AI, Energy, and Policy: The Convergence of Unintended Consequences

“There are downsides to everything; there are unintended consequences to everything.” – Steve Jobs

Over the last two weeks, Stan Druckenmiller and Paul Tudor Jones, two of the greatest macro managers of all time, have each given interviews questioning the financial situation of the U.S. and critiquing fiscal policies around government spending and monetary policy, especially the Fed’s recent decision to cut rates. Paul was particularly direct in his criticism, stating, “We are going to go broke.”

The title of this paper draws from a Steve Jobs quote, a sentiment I strongly believe, especially regarding government and Fed decisions to manage the economy and markets. Since the Great Financial Crisis, we’ve seen that printing enough money can ultimately turn both the stock market and economy around, regardless of systemic bank failures or even a global pandemic. China, too, is now dipping its toes into the printing pool, but the question remains: will they prioritize short-term gains over long-term stability, risking unintended consequences down the road? Until now, they have been reluctant to choose the U.S. path.

Currently, the U.S. appears to be cruising along: stocks and gold are at all-time highs, Bitcoin just hit record levels, and inflation has steadily declined from a 9.1% peak in 2022 to 2.4%—even lower than the 2.5% rate in January 2020, before COVID-19 hit. The M2 money supply has grown by about $6 trillion since the pandemic began, averaging over $1 trillion per year. Meanwhile, gas prices have dropped nearly 40% from their $5 peak in 2022, and despite recurrent recession warnings, GDP has averaged around 3% over the past eight quarters, with Q3 projections topping 3%, according to the Atlanta Fed’s GDPNow estimate. AI has boosted market enthusiasm, leaving profit margins near all-time highs, with expectations of a productivity boom on the horizon. As of now, government spending and a Fed eager to pursue loose monetary policy to inflate the debt away appear risk-free.

However, Stan and Paul have shared their concerns around a likely rise in inflation, which I expect as well. While AI’s deflationary potential is a hopeful offset, similar to software in the past decade, we must ask: have abrupt fiscal and monetary policies truly created a "Goldilocks" soft landing, or is it all a temporary illusion? I side with Jobs’ perspective, continually searching for areas where things could unravel in the coming years. Yes, a market correction sometime soon is likely, and based on patterns from the 1970s, it may begin in the first half of next year if inflation rises and the Fed is forced to reverse rate cut expectations in the market.

However, my growing concern is the convergence of rapid fiscal and monetary expansion during the pandemic, which has helped fuel the turbocharged rise of AI through massive investment, combined with energy transition policies. Recent advancements in AI promise productivity gains and scientific breakthroughs in the future but also bring concerns over soaring energy demands. By 2027, analysts project that AI servers may consume as much energy as entire countries like Argentina or Sweden.

On the energy transition side, experts like Daniel Yergin are skeptical of the feasibility of current energy transition policies due to their unprecedented speed and reliance on mandates rather than organic economic shifts. Unlike historical energy transitions spanning a century, today’s push to replace hydrocarbons with renewables within a few decades is fraught with challenges. Yergin stresses that comprehensive macroeconomic planning is essential to avoid supply shocks similar to those of the 1970s.

Tech leaders are also vocal about this convergence. Former Google CEO Eric Schmidt doubts that current efforts will achieve climate goals, particularly the U.S. target of net-zero emissions by 2050, due to a lack of organization and coordination. At the AI+Energy Summit, he argued that instead of relying on these targets, we should invest more in AI and data centers, which he sees as critical to solving energy challenges—albeit with trade-offs. Schmidt warned that AI’s escalating energy demand, especially from data centers, necessitates a robust power supply, or the U.S. may face energy shortages within a few years. More recently, in an interview with Tali Shine in Australia, he got more specific, referencing model simulations and saying that, at the current rate of AI advancement, America will run out of power by 2028! Despite AI’s heavy energy consumption, Schmidt believes that AI development is essential for addressing environmental issues, as conservation alone won’t suffice.

Elon Musk highlighted the exponential increase in AI computing power, roughly 10x every six months—a rate that will soon face physical limits. Comparing the demand for chips to a historic gold rush, Musk noted an emerging strain on electrical resources, humorously saying we need “transformers to run transformers.” He went on to say, “the next shortage will be electricity. They won't be able to find enough electricity to run all the chips. I think next year, you'll see they just can't find enough electricity to run all the chips.” Sam Altman added that a breakthrough in energy production is critical, as AI’s future will consume far more power than anticipated.

