Does distributing compute distribute power?

Raising $25,000active

The US-led Pax Silica initiative seeks to prevent power concentration by distributing AI compute across allied democracies. Yet, this approach overlooks concentration within these blocs.

Project Details

Having spent five months in Japan and mapping on the ground the physical layers of AI transition in Asia through 'middle power chokepoints' like South Korea, Taiwan, Singapore and Japan, a central question that emerged during this time was:

does distributing compute across democracies as implied by plurilateral blocs like Pax Silica actually prevent centralisation of power or would the bloc itself become a concentrating mechanism.

or more simply – when compute is "distributed across democracies" via a plurilateral coalition (Pax Silica as a case study), does that genuinely disperse power or does it just relocate concentration into a smaller, opaque, intergovernmental club that itself becomes a durable chokepoint, and thus harder for citizens or even weaker member states to check?

WHAT I'LL DO:

As I now transition to complete my Masters in Public Policy at McGill University, I want to address these uncertainties/questions through my Capstone Project.

I recognise Pax Silica is a new (7-month) initiative so I can't evidence much its durability in 2026-27 but I would like to decompose it into questions that are answerable now, whose answer might also bear on the bigger question – examples could include:
- Q1: where does control actually sit inside the coalition today?
- Q2: does holding a physical chokepoint convert into proportionate influence?
- Q3: what do similar past examples tell us?
- Q4: what design or cooperative features would keep [identified problem] distributed?

This will be sharpened and polished under supervision from expert faculty at McGill including Professor Taylor Owen and Vincent Rigby and rigorously investigated and tested through: 1) the four-week intensive complexity seminars and 2) the practitioner-led policy labs.

THE CONCRETE OUTPUT:

the capstone as a client-facing policy report (June 2027)
an interim working paper or public piece [April/winter 2027]
the concentration map of the coalition as a standalone public resource (August 2027)

Theory of Impact

Advances in frontier AI systems in the upcoming years could confer enormous power on those who control them. If power does become extremely concentrated, it could translate into a global regime that would last for many generations into the future because once these structures become established, they may be difficult to reverse.

As a result, my project aims to focus on a specific international governance failure that could become locked in early, examining who ends up controlling the infrastructure needed to build and deploy increasingly capable frontier AI systems. Fundamentally for this, timing matters. As new AI coalitions (such as Pax Silica) begin to emerge, this creates a rare opportunity to examine not just if these coalitions distribute AI infrastructure geographically, but whether they can genuinely distribute decision-making power.

Correspondingly, my project's impact lies in contributing towards this evidence. If more accountable and distributed governance arrangements are adopted early, the likelihood that control over frontier AI systems becomes concentrated in ways that are difficult to reverse may be reduced. Furthermore, my project outputs will provide an openly accessible resource for the wider AI community to scrutinise, build upon, challenge and learn. Finally, regardless of the findings, the project adds evidence where little currently exists.

People

Ameema Talat

Team Member

Grants Received– no grants recorded

Add grant

Funding Asks

grantmaking.ai Launch Round

Applied

Minimum

$25,000

Ideal

$48,000

How the money will be spent

The grant would cover the costs of the MPP:

This spreadsheet below includes a ~10% contingency amongst costs with tabs on [Minimum, Average and Maximum costs].
https://docs.google.com/spreadsheets/d/1gv0ldB0ZgEKC-YB-mLfeBGBWrfX8DClvAabLXQQUcFI/edit?gid=1483731140#gid=1483731140

Without contingency:

Minimum ($25,000): covers the tuition gap for the MPP year at McGill. At this level the year proceeds, with living costs bridged through pending applications elsewhere and part-time work.

Ideal ($47,500): fully funds the year including: tuition (net of the scholarship), compulsory fees and insurance, and living costs in Montréal, allowing me to prioritise the capstone, research outputs, and relevant placements over paid work.

McGill's entrance scholarship (CAD10,000) is the only funding secured to date. Two applications are pending elsewhere; I will update this page as soon as any commitment lands, per the round's requirements.

Discussion

No comments yet. Be the first to share your thoughts.