This is part linkpost, and an expansion on a previous shortform posted on LW.
I believe a massively undermade point in favour of the belief that AI will be a very big deal is that the AI buildout over the past four years is the largest bet in the history of capitalism. A peacetime industrial effort is underway which has not had a comparison since the nineteenth century. The size of the effort and the scale of the spend signal that key decision makers and capital allocators believe that very powerful AI will be achieved in the next few years.
Many of us are vaguely aware that large sums are being spent, but the scale of the undertaking is astonishing:
The six largest spenders this year are spending more on capital expenditures this year than the Russian government, governing 11 timezones and 143 million people, fighting a war on which the fate of its regime probably depends, spent last year.
Hyperscaler spending (Alphabet, Amazon, Microsoft, Meta and Oracle) will be $785-820bn this year, about 2.4% of US GDP. There are only 11 governments on earth who spend more than that. Next year the Hyperscalers are projected to spend about $1.3 trillion dollars.
Google and Amazon are each spending about 0.6% of US GDP. The Manhattan project and Apollo programme each accounted for about 0.4% of US GDP in their peak years of spending. Both of these companies has announced their intention to spend more next year.
The dramatic capital expenditure figures actually dramatically undersell the scale of the spend, because they are commiting to future purchases and leases at an incredible rate. Just looking at the second quarter of this year:
In the three months between the start of April and the end of June this year Google’s financial commitments to future spending increased by half a trillion dollars. In the same period Meta signed $233bn of new spending commitments and added a further $68bn in data center leases in July. Microsoft signed more than $130bn of new data centre leases and Amazon increased their commitments by $74bn after issuing a separate record breaking bond for $53bn in March.
Having burned through close to all of their own free cash flow, Hyperscalers, Neo-Clouds and Labs are finding increasingly creative ways to mobilise the capital of others. Until recently issuing substantial amounts of debt was a rare thing for big-tech companies. This year hyperscalers have issued record setting corporate bonds in the US, UK, EU, Switzerland, Australia and Canada, and SoftBank has issued the largest junk bond in history to fund investment in OpenAI.
Companies in China are equally commited. Tencent, for example, (of WeChat fame) had a negative free cash flow in Q2, their first negative flow in over a decade. This year combined the Chinese hyperscalers will spend about $140bn, which is a huge amount of money, more than double their spend last year, but less than is being spent by Meta alone. They are not able to spend as much in absolute terms as the US Hyperscalers, because (1) their businesses simply don't bring in anwhere near as much money and (2) the United States has a singular talent for financial engineering.
This spending gap has very important implications for the AI race.
By mid 2025 the US owned about 75% of global frontier compute while China had secured only 15%, a gap which can only have grown in the past ~16 months given the US's superior spending, superior chips and export restrictions. America has a substantial lead, which will become a gulf if the labs are nationalised/ forced to collaborate. This gap can be used to negotiate a pause/slowdown/pacing with China from a position of strength, and failing that can be independently burned to pause/pace the efforts of US labs.
These trillions of dollars are being put to use in the real world. Thousands of data centres are under construction, the largest of which consume as much electricity as major cities and cover thousands of acres. Gas pipelines are being re-routed and nuclear power plants re-opened and constructed from scratch to fuel them. Amazon, Meta, Microsoft and Google are all investing in multiple nuclear plants, each. Google, Nvidia and Sam Altman are all investing in companies pursuing nuclear fusion. The worlds richest man is semingly seriously discussing space based data centres and Google have sent TPUs to space to test the concept. Terrestrially, next year US data centres will require 66GW of continuous power, up from 31GW in 2025. 66GW is roughly equivalent to the average amount of electricity used by Germany in wintertime. Blackrock predicts AI data centres could use 15-20% of current US electricity demand by 2030. Morgan Stanley belives the US will be short of power by six New York Cities by 2028 and around 18 New York Cities by 2029.
What this tells us.
These efforts signal some very important things about the beliefs of key decision makers on both sides of the pacific, and those enabling their spend, who are amongst the most successful capital allocators in the world:
To recoup their spend AI will soon have to generate far more revenue than it currently does, so they believe very powerful, very useful, and hence very profitable AI will be achieved.
Most of this money is being spent on chips which have lifespans of about 6 years. The life time of the patience of shareholders is likely considerably shorter, so the key decision makers believe this will be achieved soon.
