strategy tag
Compute governance.
Control flops via export controls, licensing, reporting
stated endorsers
12
no opposers yet
profiled endorsers
0
248 on the board total
endorser p(doom)
·
no estimates on record
quotes by endorsers
12
just for this tag
People on the record.
12Catherine Aiken
CSET researcher; China AI talent and capability
Analyses Chinese AI capability developments; supports chip-export-control policy informed by capability data.
Understanding Chinese AI capability requires rigorous open-source intelligence work, not rhetoric about the race.
Cliff Young
Google; TPU principal engineer
Argues custom AI silicon is reshaping the economics of frontier model training; views chip design as inseparable from AI capability progress.
TPUs were designed because we knew we needed orders-of-magnitude more compute for the kinds of models we wanted to train. Software-hardware co-design at this scale changes the cost structure of AI research.
David Patterson
UC Berkeley emeritus; Google AI hardware; 2017 Turing Award
Argues domain-specific architectures (TPUs, GPUs) are how AI scaling is being achieved; the supply and design of these chips is now a strategic concern of states.
We are entering a new golden age of computer architecture, driven by the end of Moore's Law and the rise of domain-specific accelerators. AI is the dominant such domain, and where the chips are made matters more than ever.
Elsa Kania
CNAS adjunct senior fellow; China AI specialist
Argues China's military-civil fusion strategy makes Chinese frontier AI a national-security concern for the U.S., and that calibrated export controls are the most effective response.
China's strategy is to leverage AI for what it calls 'intelligentized warfare'. Western democracies have to take that seriously without overreacting to specific capability claims.
Jordan Schneider
ChinaTalk podcast host; Rhodium Group
Argues U.S. policymakers chronically underestimate the velocity and ambition of Chinese AI strategy; supports calibrated chip export controls as the most consequential policy lever the U.S. has.
If you only read English-language coverage of Chinese AI, you will keep being surprised. The Chinese discourse is detailed, candid, and several years deeper than the West gives it credit for.
Lennart Heim
Compute governance researcher at RAND
Published foundational frameworks for compute-based AI governance including verification and on-chip mechanisms.
Compute is the most promising governance lever because it is detectable, quantifiable, and controllable.
Pat Gelsinger
Former Intel CEO; chip-supply geopolitics
Argues that semiconductor supply chains are the load-bearing element of any AI-governance regime; pushed CHIPS Act funding and has called for democracies to control fabrication.
Where the oil reserves are has defined geopolitics for the last 50 years. Where the fabs are will define the next 50.
Paul Scharre
CNAS executive VP; 'Army of None', 'Four Battlegrounds' author
Argues U.S.-China AI competition will be decided largely by compute, talent, data, and institutions; pushes export controls on advanced chips as a load-bearing policy lever.
Of the four battlegrounds, data, compute, talent, and institutions, compute is the one most amenable to export controls. The semiconductor supply chain is a chokepoint we should use carefully.
Saif M. Khan
Former NSC AI technology director
Helped design chip export controls and frontier-compute reporting requirements.
Export controls on advanced AI chips are among the most consequential US technology policy decisions of the decade.
Stephen Witt
Author of 'The Thinking Machine' (NVIDIA history)
Documents the rise of NVIDIA-as-AI-infrastructure; chronicles the structural concentration that compute governance is responding to.
Whoever controls the GPU controls the future of AI. NVIDIA holds that key.
Tim Fist
Institute for Progress AI policy researcher
Advocates US industrial policy that couples chip export controls with domestic capacity expansion.
Export controls work only if paired with domestic capacity. Otherwise they're unilateral disarmament.
Zachary Arnold
Georgetown CSET; analytics lead
Argues data-driven analysis of compute and talent flows shows export controls are working in important respects; supports continued, calibrated tightening.
Our analysis suggests U.S. chip export controls are imposing real costs on Chinese frontier-AI development, not crippling it, but slowing it in measurable ways.