The challenge of decarbonizing the energy that supports data centers is now one of the most urgent and complex systemic puzzles facing the modern age. Two recently read works (Jean-Baptiste Fressoz's More and More and More and Bill Gates's How to Avoid a Climate Disaster) offer contrasting and complementary frameworks for understanding and addressing the energy dilemmas of our era. Fressoz's book deploys a historical and critical lens (all too easy and predictable blasting of past predictions), challenging prevailing narratives about energy transitions and warning of the persistent accumulation and intertwining of energy sources. Gates, in contrast, offers solution-focused roadmaps, advocating for bold technological innovation, systemic transformation, and aggressive emissions reduction, with its erudite and arcane approaches aimed toward those who may be more in a role to do something with this advice.
With AI assisting with the coherence and analysis of the core arguments, lay readers can begin to apply this information to those supporting data centers, synthesizing their perspectives with up-to-date empirical evidence, market trends, and emerging innovations. Where instructive, tables and summary lists are integrated.
Fressoz’s thesis contests the received wisdom that energy transitions represent a clean substitution of old forms of energy by newer, cleaner alternatives in historical succession and instead asserts that historic survey reveals not replacement but mutual accumulation, bludgeoning the reader with this notion quite thoroughly by the conclusion. Wood was not replaced by coal; coal was not replaced by oil; oil was not replaced by nuclear or renewables. Instead, introductions of new energy infrastructures stimulated greater extraction and consumption of existing fuels. For example, coal’s ascent demanded enormous amounts of timber for mine supports and rails, while modern oil and gas extraction and transport remain dependent on coal-intensive steel, concrete, and plastics production.
This thesis reorganizes how the history of energy should be told. Instead of tidy eras divided by fuel type, we face a web of entangled, mutually reinforcing resource streams. Fressoz’s evidence is insurmountable, ranging from 19th-century mine operations to contemporary global raw material flows, and undermines the optimism of “transitions,” revealing that new energy systems often deepen total resource consumption rather than displace older, more polluting forms.
Fressoz also devotes a good number of words to scrutinize the intellectual and political roots of the energy transition narrative. He argues that the term “energy transition” originated more as a promotional or ideological construct than as an empirically grounded concept. Industrialists seized on it to pitch new products (such as atomic power), and policy thinkers and climate officials followed suit, projecting an imminent changeover to renewables. In reality, he warns, this faith in innovation as panacea, embodied in “transition” rhetoric, has often delayed serious action on resource consumption and emissions, hence the coupling with the second book that relies so heavily on this narrative. Institutions championing this approach, Fressoz contends, have “procrastinated,” eagerly prioritizing hypothetical technological fixes while deferring the hard questions of reducing material throughput and confronting economic growth imperatives. It is not to point out that Gates ignores this, however, placing such weighted emphasis may be inappropriate given the haste at which we may need to apply solutions.
A key insight of Fressoz’s work is the symbiosis of the physical infrastructure that props up energy production, all driven home with extensive numbers. Solar panels require silicon, rare earths, steel, and glass; wind turbines are feats of steel and concrete engineering; electric vehicles rely on mining and petrochemicals. Technologies conceived as substitutes end up tightly integrated into, and often expanding, existing extractive systems. Fressoz’s analysis underscores that decarbonizing one aspect of energy use, such as electricity generation, leaves untouched or even exacerbates emissions from cement, steel, fertilizer, and plastics, which underpin the build-out of even the greenest infrastructures. This again undermines the intent to leverage the ambiguous technology approach as a definitive solution.
Fressoz’s perspective suggests that as data centers electrify and adopt renewables, they do not simply phase out fossil fuels, especially given the need for reliable service. The construction of each facility requires more mining, refining, and fossil energy throughout the global commodity web, and the offsets of the data center betterment to the environment has yet to be calculated. The accumulation models seen thus far predict that data center energy use, even as becomes cleaner, will amplify total resource throughput unless paired with broader strategies for absolute reductions to reach the label of “sufficiency” and systemic reconsiderations of economic growth and digital expansion.
Gates’s book is an engineer’s manifesto for systemic climate action, taking in vast amounts of information to lay out the most crucial metrics and then working toward explanations of solutions. He anchors his argument in a stark quantitative challenge: the world must reduce annual greenhouse gas emissions from 51 billion tons to net zero, ideally by 2050, to avert climate catastrophe. Crucially, “net zero” means not just reductions, but balancing any residual emissions with removal, whether by land management, carbon capture, or direct-air processes.
Gates is optimistic (and gives compelling reasons why) and breaks down the targets to be across five broad sectors: how we plug in (electricity); make things (manufacturing); grow things (agriculture); get around (transportation); and heat or cool spaces (buildings). He is clear-eyed about the scale and difficulty of this “all-of-the-above” effort, so it is not without pause or conjecture about the scale and thankfully uses comparisons throughout for easier measure. There is no single silver bullet: every category of emissions must be addressed, and often through difficult, expensive, or unfamiliar routes.
A central item in the book is the “Green Premium” which is the extra cost of decarbonized commodities, technologies, and anything made or consumed compared to their polluting counterparts. Gates’s approach is pragmatic: instead of wishing for consumers to absorb Green Premiums, he calls for policy, investment, and innovation to drive these costs down through economies of scale, R&D, and smarter regulation. In closer review of policy implementation, there is likely to be organized resistance to changes affecting business and companies and a likely lesser amount of push back from consumers as a whole, meaning policies might affect the individuals much more than those that use large scale quantities of, really, anything.
