The article has one crucial omission, maybe it will be addressed later. It takes the scenario where AI in its current form is going to be massively useful at face value.
What if that's not the case? What if it is, but does not need these massive compute clusters with chips designed by NVDA? It's too early to consider a "compudollar".
1) I do not think a compute-dollar should happen or even will happen. But I do think its worth laying out the proposal because if an attempt is made, the changes it makes to export controls and extraterritorial control will be material even if the scheme fails
Brilliant work connecting Petrodollar history to compute diplomacy. The "comprador bourgeoisie" framing for Gulf states is devastating, they're basically becoming glorified landlords while calling it sovereignity. I've watched the SoftBank numbers unfold and it's wild how transparently dependent the whole structure is. That cannibalization trap section raises questions nobody's addressing, what happens when AI infrastructure starts eating the actual export revenues its suppposed to diversify away from.
Great article and I love how you string the commentary together! I wonder what you might think of an article a colleague and I published on ‘national computing power’ in the U.S v China (sorry for the shameless plug): https://journals.sagepub.com/doi/full/10.1177/0308518X251369704
I really really enjoyed this! I wish i saw this earlier but will amend some later parts to include the analysis cause there's a lot of points that intersect with what i'm interested in and the article raises a lot of questions for me / inspires some line of inquiry i want to follow up next year. Is it too late to curb the power of state-aligned hyperscalers? How hard will it be to roll back the AI sector grafting itself onto US political economy or if we will just have to suffer the latest wave of privatization and whatever happens as asset managers/hyperscalers/utilities/natsec merge their interests into one delicious bowl? What damage will this new Cold War do in terms of permanently foreclosing certain green developments here or harming them elsewhere by racing for critical minerals while massively building out fossil fuel infrastructure. Is the US gonna try to emulate elements of Chinese state capitalism in the course of the overbuild? Are we gonna sabotage their green development investment overseas? How much of the coming techno-industrial order is gonna be built with executive action and how much of that can be rolled back through non-executive means since we're fucked for the next few years. Really really fascinating article
fascinating....adding to the question below, isn't there another key assumption here that US-based hardware/stack is going to remain superior to China's for the foreseeable future?
which seems a reasonable assumption currently, but may well turn out differently ? in particular China's tech infrastructure seems to be configured for building much larger interconnected clusters of (indivuidually less powerful) chips - which may be especially well suited to gulf states with energy to burn?
I was going to ask about what happens if AI demand doesn't actually match up what their hopes are, but you're going to cover it in subsequent parts, so I'll look forward to your subsequent posts.
That being said, I just forwarded this to all the policy nerds I know for their commentary and edification. Looking forward to hearing more about about this in the… uh… oh, crap, it's two more weeks in the year.
Great stuff—feels like a first glance (for me) at the kind of classic “triangle trade” diagram that will appear in future textbooks to describe our era (if these are even a thing...)
I'm interested in it too as later sections will lay out! But it would be a mistake to try to build a technodollar system (don't think it would even work) and it would be a mistake to export compute to Gulf and it is a mistake to sell Sovereign AI. All of these developments would subsume whatever we might find exciting and interesting into horrific systems
The capex-to-revenue cascade. You need enormous chips to train models that generate marginal improvements. At some point, the cost of marginal improvement exceeds the value it creates in deployed systems. We're very close. The unwind will be fast because the cliff is sharp—nothing looks broken until it does.
Phenomenal work bringing together so many disparate threads into a convincing (and disturbing) vision of an "imperial strategy". This made me think of a recent post from another of my favorite tech critics (and AI security expert), the brilliant Disesdi Shoshana Cox (https://open.substack.com/pub/disesdi/p/trump-is-officially-the-ai-president?utm_campaign=post-expanded-share&utm_medium=web). Here Cox proposes Trump is not just interested in profiteering from AI or making his tech bros happy. He wants to leave a legacy - of AI in warfare. "To the Trump administration, the “AI arms race” is literal." Military domination through AI would be the other side of the coin of imperial strategy - for those cases economic domination doesn't cut it. Let's hope none of this actually happens.
A masterclass in geopolitical foresight. This unpacks AI not just as a technological force, but as a global chessboard of power, capital, and dependency. Your analysis makes clear that the stakes extend far beyond Silicon Valley, this is about who controls the digital infrastructure of the 21st century.
I talk about the latest AI trends and insights. If you’re interested in practical AI strategies for navigating the geopolitical and industrial forces shaping the AI economy, check out my Substack. I’m sure you’ll find it very relevant and relatable.
The article has one crucial omission, maybe it will be addressed later. It takes the scenario where AI in its current form is going to be massively useful at face value.
What if that's not the case? What if it is, but does not need these massive compute clusters with chips designed by NVDA? It's too early to consider a "compudollar".
