Silicon boundaries: Understanding, anticipating, and managing the societal instability that can be caused by AI and related technologies
This report introduces Silicon Boundaries, thresholds where AI and compute-powered technologies create systemic societal instability. Mirroring planetary boundaries, the framework identifies nine risk categories—from information integrity to environmental sustainability—and outlines scenarios for a potential ‘Silicon Bubble’, offering guidance for investors to manage these emerging technological risks.
Please login or join for free to read more.
OVERVIEW
Foreword
Developed in September 2025, the concept of Silicon Boundaries aims to address the societal instability potentially caused by artificial intelligence and related technologies. This report serves as a public launch for the concept, which is intended to open a joint conversation on how to avoid crossing these thresholds or return to safety where they have already been breached. The intellectual origin of the framework is based on planetary boundaries in environmental science, translating those concepts for financial markets, policymakers, and the real economy.
Evidence of societal risks
There is growing evidence that technologies powered by compute and silicon create significant societal risks. This is illustrated by recent developments such as legal settlements over copyright, national security concerns regarding Artificial General Intelligence, and the impacts of excessive screen time on physical and mental health. Specific examples include AI bots impacting student performance in rural Colombia and the alarm sounded regarding AI chatbots following teenage suicides.
Compute-powered technologies
While the products and services underpinning these risks are varied, they all realise their potential via the mobilisation of silicon-based computing power. These include Large Language Models (LLMs), social media, screen use, autonomous agents, robotics, and quantum computing.
Defining Silicon Boundaries
Silicon Boundaries are defined as a set of thresholds where the benefits of AI and related technologies are outweighed by their contributions to societal instability. Crossing these thresholds can trigger non-linear change and irreversible tipping points. The concept argues that as compute scales, these technologies could cross societal boundaries unsustainably, impacting economic, financial, and political systems at national and global scales.
The silicon bubble
Mirroring the ‘carbon bubble’, there is a forecasted amount of compute that may become ‘stranded’ if society chooses to avoid breaching Silicon Boundaries. This potential ‘Silicon Bubble’ represents the gap between planned compute by technology companies and compute consistent with safe societal boundaries.
Differences from carbon
Unlike carbon, the science relating compute to Silicon Boundaries is not yet settled, and the relationship is less linear due to the nature of deployment and societal response. Two crucial differences are highlighted: the aggregate level at which costs outweigh benefits is unknown, and boundaries differ between ‘personal use’ and ‘societal use’. For instance, limiting compute for personal consumption may be desirable while expanding it for societal benefits in healthcare remains viable.
Preliminary framework
The Silicon Boundaries 0.1 framework identifies a preliminary set of nine categories and over thirty individual boundaries. This framework is expected to evolve as science and the world change, incorporating engagement from various stakeholders.
Nine boundary categories
The framework spans information integrity, social cohesion, democratic integrity, economic participation, physical and mental health, safety and security, financial stability, rights and consent, and environmental sustainability. Individual boundaries include disinformation and synthetic media, screen-mediated displacement of in-person interaction, and computational propaganda.
Individual boundaries by category
Information integrity involves the large-scale infiltration of social systems by AI agents and the erosion of trust infrastructure. Social cohesion is threatened by algorithmic polarisation and filter bubbles, while political stability faces risks from micro-targeted manipulation and ‘loss of human control’ scenarios. Economic boundaries include labour displacement and winner-takes-all market concentration. Health risks include addictive design and youth developmental harms. Safety and security involve autonomous weapons and AI-enabled cyber-attacks. Financial stability is threatened by algorithmic herding and overinvestment in compute infrastructure relative to societal returns. Rights and consent involve mass surveillance and the deployment of irrevocable AI systems. Finally, environmental sustainability is impacted by the energy and water consumption of data centres.
Tipping points
Crossing Silicon Boundaries may lead to irreversible tipping points. Loneliness driven by digitisation increases mortality risk by 26%, equivalent to the health effect of smoking 15 cigarettes a day. A Silicon Tipping Point could thus have the health effect of turning the majority of the population into chain smokers. Politically, it could drive the collapse of liberal democracies, while economically it could create monopolies similar to the ‘Seven Sisters’ oil companies.
NGFS scenario framework
The report applies the Network for Greening the Financial System (NGFS) scenario framework to silicon risks. Scenarios include the ‘Luddite Scenario’ (high transition, low societal risk), the ‘Icarus Scenario’ (high transition, high societal risk), the ‘Goldilocks Scenario’ (low transition, low societal risk), and the ‘Gilded Cage Scenario’ (low transition, high societal risk).
Transition risks if the silicon bubble bursts
The technology sector’s market capitalisation is two to three times higher than that of ‘carbon bubble’ companies during their original publication. Morgan Stanley estimates $2.9 trillion in data centre capex for 2025–2028. Financial markets expect these companies to capture more than two-thirds of all stock market earnings growth over the next five years. Potential litigation risks also represent existential threats to business models.
Societal risks if the silicon bubble bursts
Mental health costs already equate to approximately 2% of US GDP, a figure that could double under pessimistic scenarios. Political instability may reduce global growth by 10% based on historical literature. Unemployment rates driven by AI could increase by 5–10% in a pessimistic scenario.
Next steps for research
Research priorities include quantifying societal boundaries and developing scenarios, data, and models to measure impacts on companies and portfolios. There is also a need to understand synergies between the Silicon Bubble and other sustainability themes.
Next steps for investors
Investors are encouraged to implement scenario analysis under ‘Silicon scenarios’ and ramp up capabilities for engagement with policymakers and companies. They should also develop portfolio allocation strategies that meaningfully differentiate between silicon companies based on risks and opportunities.