Dearest readers,
Artificial Intelligence has become the darling of the tech world; everyone whispers its name with that breathless mixture of awe and anxiety usually reserved for forbidden lovers or luxury handbags on clearance. We are told it will transform our lives; rewrite our jobs; conquer inefficiency; maybe even solve world peace before lunch. Yet, beneath this shimmering surface lies a cliff; a very steep cliff; made of silicon, supply chains, and a healthy dose of geopolitical fantasy.
This is the tale of that cliff; and how close we are to sliding right off it. You will not find any of this information in the mainstream, or alt-stream ahha. I can tell you where I got it (no I did not get info on potential doom of AI from AI), but I would rather your find it on your own because it will empower you if this topic concerns you in any way. ;)
The Silicon Cliff; A Love Story with Commitment Issues
AI today, especially the large language models we chat with for fun, flirtation, and frantic midnight troubleshooting, is dazzling on the surface. It codes; it summarizes; it even drafts breakup texts with unsettling emotional intelligence. The hype feels intoxicating; society is already pre-writing the wedding vows for the “AI revolution.”
But here is the messy truth, the kind I have no patience sugar coating: AI is still a creature of the information realm; not the physical one. It can reshuffle words and numbers entirely, but ask it to fix your dishwasher or run a mining operation and it becomes that friend who offers to “supervise” while you do the actual work. I actually tried to use it to fix my mom's dishwasher btw- lots of confusion there. Real-world transformation requires robotics; robotics requires hardware; hardware requires the most fragile global machine humanity has ever stitched together.
That machine is groaning.
A Hardware House of Cards; Built on Wishes and Nanometers
AI’s brain lives inside impossibly tiny semiconductors; chips so small they might as well be magic stones from a fantasy novel. Anything under ten nanometers relies on a production ecosystem so delicate it feels like a cosmic joke. Seriously, you can either laugh or stand dumbfounded at how this is even possible.
To create a single advanced chip, the world relies on:
thirty thousand individual parts;
one hundred thousand supply chain steps;
nine thousand specialized companies;
over four thousand five hundred single point failures;
and a cooperation pact spanning eighty countries.
A breakup with even one mid-sized country and poof; the magic stops. You cannot whip up replacements in a garage; recreating this ecosystem would take at least fifteen years. Not “Silicon Valley years,” where people claim teleportation is close; real years.
The Ghost of 14 Nanometers Now for the plot twist.
The chips powering today’s AI were originally built for gaming consoles. I am talking about graphics cards that are now large and complex enough to be computers onto themselves. Ever seen a RTX of any kind hooked up to a small SBC? You would be like, which one is the computer and which one is a component?
True AI-specific chips still do not exist; they live on PowerPoints and optimistic supply chain projections. Timelines for their debut drift toward 2029, but anyone who has met reality knows better.
If today’s delicate semiconductor ecosystem falters, the industry would fall back to fourteen nanometer chips. That is like switching from a racing bike to a rusty tricycle; still technically transportation, but do not expect to win anything.
Running modern AI models on older chips would mean:
efficiency plunges by ninety to ninety five percent;
performance collapsing to last year’s levels;
energy costs exploding fivefold.
AI would still exist, but only in a weaker, hungrier, dramatically more expensive form. The magic would sputter; the hype bubble would hiss; and the world would quietly pretend it never made those bold predictions.
A Conversation Happening Before the Technology Exists
The charming irony is that we are debating the philosophical, ethical, and societal implications of a technology that may not survive its own childhood. The hardware is brittle; the supply chain is shaky; the geopolitics are turning medieval. Humanity is arguing about AI’s future rights while forgetting that this entire empire rests on nanometers and international group projects; we all know how those usually end.
Experts now whisper that the acceleration they feared two years ago has slowed; the timeline has loosened; the world has probably bought itself until the 2040s before the next existential AI panic returns.
Until we solve the hardware bottleneck; until the semiconductor tower stops wobbling; AI remains a breathtaking illusion balanced on unstable ground.
Imagine a Formula 1 car that goes two hundred miles per hour, but only if thousands of mechanics across eighty countries cooperate without interruption. If even one mechanic oversleeps, the car breaks down; suddenly you are racing in a decade-old model burning five times the fuel. That, dearest readers, is today’s AI industry.
Novel; glamorous; brilliant; and one supply shock away from rolling back into the garage with a very confused expression. This is the real tension of the moment; not whether AI will replace us, but whether its own infrastructure can survive long enough to even try.
Now, in order for this to not be true, I think that any country that wants to even maintain the current level of AI power would have to do the following:
Guess what, dearest readers? If these points sound familiar, it is precisely because they are doing this right now! Here we are, sitting and thinking they cannot justify spending all this money on AI physical infrastructure at all based on the current revenue and pricing. But, if you factor in what we now know about the hardware side supply chain fragility and a supply that is not long for this world, what they are doing makes perfect sense. Also, it makes perfect sense for companies to tinker with smaller, more efficient or task/sector oriented LLM's as well as better RAG (look it up babes) approaches.