I will address this two ways in later parts:
1) I do not think a compute-dollar should happen or even will happen. But I do think its worth laying out the proposal because if an attempt is made, the changes it makes to export controls and extraterritorial control will be material even if the scheme fails
2) There will be a part on use value!
Brilliant work connecting Petrodollar history to compute diplomacy. The "comprador bourgeoisie" framing for Gulf states is devastating, they're basically becoming glorified landlords while calling it sovereignity. I've watched the SoftBank numbers unfold and it's wild how transparently dependent the whole structure is. That cannibalization trap section raises questions nobody's addressing, what happens when AI infrastructure starts eating the actual export revenues its suppposed to diversify away from.
SoftBank has always been honest about what it is!
Great article and I love how you string the commentary together! I wonder what you might think of an article a colleague and I published on ‘national computing power’ in the U.S v China (sorry for the shameless plug): https://journals.sagepub.com/doi/full/10.1177/0308518X251369704
I really really enjoyed this! I wish i saw this earlier but will amend some later parts to include the analysis cause there's a lot of points that intersect with what i'm interested in and the article raises a lot of questions for me / inspires some line of inquiry i want to follow up next year. Is it too late to curb the power of state-aligned hyperscalers? How hard will it be to roll back the AI sector grafting itself onto US political economy or if we will just have to suffer the latest wave of privatization and whatever happens as asset managers/hyperscalers/utilities/natsec merge their interests into one delicious bowl? What damage will this new Cold War do in terms of permanently foreclosing certain green developments here or harming them elsewhere by racing for critical minerals while massively building out fossil fuel infrastructure. Is the US gonna try to emulate elements of Chinese state capitalism in the course of the overbuild? Are we gonna sabotage their green development investment overseas? How much of the coming techno-industrial order is gonna be built with executive action and how much of that can be rolled back through non-executive means since we're fucked for the next few years. Really really fascinating article
Great article, thank you for this! Looking forward to the subsequent parts.
fascinating....adding to the question below, isn't there another key assumption here that US-based hardware/stack is going to remain superior to China's for the foreseeable future?
which seems a reasonable assumption currently, but may well turn out differently ? in particular China's tech infrastructure seems to be configured for building much larger interconnected clusters of (indivuidually less powerful) chips - which may be especially well suited to gulf states with energy to burn?
I was going to ask about what happens if AI demand doesn't actually match up what their hopes are, but you're going to cover it in subsequent parts, so I'll look forward to your subsequent posts.
That being said, I just forwarded this to all the policy nerds I know for their commentary and edification. Looking forward to hearing more about about this in the… uh… oh, crap, it's two more weeks in the year.
Great stuff—feels like a first glance (for me) at the kind of classic “triangle trade” diagram that will appear in future textbooks to describe our era (if these are even a thing...)
Uglier not "more ugly." Well written otherwise and quite fascinating.
thanks for reading to the end o7
Lol
Really well reported and written, though I find gen AI to be a cool branch of technology.
I'm interested in it too as later sections will lay out! But it would be a mistake to try to build a technodollar system (don't think it would even work) and it would be a mistake to export compute to Gulf and it is a mistake to sell Sovereign AI. All of these developments would subsume whatever we might find exciting and interesting into horrific systems
The capex-to-revenue cascade. You need enormous chips to train models that generate marginal improvements. At some point, the cost of marginal improvement exceeds the value it creates in deployed systems. We're very close. The unwind will be fast because the cliff is sharp—nothing looks broken until it does.
Phenomenal work bringing together so many disparate threads into a convincing (and disturbing) vision of an "imperial strategy". This made me think of a recent post from another of my favorite tech critics (and AI security expert), the brilliant Disesdi Shoshana Cox (https://open.substack.com/pub/disesdi/p/trump-is-officially-the-ai-president?utm_campaign=post-expanded-share&utm_medium=web). Here Cox proposes Trump is not just interested in profiteering from AI or making his tech bros happy. He wants to leave a legacy - of AI in warfare. "To the Trump administration, the “AI arms race” is literal." Military domination through AI would be the other side of the coin of imperial strategy - for those cases economic domination doesn't cut it. Let's hope none of this actually happens.
Persian gulf, not just gulf
A masterclass in geopolitical foresight. This unpacks AI not just as a technological force, but as a global chessboard of power, capital, and dependency. Your analysis makes clear that the stakes extend far beyond Silicon Valley, this is about who controls the digital infrastructure of the 21st century.
I talk about the latest AI trends and insights. If you’re interested in practical AI strategies for navigating the geopolitical and industrial forces shaping the AI economy, check out my Substack. I’m sure you’ll find it very relevant and relatable.
you forget something crucial. Interesting analysis. Energy still rules. nothing matters without that