No-one is content to wait and see if AGI is achieved and then invest. They are investing ahead of time. They believe that such will be the bonanza when it is achieved that they don't want to miss out on even a few years of it.
These beliefs that very powerful AI can be achieved and will be achieved soon are held very strongly, not weakly. The companies are not placing small bets. They have burned through close to all of their free cash flow, have issued record setting corporate bonds in the US, UK, EU, Switzerland, Australia and Canada. They have spending commitments for beyond this year in the trillions of dollars, etc. Oracle and both the leading labs are commited to a point where there companies probably won't survive a failure to reach AGI.
Leaving aside the question of whether very powerful AI can be built safely using anything like current methods, because of the scale of the spending key decision makers need to achieve it soon to justify their spend. I think this is extremely significant; it will encourage recklessness, to say the least.
The above is a summary of the key points of section one of the essay. Section two explores the current capabilities of the technology, (something most LWers can probably skip, though I believe there is value there) and section three explores what comes next; where this titanic industrial effort is leading us as a civilisation.
This is part linkpost, and an expansion on a previous shortform posted on LW.
I believe a massively undermade point in favour of the belief that AI will be a very big deal is that the AI buildout over the past four years is the largest bet in the history of capitalism. A peacetime industrial effort is underway which has not had a comparison since the nineteenth century. The size of the effort and the scale of the spend signal that key decision makers and capital allocators believe that very powerful AI will be achieved in the next few years.
Many of us are vaguely aware that large sums are being spent, but the scale of the undertaking is astonishing:
Above quote is from the essay, I.II.
Having burned through close to all of their own free cash flow, Hyperscalers, Neo-Clouds and Labs are finding increasingly creative ways to mobilise the capital of others. Until recently issuing substantial amounts of debt was a rare thing for big-tech companies. This year hyperscalers have issued record setting corporate bonds in the US, UK, EU, Switzerland, Australia and Canada, and SoftBank has issued the largest junk bond in history to fund investment in OpenAI.
Companies in China are equally commited. Tencent, for example, (of WeChat fame) had a negative free cash flow in Q2, their first negative flow in over a decade. This year combined the Chinese hyperscalers will spend about $140bn, which is a huge amount of money, more than double their spend last year, but less than is being spent by Meta alone. They are not able to spend as much in absolute terms as the US Hyperscalers, because (1) their businesses simply don't bring in anwhere near as much money and (2) the United States has a singular talent for financial engineering.
This spending gap has very important implications for the AI race.
By mid 2025 the US owned about 75% of global frontier compute while China had secured only 15%, a gap which can only have grown in the past ~16 months given the US's superior spending, superior chips and export restrictions. America has a substantial lead, which will become a gulf if the labs are nationalised/ forced to collaborate. This gap can be used to negotiate a pause/slowdown/pacing with China from a position of strength, and failing that can be independently burned to pause/pace the efforts of US labs.
These trillions of dollars are being put to use in the real world. Thousands of data centres are under construction, the largest of which consume as much electricity as major cities and cover thousands of acres. Gas pipelines are being re-routed and nuclear power plants re-opened and constructed from scratch to fuel them. Amazon, Meta, Microsoft and Google are all investing in multiple nuclear plants, each. Google, Nvidia and Sam Altman are all investing in companies pursuing nuclear fusion. The worlds richest man is semingly seriously discussing space based data centres and Google have sent TPUs to space to test the concept. Terrestrially, next year US data centres will require 66GW of continuous power, up from 31GW in 2025. 66GW is roughly equivalent to the average amount of electricity used by Germany in wintertime. Blackrock predicts AI data centres could use 15-20% of current US electricity demand by 2030. Morgan Stanley belives the US will be short of power by six New York Cities by 2028 and around 18 New York Cities by 2029.
What this tells us.
These efforts signal some very important things about the beliefs of key decision makers on both sides of the pacific, and those enabling their spend, who are amongst the most successful capital allocators in the world:
Leaving aside the question of whether very powerful AI can be built safely using anything like current methods, because of the scale of the spending key decision makers need to achieve it soon to justify their spend. I think this is extremely significant; it will encourage recklessness, to say the least.
The above is a summary of the key points of section one of the essay. Section two explores the current capabilities of the technology, (something most LWers can probably skip, though I believe there is value there) and section three explores what comes next; where this titanic industrial effort is leading us as a civilisation.