Peter, you asked the single most important question of our era: “Who owns the machines when machines do the work?” You correctly identified that if ownership centralizes, we get jobless poverty with better branding and that the public is beginning to physically attack centralized AI infrastructure because it acts as a thermodynamic tax on their local grid with no visible local benefit.
But you are treating these two problems as separate contradictions. They are not. They are the exact same architectural error. The tension you are sensing is the friction of the Hub-and-Spoke mainframe model.
The solution to the ownership crisis and the infrastructure crisis is the same: We must deprecate the Data Center. The "Abundance" you advocate for cannot be achieved by taxing a monopoly and distributing the fiat; it must be achieved at the protocol layer.
Look at what is happening right now with Covenant-72B, Bittensor, and the broader DePIN ecosystem. Covenant-72B is an open-weight, heavily quantized frontier model designed specifically for edge inference and distributed swarm training. It doesn't require a 1.2 GW fortress in Texas to run. It is designed to run across the millions of M-series MacBooks, gaming GPUs, and local solar-powered microgrids that already sit idle 80% of the day across the globe.
When you shift from Cloud Compute to Fog Compute, you solve both issues simultaneously:
1. You solve the Infrastructure Backlash.
We do not need to fight local zoning boards or strain the PJM grid for massive new data centers. The compute already exists, distributed globally across consumer hardware. By routing the AI workload through a decentralized network (like TON, Akash, or Bittensor), the heat and power draw are dissipated across millions of homes. You remove the physical target.
2. You solve the Ownership Crisis.
If the network is powered by the decentralized hardware of the users, the users are the infrastructure. When Covenant-72B executes a task, the cryptographic micro-payment for that compute goes directly to the owner of the local node. We earn the "Automation Dividend" directly, because we own the metal the intelligence runs on.
You built your career teaching us that technology moves from Scarcity to Abundance by becoming decentralized and demonetized. Do not abandon your own thesis just because the mega-cap tech incumbents are terrified of losing control of the transition.
Power consumption is power consumption regardless of whether it's distributed on not. There will be a price to pay, by all. And maybe I'm missing something, but how does distribution of hardware increase agency?
If your GPU is contributing to AI deployment, you stop being a consumer of compute and become a sovereign. Our deployment of intelligence becomes the source of our Abundance.
Moreover, when training runs are coordinated on a blockchain, every participant gets a vote on what to focus next in proportion to their stake and/or compute. And because inference also depends on decentralized compute, no centralized provider can unilaterally decide to lock the model away like Anthropic just did.
I was reading a few days ago of an invention that allows computer chips to operate on 1/10 or less of current power requirements. That would reduce the problem with AIs power consumption.
Also, fusion power will be basically unlimited and so should be relatively cheap, while we still use archaic old concepts like money.
Fusion is a pipe dream. And, as Elon says, we have a massive fusion reactor in the sky. Solar/batteries continues to drop tho and is immediately deployable, unlike nuclear. Not sure why we're not seeing these in the desserts - maybe cause of lack of water. And yes, chips are getting more efficient (as are AI models) but demand is greatly outpacing that progress.
You could argue that NVIDIA and TSMC already have that control architecture at the chip level. Assembly firms like Dell and Lenovo don't really matter compared to that.
Above are links to companies that build data centers underwater . Uses seawater cooling. There maybe other companies out there, but this seems to be able to power the compute and is environmentally friendly no stress on local inhabitants.
I’m concerned that government inflation will steal the productivity gains that would normally flow to all people in the form of lower prices and instead have it flow to the select few near the money printer.
The "contradictions are signals" framing is the useful part. Most disruption analysis treats the friction as noise to be optimized away, when it's usually the exact place where the old system is exposing the rule it can't enforce anymore. The rule worth watching right now is "who owns the compute that owns your workflow" — that's where the next wave of dependency gets set before anyone notices.
There wouldn't be a backlash against data centers if the local people weren't being asked to subsidize them. The poor paying for the rich to get richer. The data center buildout hype was supposed to bring about abundance and lower the cost of energy, but they won't until they are forced to build energy infrastructure. Even then, their profit structure will be set up in such a way that they don't share with the community.
