day fallout. Other parts of the AI ecosystem have slipped as well, feeding a new steady drumbeat of “AI anxiety.” I know this has happened because I get the calls and emails. The peak in NVIDIA’s performance versus the index came on August 8th, the day after ChatGPT-5 was released to largely negative reviews. GPT5 is amazing for what it is worth. That was quickly followed by an MIT report downplaying AI’s corporate value and Sam Altman cautioning that AI might be in a bubble. Because so few people truly use or understand AI, many still think of it as just another technology, a product cycle where companies are simply part of a supply chain, the way autos depend on steel, rubber, and parts. That perspective naturally breeds fear: if AI slows, then every link in the chain is at risk. But AI is not a technology in that sense. It is more like electricity, a general-purpose force that spreads everywhere, invisible yet indispensable. As Andrew Ng put it, “AI is the new electricity. Electricity transformed industries: agriculture, transportation, communication, manufacturing.” The same is happening now, only faster and on a broader canvas. AI cuts costs, sparks creativity, and reshapes management all at once. It is not a tool in one sector but a hybrid solution reshaping every sector.
The essence of this hybrid character is that AI collapses boundaries that once separated distinct roles. In some cases, it acts like automation, driving down costs by replacing repetitive tasks. In others, it serves as a creative partner, enabling new products, services, and even entire business models. It also functions as a managerial tool, scanning vast datasets to provide early-warning signals and guide strategic decisions. In this way, AI mirrors the brain itself, a biological technology that is inherently hybrid, capable of mechanical, creative, and managerial functions all at once. No previous innovation has combined these dimensions in a single force. That is why AI cannot be confined to one sector; it operates as a horizontal layer of intelligence that cuts across them all, creating compounding effects that only become visible when the system is viewed as a whole. And just as machines only delivered efficiency once they were plugged into electricity, AI only realizes its potential when plugged into the vast compute and data infrastructure that fuels it.
As I have been writing about the power needs for AI, the questions in response have led to this paper. The irony is that the infrastructure required to unlock AI’s potential reflects the same hybrid character. AI does not run on a single, uniform power source any more than it belongs to a single industry. Instead, its energy demands span fossil fuels for baseline loads, renewables for intermittent surges, nuclear for long-term stability, and storage technologies to smooth the volatility in between. No one source is sufficient on its own, just as no one sector fully contains AI’s impact. The grid itself is becoming hybrid to support a hybrid technology. This creates a mirror image: AI, the “new electricity,” is entirely dependent on a patchwork of old and new electricity sources, each stitched together to deliver the resilience and scale it needs. In effect, AI does not just demand more power, it demands a more complex, integrated power system, one that mirrors its own hybrid nature.
In the age of oil, the energy for transportation equation was comparatively simple. Crude could be refined into gasoline for cars, diesel for trucks and ships, and kerosene-based jet fuel for airlines. A single feedstock was transformed into multiple outputs, each matched to its use case. AI’s energy story is far more complicated. Instead of refining one resource into different fuels, the challenge is the reverse: combining multiple resources, natural gas, renewables, nuclear, and storage into one consistent stream of electricity. That electricity, in turn, becomes the lifeblood of AI’s data centers, where semiconductors convert raw power into compute. Unlike oil, where conversion was linear and predictable, AI’s hybrid power system requires constant balancing, integration, and optimization. And because it depends on multiple sources stitched together, the risks of supply chain disruption multiply: a shortfall in uranium, copper, battery materials, or natural gas can ripple through the entire system, threatening the stability of the unified power output. Hybrid grids are just the first example, the entire world of AI runs on hybrid solutions.
Elon Musk’s four Tesla Master Plans make clear that he has always viewed technology through this cross-silo, hybrid lens. The first began with cars, but each successive plan expanded outward into batteries, solar power, autonomous driving, and now humanoid robotics. Just like AI is difficult for investors or analysts, Tesla is not a car company. This is why I always enjoyed watching the transformation of Adam Jonas at Morgan Stanley into a hybrid auto analyst. What looks like separate industries on Wall Street’s research desks are, in Musk’s framework, a single integrated system built on energy and intelligence. As he has noted, “AI is going to have a bigger impact than almost anything in history. It will change every single industry.” His roadmap shows how: electricity, mobility, storage, and AI are not parallel bets but interlocking components of the same future. In his own words, “The AI scaling constraint will move from chips to voltage transformers to electricity generation,” a reminder that the hybrid nature of AI’s demands extends well beyond technology into the infrastructure that underpins the entire economy.
