Silicon Boundaries

Silicon Boundaries

Understanding how much AI society can productively absorb, and how to raise that amount over time

Silicon Boundaries

Silicon Boundaries is a joint initiative of the Oxford Sustainable Finance Group at the University of Oxford and Theia Finance Labs.

Societies, like ecosystems, have conditions they need to function. They need a certain level of trust between people. They need institutions that work. They need economic systems that allow broad participation. They need people to be healthy enough, both physically and mentally, to engage in civic and economic life. These are not luxuries. They are preconditions for everything else, including the financial returns that investors depend on.

Silicon Boundaries are the thresholds beyond which the benefits of AI and related technologies are outweighed by their contribution to societal instability. Together they define a safe operating space. When the use of AI, in its scale or its nature, outruns what society can absorb, it begins to erode these preconditions and pushes society towards instability. Crossing these thresholds can trigger non-linear change and, in some cases, irreversible tipping points. Using too little AI has costs too, in forgone benefits and in stranded capital built for greater use.

A chart of the use of AI in society over time against a shaded safe operating space between two dashed boundaries. Where the path rises above the space it is marked as societal risk; where it falls below, as missed opportunity and transition risk.
Read the Report

Similarities with Planetary Boundaries

Silicon Boundaries borrow the idea of a safe operating space from the Planetary Boundaries of Earth-system science.

But one of the crucial differences is that the thresholds of Silicon Boundaries are not fixed. With climate, for example, you have relatively clear physical thresholds: 1.5 degrees, 2 degrees, the point at which ice sheets become unstable. The physics does not negotiate.

With Silicon Boundaries, the thresholds depend on context. They depend on how the technology is deployed, what safeguards are in place, how society responds, and what alternatives exist. That means boundaries can move, and are not the same in every society. The evidence behind them is also mainly social-scientific, and less settled than the physics behind Planetary Boundaries.

Society can increase the safe operating space by adapting in ways that make it more resilient to technology risks: stronger regulation, better training and digital literacy, investment in real-world alternatives to screen-based activity, and robust institutions. All of these let society productively absorb more AI without crossing the threshold into instability. But society can also decrease the safe operating space when safeguards are absent or eroded.

Four small charts showing the same path of AI use against a safe operating space that rises with use, holds flat, increases and decreases. Where the space holds flat or decreases, the path rises above it into societal risk.

Investors and companies are betting trillions on continued compute expansion. What those investments are worth depends on how society responds to AI, through its politics, its courts and its rules, yet capital allocators have no systematic way to think about that response. If society acts to constrain AI in order to stay within Silicon Boundaries, capital allocated to AI infrastructure and technology companies could lose value. We call this risk the Silicon Bubble, a concept that draws on our previous work on stranded assets, unburnable carbon, and the carbon bubble.

We have set out four scenarios to help investors think about possible futures and how they might influence them. They are defined by two kinds of risk: transition risk, the chance that AI-related investments lose value, and societal risk, the chance that AI pushes society past its boundaries. In Goldilocks, both risks stay low. In Luddite, regulation, backlash and poor economics hold AI back, sparing society but not the sector. In Gilded Cage, AI spreads, does serious harm and leaves a two-tier economy. In Icarus, harm and losses arrive together.

A two-by-two grid of the four scenarios, with transition risk rising up the side and societal risk across: Goldilocks with both risks low, Luddite with transition risk high, Gilded Cage with societal risk high, and Icarus with both high.

The Silicon Boundaries Framework

Our aim is to provide tools that can help companies, financial institutions, and governments contribute to society remaining within a safe operating space while realising the benefits of AI and related technologies.

To do this, we need to better understand how societal stability is affected by AI and related technologies, identify and monitor relevant thresholds, and develop responses that enhance resilience while dampening the drivers of instability.

The first version of our framework (v0.1) sets out 45 individual boundaries in nine categories:

  • Information Integrity & Shared Reality
  • Social Cohesion & Connectedness
  • Political Stability
  • Economic Participation & Distributional Stability
  • Physical & Mental Health
  • Safety, Security, & the State’s Monopoly on Violence
  • Financial Stability & Capital Misallocation
  • Rights, Consent, & Human Dignity
  • Environmental Sustainability

 

With the framework, we can analyse any company and ask whether, and how, it increases or decreases the safe operating space.

A circle with the safe operating space at its centre, ringed by a dashed boundary, and the 45 individual boundaries of the v0.1 framework arranged around it in nine categories: information integrity, social cohesion, political stability, economic participation, health, safety and security, financial stability, rights and dignity, and environment.
Find out more

Read the Report and listen to the Podcast

You can read our first report here. You can also listen to our accompanying podcast, which discusses the framework and its implications.

Our research note, What do you believe about AI?, sets out fifteen questions on which reasonable investors might disagree about how AI should develop. Our beliefs tool turns an investor’s answers into a draft statement of its AI investment beliefs.

A matrix of the fifteen questions, grouped under the frontier, power and deployment, with each question's two sides, against the seven archetypes, showing the side each archetype takes.

We are now defining workable thresholds and developing indicators, scenarios and tools, so that investors can form their own views on AI, assess companies and portfolios against the boundaries, and engage the companies they invest in or finance. We also work with policymakers and supervisors.

Read the Report
Listen to the Podcast
Try the Beliefs Tool

Events

Silicon Boundaries in Practice

Directors

Directors

Dr Jakob Thomä

Co-Founder & Research Director, Theia Finance Lab

Jakob Thomä is one of the leading global experts on sustainable finance and long-term risks, and entrepreneur with a track record of incubating commercial and non-commercial sustainability solutions. He co-founded the 2° Investing Initiative entities in New York and Berlin and serves as Research Director of Theia Finance Labs. His book “Der Kill Score: Auf der Spur unseres ökologischen und sozialen Fußabdruck” explores the impact of sustainability on human lives.

Dr Ben Caldecott

Ben-Caldecott

Director, Oxford Sustainable Finance Group. Associate Professor and Senior Research Fellow of Sustainable Finance, Smith School of Enterprise and the Environment, University of Oxford

Dr Ben Caldecott is the founding Director of the Oxford Sustainable Finance Group and the inaugural Associate Professor of Sustainable Finance at the University of Oxford, a post maintained by an endowment and the first ever endowed professorship of sustainable finance. He is also a Supernumerary Fellow at Oriel College, Oxford. Ben operates across finance, policy, and research, working as part of boards and senior management, and with policymakers, regulators, and civil society on how to interpret and shape fast-moving energy, climate, environment, and wider sustainability topics.

Sign up to the newsletter

* indicates required
Email Format

Silicon Boundaries is a joint initiative of the Oxford Sustainable Finance Group at the University of Oxford and Theia Finance Labs. It asks how much AI society can productively absorb, and how to raise that amount over time, while staying within a safe operating space.