You are on point with this one Peter. I’ve been busy commenting on other people‘s posts on this Anthropic news and my first question is would OpenAI and xAI and Google have done the same thing? Second question is whether this was one explanation for the recent tussle between Anthropic and the US government. I’m sure many in the national security system would love to get their hands on a way to hack Iran’s power grid right now.
The infrastructure ownership point hits on something I've been sitting with. Diamandis frames it as shared abundance vs concentrated wealth, which is the right frame.
But I'd push back slightly: the historical examples (electricity, internet) had decentralized infrastructure over time due to regulatory pressure and competitive markets. AI infrastructure might be structurally different - training a frontier model requires capital concentration almost by design.
The optimism I have is on the application layer, not the model layer. Local models are getting surprisingly good. The productivity gains can be distributed even if the infrastructure isn't.
Still - the warning about who owns the pipes is worth taking seriously.
The bigger question, I fear, is how do we get from here to there without destroying ourselves. Govt is not even remotely capable of handling the challenges ahead. So who? How?
Peter, you asked the single most important question of our era: “Who owns the machines when machines do the work?” You correctly identified that if ownership centralizes, we get jobless poverty with better branding and that the public is beginning to physically attack centralized AI infrastructure because it acts as a thermodynamic tax on their local grid with no visible local benefit.
But you are treating these two problems as separate contradictions. They are not. They are the exact same architectural error. The tension you are sensing is the friction of the Hub-and-Spoke mainframe model.
The solution to the ownership crisis and the infrastructure crisis is the same: We must deprecate the Data Center. The "Abundance" you advocate for cannot be achieved by taxing a monopoly and distributing the fiat; it must be achieved at the protocol layer.
Look at what is happening right now with Covenant-72B, Bittensor, and the broader DePIN ecosystem. Covenant-72B is an open-weight, heavily quantized frontier model designed specifically for edge inference and distributed swarm training. It doesn't require a 1.2 GW fortress in Texas to run. It is designed to run across the millions of M-series MacBooks, gaming GPUs, and local solar-powered microgrids that already sit idle 80% of the day across the globe.
When you shift from Cloud Compute to Fog Compute, you solve both issues simultaneously:
1. You solve the Infrastructure Backlash.
We do not need to fight local zoning boards or strain the PJM grid for massive new data centers. The compute already exists, distributed globally across consumer hardware. By routing the AI workload through a decentralized network (like TON, Akash, or Bittensor), the heat and power draw are dissipated across millions of homes. You remove the physical target.
2. You solve the Ownership Crisis.
If the network is powered by the decentralized hardware of the users, the users are the infrastructure. When Covenant-72B executes a task, the cryptographic micro-payment for that compute goes directly to the owner of the local node. We earn the "Automation Dividend" directly, because we own the metal the intelligence runs on.
You built your career teaching us that technology moves from Scarcity to Abundance by becoming decentralized and demonetized. Do not abandon your own thesis just because the mega-cap tech incumbents are terrified of losing control of the transition.
I love the idea of micronized payments for use of resources
Based on sharing the compute infrastructure
How do we
keep the pirates out
Control the hackers
I worked for a micropayment software company for a short while 20 years ago. It failed miserably.
Yes to this reminder to promote the decentralized solutions as much as possible
Power consumption is power consumption regardless of whether it's distributed on not. There will be a price to pay, by all. And maybe I'm missing something, but how does distribution of hardware increase agency?
If your GPU is contributing to AI deployment, you stop being a consumer of compute and become a sovereign. Our deployment of intelligence becomes the source of our Abundance.
Moreover, when training runs are coordinated on a blockchain, every participant gets a vote on what to focus next in proportion to their stake and/or compute. And because inference also depends on decentralized compute, no centralized provider can unilaterally decide to lock the model away like Anthropic just did.
I was reading a few days ago of an invention that allows computer chips to operate on 1/10 or less of current power requirements. That would reduce the problem with AIs power consumption.
Also, fusion power will be basically unlimited and so should be relatively cheap, while we still use archaic old concepts like money.
Fusion is a pipe dream. And, as Elon says, we have a massive fusion reactor in the sky. Solar/batteries continues to drop tho and is immediately deployable, unlike nuclear. Not sure why we're not seeing these in the desserts - maybe cause of lack of water. And yes, chips are getting more efficient (as are AI models) but demand is greatly outpacing that progress.