For investors, this hybrid architecture creates analytical challenges that mirror AI itself. Oil had the virtue of simplicity, energy companies, refiners, and transportation firms could be modeled within clear boundaries, and shocks to supply were broadly understood. With AI, the story runs in the opposite direction. The need to weave together natural gas, renewables, nuclear, and storage into one output means that disruption in any of these inputs can cascade into the performance of technology companies. Yet Wall Street remains structurally siloed: the utilities analyst looks only at grid operators, the tech analyst at semiconductors, the industrials analyst at robotics, and the energy analyst at natural gas or uranium. No one “owns” the cross-cutting risk of AI’s hybrid power demand, just as no one fully owns the cross-cutting story of AI’s hybrid efficiency. As Jensen Huang put it, “AI is fundamental infrastructure, like electricity,” but the very fact that it sits everywhere means that it belongs to no single coverage model on the street.
The hybrid nature of AI’s supply chain means that doubt in one corner of the story can ripple across the entire ecosystem. In traditional sectors, supply chains were more linear: weakness in autos might not affect airlines, and oil demand shocks could be absorbed differently by diesel versus gasoline. But in AI, the dependencies run across silos. A pause in data center capex, like we saw with DeepSeek’s disruption to sentiment, can send shockwaves from semiconductors to utilities to industrial equipment suppliers. Suddenly, NVIDIA’s order book or a hyperscaler’s quarterly capex number becomes a proxy for the health of the whole AI buildout. Because every link in the chain from GPUs to copper to power infrastructure is feeding into the same output of electricity-driven compute, skepticism about AI demand tightens correlations across sectors that would normally diversify each other. What emerges is not just a technology cycle, but a hybrid risk cycle, where doubt in one input threatens the valuation of the entire AI universe. And just as in August investors misread NVIDIA’s underperformance as the end of an AI product cycle, markets risk misreading systemic hybrid risks as isolated sector issues.
This is especially frustrating for investors because the search for idiosyncratic risk, the stock or theme that escapes broad factor exposure, is central to the craft of active management. A clean stock pick that is insulated from macro swings or sector cycles is the holy grail. But AI refuses to cooperate. It is not a technology subsector that can be carved out and modeled separately; it is a factor in its own right, like interest rates or liquidity. The hybrid supply chain means that correlations rise across semiconductors, power producers, reits, commodities, and even industrials, leaving little room for diversification. An investor may think they are playing a utility capex story or a copper cycle, but in reality, they are holding AI beta. As Marc Andreessen observed, “AI is quite possibly the most important and best thing our civilization has ever created, certainly on par with electricity and microchips.” In this way, the effort to isolate idiosyncratic risk collapses into exposure to the AI factor, whether or not it sits in the “technology” bucket.
This makes risk management more challenging than in past cycles. By definition, factors are broad and hard to hedge, but AI’s reach makes it especially pervasive. As it migrates from a technology story into the trend driving nearly every corner of the economy, traditional diversification tools lose their effectiveness. Utilities, semiconductors, cloud providers, miners, and industrials now share the same AI demand driver. What once would have been offsetting exposures become correlated risks. For risk managers, this creates a dilemma: trimming AI exposure means reducing the very growth engine propelling markets, yet embracing it means accepting a level of systemic concentration that is uncomfortable by historical standards. AI is not just a theme it has become the factor around which risk management itself must be rethought.
The implication is clear: investors cannot rely on the old playbook. When AI functions as both a technology driver and a systemic factor, risk cannot be managed within traditional sector silos. What is required instead is a hybrid investment lens one that mirrors the hybrid nature of AI and its energy backbone. This means analyzing semiconductors alongside utilities, following copper and uranium supply chains as closely as cloud revenues, and treating hyperscaler capex with the same weight as monetary policy. It also means embracing polymathic thinking, where insights from technology, energy, commodities, and macroeconomics are combined into a single framework. Just as AI itself collapses boundaries between efficiency, creativity, and management, investors must collapse boundaries in their analysis.
Artificial intelligence began this journey framed as “the new electricity,” in Andrew Ng’s words, and the analogy has proven deeper than many realized. Like electricity, AI is not confined to one industry but diffuses across every sector, reshaping production, consumption, and organization itself. Unlike electricity, however, AI’s hybrid efficiency and hybrid power demands create a tangle of correlations that investors struggle to navigate. What looks like diversification in old models becomes concentration under the new factor of AI. This makes risk management more complex, but it also makes the opportunity larger. The challenge for investors is not simply to find winners within technology, energy, or industrials, but to adopt a hybrid lens that sees across them all. To miss this is to miss the system itself. In that sense, AI is not just the new electricity, it is the hybrid solution redefining investment weather, the background condition shaping everything else. Hybrid solutions are not the exception they are the defining condition of the AI age.
