Geology, a cornerstone of the oil and gas industry, is emerging as the subsurface code powering the current AI revolution. Beneath this code lies an immutable physical constraint: energy, a resource with an unavoidable geological foundation that this article explores.
Google exemplifies this approach by aiming to couple natural gas with carbon capture and storage (CCS) to generate low-carbon power for data centers. In such a configuration, geology serves a dual purpose: providing the energy resource while supporting the intended permanent storage of the CO2 generated from power production. Additionally, Fervo Energy’s 2026 Nasdaq debut at a multibillion-dollar valuation elevates geothermal as a scalable pillar of the subsurface code, building on its 2023 Project Red pilot with Google.
Similarly, Meta's 2025 agreement with XGS Energy underscores the growing demand for firm baseload power to support data centers and expanding AI infrastructure. Finally, Deep Fission's recent Nasdaq debut and advancement of nuclear technology in downhole environments that follow oil and gas principles suggest that the subsurface code for data centers extends beyond natural gas, carbon management, and geothermal.
The AI-Energy-Carbon Management thesis holds that the AI revolution is collapsing the historical boundaries between AI, energy, and carbon management into a unified industrial architecture bridging the digital and physical worlds. This thesis was coined by the author while co-moderating a panel featuring NVIDIA, XGS Energy, Microsoft, SLB, Fugro, and Occidental at the 2025 SPE Energy Transition Symposium.
Frontier Large Language Models and Carbon Management
Oracle further illustrates the subsurface dependency by securing the right to acquire a stake in Bloom Energy while deploying its fuel cells and, in the process, circumventing power-grid backlogs to generate lower-emissions, onsite power from natural gas for additional data center capacity. Together, these investments align with NVIDIA CEO Jensen Huang's description of the AI ecosystem as a five-layer stack, with energy serving as its foundational layer powering the expanding AI economy.
The AI-Energy-Carbon Management thesis extends across the distinct strategies adopted by the frontier large language model (LLM) ecosystem, including Anthropic (Claude), Google (Gemini), and OpenAI (ChatGPT). Frontier AI companies are increasingly investing in carbon management to address the emissions associated with rapidly expanding computational demand.
Case in point: Anthropic has officially joined Frontier as a buyer within a $915 million expansion of the advance market commitment for permanent carbon removal. Alongside Google, Stripe, Shopify, and other participating companies, Anthropic is helping create demand for engineered carbon-removal solutions, including direct air capture (DAC). In parallel, Occidental is advancing DAC toward commercial operation by constructing the world's largest facility of its kind, illustrating how legacy oil and gas operators are scaling permanent carbon removal, while AI companies are emerging as a new class of participants in the carbon-management domain.
Google's participation in Frontier reinforces this strategy. Beyond operating Gemini, the company simultaneously advances subsurface carbon management through its natural gas coupled with CCS strategy while supporting engineered carbon removal through Frontier. Meanwhile, OpenAI’s primary cloud partner, Microsoft, is a major corporate purchaser of carbon dioxide removal, helping accelerate the engineered carbon removal market through commitments that include DAC.
The Convergence of AI and Energy's Industrial Ecosystem
OpenAI is advancing large-scale AI infrastructure through initiatives such as Stargate, alongside CEO Sam Altman's strategic investments in advanced nuclear fission and fusion ventures, including Oklo and Helion Energy. Deep Fission exemplifies this convergence by adapting oilfield drilling techniques to deploy nuclear reactors in downhole environments.
This emerging approach repurposes decades of subsurface engineering expertise for AI-scale, low-carbon baseload power. Unlike geothermal or natural gas, this approach is largely independent of geological resource extraction, transforming the subsurface into an engineered platform for energy generation. Consequently, the commercial implications of a dry hole are fundamentally different, as project value is no longer principally determined by the discovery of an economically productive subsurface resource. Instead, deployment is equally anchored to data center demand, representing a commercial market for Deep Fission’s downhole nuclear technology.
Collectively, these developments illustrate a broader industrial convergence. To power the next generation of frontier LLMs and the broader AI economy, the fast-moving digital world is increasingly integrating with heavy industry. While AI represents capabilities accumulated across a relatively young technology sector, the energy industry represents capabilities refined over centuries of industrial development, with carbon management building upon that foundation. Together, these developments establish a unified industrial architecture in which digital scaling is fundamentally anchored in physical and geological reality, thereby defining the AI-Energy-Carbon Management thesis.
Capital Allocation and the New Economics of Carbon Management
Since the emergence of frontier LLMs following the release of ChatGPT in 2022, capital allocation in the global economy has grown increasingly sensitive not just to the price of a barrel of oil, but to the energy required behind a digital prompt. As the economics of a barrel and the economics of compute converge, carbon management, historically viewed as a legacy oilfield cost center, is being repositioned as an area of strategic capital allocation for the AI economy, driving a transformation as important as the buildout of energy infrastructure.
Sleipner's offshore CO2 storage project, operated by Equinor since 1996 with early ExxonMobil participation, remains the baseline precedent for this transformation. Today, Microsoft underscores this model by acting as a digital partner within the evolving carbon market. Working with Northern Lights (a venture that includes Equinor) and other partners, Microsoft supports the digitalization of the subsurface CCS value chain while helping build the digital and commercial infrastructure required to scale geological storage. In parallel, Northern Lights demonstrates how geological storage is evolving from what was historically viewed as an oilfield cost center into an emerging investment opportunity (as outlined herein). This evolution has the potential to redistribute project risk while broadening institutional investment profiles.
Concurrently, ExxonMobil is interfacing directly with hyperscale data centers, partnering with NextEra Energy on a proposed 1.2-gigawatt natural gas power plant located near its CO2 pipeline infrastructure and designed to capture and sequester emissions, building on the company’s early CCS experience dating back to 1996.
Conclusion
Viewed collectively, these investments and market movements represent the microeconomic signals of a fundamental macroeconomic shift. In the era following the emergence of frontier LLMs and the AI economy, computational scaling triggered an unprecedented physical power race, forcing hyperscale digital technology and heavy industry into structural convergence. The defining strategic question of this century is no longer which enterprise will train the most capable AI model, but which will construct the integrated digital-physical platform capable of unifying compute, energy, geology, carbon management, and their supporting infrastructure into a single, vertically integrated ecosystem.
The race is underway to build Rockefeller 2.0: the first enterprise capable of this integration. The enterprise that achieves this synthesis will not merely power the future of AI but will help define the industrial architecture of the modern economy.