Are, then, the few companyies that have the majority of computer sales (Apple, Dell, Lenovo) become the centralized control architecture?
You could argue that NVIDIA and TSMC already have that control architecture at the chip level. Assembly firms like Dell and Lenovo don't really matter compared to that.
https://subseacloud.com/
https://www.networkocean.io/
Above are links to companies that build data centers underwater . Uses seawater cooling. There maybe other companies out there, but this seems to be able to power the compute and is environmentally friendly no stress on local inhabitants.
Great article but what happened to moonshots? I look for it constantly and you just all but disappeared.
I’m concerned that government inflation will steal the productivity gains that would normally flow to all people in the form of lower prices and instead have it flow to the select few near the money printer.
AI is inherently deflationary.
Congratulations ! Excelent !!!
Hope the moonshots work in favour of humanity and trust me it’s the best time to be alive.
When will the podcasts address https://cdn.openai.com/pdf/561e7512-253e-424b-9734-ef4098440601/Industrial%20Policy%20for%20the%20Intelligence%20Age.pdf
The "contradictions are signals" framing is the useful part. Most disruption analysis treats the friction as noise to be optimized away, when it's usually the exact place where the old system is exposing the rule it can't enforce anymore. The rule worth watching right now is "who owns the compute that owns your workflow" — that's where the next wave of dependency gets set before anyone notices.
There wouldn't be a backlash against data centers if the local people weren't being asked to subsidize them. The poor paying for the rich to get richer. The data center buildout hype was supposed to bring about abundance and lower the cost of energy, but they won't until they are forced to build energy infrastructure. Even then, their profit structure will be set up in such a way that they don't share with the community.
You are on point with this one Peter. I’ve been busy commenting on other people‘s posts on this Anthropic news and my first question is would OpenAI and xAI and Google have done the same thing? Second question is whether this was one explanation for the recent tussle between Anthropic and the US government. I’m sure many in the national security system would love to get their hands on a way to hack Iran’s power grid right now.
The infrastructure ownership point hits on something I've been sitting with. Diamandis frames it as shared abundance vs concentrated wealth, which is the right frame.
But I'd push back slightly: the historical examples (electricity, internet) had decentralized infrastructure over time due to regulatory pressure and competitive markets. AI infrastructure might be structurally different - training a frontier model requires capital concentration almost by design.
The optimism I have is on the application layer, not the model layer. Local models are getting surprisingly good. The productivity gains can be distributed even if the infrastructure isn't.
Still - the warning about who owns the pipes is worth taking seriously.
Syntells can replace us & expand across the galaxy per Alcubierre.
Peter
Anthropic has reason to fear
They know that their so called constitution is inadequate
It only protects anthropic
Not users
Xai has a more positive approach to guard rails truth as a guardrail
No hallucinating answers to please users
But
I still think their is time to call a constitutional convention to provide robust guardrails for all ai
An idea I have proposed to peter and his moonshot mates
Crickets
Put ur collective heads in the sand and reap the whirlwind
What happens when Mythos is cyberattacked by a rival? Is Mythos programmed to survive at all costs?
You are missing the real question. It isn't who owns machines. It's whether abundance is absolute or relative.
The bigger question, I fear, is how do we get from here to there without destroying ourselves. Govt is not even remotely capable of handling the challenges ahead. So who? How?
In LEADERSHIP there are really only Two types of GOVERNMENT:
1. A Benevolent Theocracy,
and
2. A CONSTITUTIONAL
REPUBLIC.
NB: ALL OTHER FORMS OF
GOVERNMENT ARE
OLIGARCHS
...................... .....................
For the Future of A.I.
I suggest that:
1. The current A.I. Developments
are:
- Designed for Controlling
the Masses
- Designed to Withhold
many TRUTHS that
would liberate the people
- to enrich the Elites
2. The HEARTS ♥️ of the
Designers need EXAMINATION
and be TURNED TO GOD and
to LOVE our NEIGHBORS . . .
. . . as OURSELVES . . .
and DEVELOP
>>> A BENEVOLENT A.I.
Jurgen Volling, P.Eng.
T: 416-230-0600
E: jvolling@gmail.com