If you don’t believe these dire forecasts, actions speak louder than words. Recently, tech giants have made historic news as they turned to nuclear power as a stable, substantial source of energy needed to support their AI ambitions. Did anyone expect to see a reopening of Three Mile Island at the beginning of the year? When Microsoft, Google, and Amazon start investing in nuclear energy, it signals their understanding that solar and wind alone may not suffice. Needless to say, but I can’t imagine ramping up nuclear will go smoothly from a policy perspective.

Meanwhile, another consequence of the energy transition policy unfolds: government regulations and ESG pressures have actively discouraged investment in fossil fuel infrastructure in recent years. Exxon’s recent Global Outlook report underscores this growing risk, estimating that global oil production could decline by 15% per year without continued investment—nearly double previous estimates from the International Energy Agency. Without new investments, global oil supplies could drop by over 15 million barrels per day in just one year. The report says, “At that rate, by 2030, oil supplies would fall from 100 million barrels per day to less than 30 million – that’s 70 million barrels short of what’s needed to meet demand every day.” This creates a fascinating paradox: while fiscal policies accelerate innovation requiring increased power, environmental policies restrict traditional energy investments.

And don’t forget, the green transition itself is energy-intensive. Building solar panels, wind turbines, and electric vehicle batteries requires substantial energy. The International Energy Agency estimates that meeting the Paris Agreement’s goals will demand a massive increase in critical minerals, the mining and processing of which also require significant energy.

Three forces converge:

Accelerated AI development demanding enormous power

Reduced investment in traditional energy infrastructure reducing future supply

Increased energy needs for the green transition

This creates what you can call a “power paradox.” The very policies intended to foster innovation and combat climate change may inadvertently trigger an energy shortage that hampers both goals. As energy expert Vaclav Smil says, “Energy is the only universal currency: one of its many forms must be transformed to another in order for stars to shine, planets to rotate, plants to grow, and civilizations to evolve.”

The timing makes this especially concerning. The rapid advancement of AI, particularly with AI agents, suggests that power demand may escalate even faster than projections indicate. Microsoft's president, Brad Smith, recently acknowledged that their “carbon moonshot” goals are now five times harder to achieve due to AI’s energy demands.

Consider this: data centers currently consume 1-2% of global electricity. By 2030, this could rise to 3-4%. While this may sound minimal, it represents a demand increase equivalent to the current electricity consumption of Germany—and these estimates might be conservative if AI adoption continues to accelerate.

The implications are not merely theoretical. Strains are already evident, as seen in Virginia’s “Data Center Alley,” where power capacity is limited. Ireland has implemented a de facto moratorium on new data centers, which already consume nearly a fifth of the country’s electricity.

This scenario reminds me of my “Adapt or Die” framework, stressing the need to unlearn outdated assumptions during this world of exponential innovation. Current policies are based on linear projections that were made before the money printing during the pandemic and before the release of ChatGPT. We’re entering a sooner and faster-than-expected exponential phase in technological development with AI, and the green transition will likely demand far more energy than anticipated by linear models. Hopefully, AI can help solve some of these demand issues sooner than expected as well because I can’t see another solution.

Ironically, efforts to continue to stimulate growth to inflate our way out of debt, as Paul Tudor Jones said, will continue to power growth of innovation while addressing climate change and may create conditions that impede both objectives. This classic example of complex systems producing unexpected outcomes illustrates the risk of multiple policy interventions intersecting.

The path forward demands a delicate balance between fostering innovation, ensuring energy security, and meeting environmental goals. This challenge will test our ability to adapt and innovate in unexpected ways. As always, those who can see the bigger picture and adapt quickly will be best positioned to thrive.

In terms of signals in the market that we have reached a danger point, it has to show up in the price of crude oil and copper, in my opinion. Although most of the forecasts for demand and supply issues are likely a few years away, the race for AI dominance is a global competition and, as we see, energy security remains a constant problem. As long as oil and copper prices remain contained, the market is telling you this unintended consequence is not something to worry about yet. I believe we will be seeing both of them move higher next year as the global rate cuts and China stimulus increase demand for goods and as fears of potential AI bottlenecks become more obvious. Until then, I think markets will be able to focus on “Goldilocks. “ In your portfolios though, I highly recommend having some energy and copper positions as a hedge to “Goldilocks” especially when bonds are not helping.

As Steve Jobs said in his Stanford Commencement speech, “It is incredibly easy to fill in the unknown areas with false facts. In fact, science has proved that your brain will automatically make connections and beliefs to help fill in your knowledge gaps. Accept and then remind yourself that there are certain truths you won't know until later. Otherwise, you aren't just making decisions on bad information, but are less likely to recognize good information when you finally come across it.”  I think the government and Fed should listen to his speech and listen to the leaders in the industries of energy and technology warning them so that they can hear the truths now rather than later.