Gates lays out that “doubling down on innovation,” massive investment, and coordinated policy can shift the system slowly but surely in the right direction, leaving aside the measure of rubber-banding back to previous ways or making incorrect shifts that negate benefits. He highlights priorities such as:
Gates emphasizes both the importance of government action (standards, carbon pricing, procurement, incentives) and the catalytic role of private capital (such as his Breakthrough Energy coalition) in commercializing new technologies. With the numbers given, the carrot to be chased seems immense for any energy hungry rabbit to ignore, should the incentives line up properly.
For Gates, data centers are a frontier for decarbonization challenges. They concentrate massive, growing electrical loads, require near-constant reliability, and often demand growth in regions where clean power is scarce or grid integration is complex. The task is to power an accelerating AI-driven digital economy with zero emissions, rapidly and affordably. Gates sees this as emblematic of the broader energy "transition" (at the peril of not recognizing that transition is mostly a fallacy per Fressoz), with high technical barriers but rich potential for transformative innovation and learning by doing.
Historically, data center energy consumption was stabilized through extraordinary gains in IT efficiency. Between 2010 and 2018, global data center energy use remained relatively flat, even as server deployment exploded, due to improvements in virtualization, cooling, and hardware design. Over the last decade we've seen surging demand for AI, machine learning, blockchain, and cloud computing that has led to rapid expansion in both the number and size of data centers, as well as the computational intensity per facility. Recent U.S. Department of Energy (DOE) and Lawrence Berkeley National Laboratory (LBNL) reports project that U.S. data centers alone consumed 176 TWh in 2023 (about 4.4% of U.S. electricity), a load expected to double or triple by 2028 to 2030, reaching upwards of 325‒580 TWh per year, or up to 12% of total national consumption if the power generation and transmission will accommodate it.
The drivers of this explosive demand primarily include:
Notably, the energy demand of commercial computing, including data centers, is projected by the U.S. Energy Information Administration to become the largest single use in the commercial sector, surpassing space cooling and lighting by 2050.
Similar to Fressoz’s thesis, each technological leap, from compute, IoT, cloud, blockchain, to AI, has led not to the replacement of earlier digital systems but to their coexistence and accumulation. As even since the last millennium, older data centers remain online as newer, denser facilities are built. The resource intensity of constructing and maintaining these centers (from steel to silicon wafers) only accelerates overall energy flows.
The advent of AI training and inference workloads has meant new efficiency curves needed to be created. While chips and servers continue to improve in performance per watt, the sheer quantity of computation overwhelms many efficiency gains. New AI training clusters and language models demand hundreds of megawatts of dedicated capacity per facility or campus. Each new model and service shifts the baseline upward.
Major jurisdictions (EU, certain U.S. states, China’s national government) have moved to stricter rules:
Robust incentives exist for:
| Theme/Concept | More and More and More (Fressoz) | How to Avoid a Climate Disaster (Gates) | Relevance to Data Center Energy Use |
|---|---|---|---|
| Energy “Transition” | Seen as a myth; history of accumulation not substitution. | Frames transition as urgent and doable with innovation and policy. | Warns against complacency or exclusive reliance on new tech to decarbonize data centers; about total resource use. |
| Technological Solutionism | Critical of blind faith in future innovation; “innovation as procrastination.” | Optimistic, innovation is central (R&D, deployment, investment). | Both support efficiency but differ: Fressoz urges scaling back, Gates focuses on making decarbonization affordable/practical. |
| Material Interdependence | New energy systems depend on old (steel, plastic, chemicals) and accumulate resource use. | Emphasizes need to decarbonize all sectors (cement, steel, etc.) for net zero. | Data center buildout and “green” supply must address embedded emissions, not just operational ones. |
| Green Premium | Not primary focus; skeptical that lower cost alone will decouple growth from emissions. | Core concept, must be reduced to scale green solutions equitably. | Cost premium for renewables, storage, and low-carbon materials must be brought down for sustainable data center growth. |
| Policy Role | Critiques procrastinating policies, urges sufficiency and degrowth. | Calls for strong policy: standards, incentives, public/private R&D. | Policy needed to mandate and accelerate sustainable data center operations. |
| Absolute Reductions vs. Efficiency | Calls for sufficiency, degrowth, not just efficiency. | Targets absolute net-zero, but within growth paradigm. | Data center sector must address rapid demand growth as well as decarbonization—efficiency is not enough alone. |
Analysis: Both authors agree that decarbonizing energy for data centers require massive effort, but their worldviews differ. Fressoz warns against mistaking “energy transitions” for meaningful decoupling from fossil resource use. Gates, by contrast, insists that innovation, investment, and scale can bend the arc to net zero, if “green premiums” can be closed. The relevance for data center power systems is profound: genuine sustainability will demand both technological and material transformation, but also a reckoning with runaway demand and “scope 3” supply chain impacts.
Both books, when applied to the energy sector underpinning data centers, expose central tensions:
In pragmatic terms, the pathway to a sustainable energy paradigm for data centers in the next 25 years will require:
Ultimately, the most resilient, cost-effective, and sustainable future for data center energy will be forged at the intersection of the hard lessons of history (Fressoz) and a pragmatic optimism rooted in innovation and policy (Gates). The next 25 years will test whether our global appetite for “more and more and more” can be balanced by a planetary strategy that offers enough, efficiently, equitably, and within the boundary conditions of our environment.