Bottom line, when the supply chain likely breaks, do not expect it will be cheap or even easy for you to use the latest version of Nano Banana Pro or the latest Kling AI video to generate low brow clickbait easily for your YouTube Shorts channel side hustle. They will make it more expensive or lock it down completely unless they develop ways to make it less resource intensive. They will need that power for the office, the government and the military as we might find ourselves in the world of cutting edge hardware scarcity. So, I will make a prediction that, if you care about AI and use it now for text, image or video generation, you may want to build or buy a PC in the used or new market, whatever you can afford, that can at least run a decent LLM now. I was curious so I looked around, and honestly most GPU's that would do the trick are closer to 1,000-1,200 CAD here in Canada (obviously ahha). So, you may be looking at a custom PC upwards of 2,000 CAD when all is said and done.
Or, the most fragile supply chain ecosystem ever does not get sabotaged, harmed or destroyed even if we go into World War 3, right? Aaand, everything will be just peachy, right? ;)
Sigh...
What do you think? There is plenty to consider here, and not just whether to do calls or puts on NVIDIA with a long expiration date...
Artificial Intelligence has become the darling of the tech world; everyone whispers its name with that breathless mixture of awe and anxiety usually reserved for forbidden lovers or luxury handbags on clearance. We are told it will transform our lives; rewrite our jobs; conquer inefficiency; maybe even solve world peace before lunch. Yet, beneath this shimmering surface lies a cliff; a very steep cliff; made of silicon, supply chains, and a healthy dose of geopolitical fantasy.
This is the tale of that cliff; and how close we are to sliding right off it. You will not find any of this information in the mainstream, or alt-stream ahha. I can tell you where I got it (no I did not get info on potential doom of AI from AI), but I would rather your find it on your own because it will empower you if this topic concerns you in any way. ;)
The Silicon Cliff; A Love Story with Commitment Issues
AI today, especially the large language models we chat with for fun, flirtation, and frantic midnight troubleshooting, is dazzling on the surface. It codes; it summarizes; it even drafts breakup texts with unsettling emotional intelligence. The hype feels intoxicating; society is already pre-writing the wedding vows for the “AI revolution.”
But here is the messy truth, the kind I have no patience sugar coating: AI is still a creature of the information realm; not the physical one. It can reshuffle words and numbers entirely, but ask it to fix your dishwasher or run a mining operation and it becomes that friend who offers to “supervise” while you do the actual work. I actually tried to use it to fix my mom's dishwasher btw- lots of confusion there. Real-world transformation requires robotics; robotics requires hardware; hardware requires the most fragile global machine humanity has ever stitched together.
That machine is groaning.
A Hardware House of Cards; Built on Wishes and Nanometers
AI’s brain lives inside impossibly tiny semiconductors; chips so small they might as well be magic stones from a fantasy novel. Anything under ten nanometers relies on a production ecosystem so delicate it feels like a cosmic joke. Seriously, you can either laugh or stand dumbfounded at how this is even possible.
To create a single advanced chip, the world relies on:
thirty thousand individual parts;
one hundred thousand supply chain steps;
nine thousand specialized companies;
over four thousand five hundred single point failures;
and a cooperation pact spanning eighty countries.
A breakup with even one mid-sized country and poof; the magic stops. You cannot whip up replacements in a garage; recreating this ecosystem would take at least fifteen years. Not “Silicon Valley years,” where people claim teleportation is close; real years.
The Ghost of 14 Nanometers Now for the plot twist.
The chips powering today’s AI were originally built for gaming consoles. I am talking about graphics cards that are now large and complex enough to be computers onto themselves. Ever seen a RTX of any kind hooked up to a small SBC? You would be like, which one is the computer and which one is a component?
True AI-specific chips still do not exist; they live on PowerPoints and optimistic supply chain projections. Timelines for their debut drift toward 2029, but anyone who has met reality knows better.
If today’s delicate semiconductor ecosystem falters, the industry would fall back to fourteen nanometer chips. That is like switching from a racing bike to a rusty tricycle; still technically transportation, but do not expect to win anything.
Running modern AI models on older chips would mean:
efficiency plunges by ninety to ninety five percent;
performance collapsing to last year’s levels;
energy costs exploding fivefold.
AI would still exist, but only in a weaker, hungrier, dramatically more expensive form. The magic would sputter; the hype bubble would hiss; and the world would quietly pretend it never made those bold predictions.
A Conversation Happening Before the Technology Exists
The charming irony is that we are debating the philosophical, ethical, and societal implications of a technology that may not survive its own childhood. The hardware is brittle; the supply chain is shaky; the geopolitics are turning medieval. Humanity is arguing about AI’s future rights while forgetting that this entire empire rests on nanometers and international group projects; we all know how those usually end.
Experts now whisper that the acceleration they feared two years ago has slowed; the timeline has loosened; the world has probably bought itself until the 2040s before the next existential AI panic returns.
Until we solve the hardware bottleneck; until the semiconductor tower stops wobbling; AI remains a breathtaking illusion balanced on unstable ground.
Imagine a Formula 1 car that goes two hundred miles per hour, but only if thousands of mechanics across eighty countries cooperate without interruption. If even one mechanic oversleeps, the car breaks down; suddenly you are racing in a decade-old model burning five times the fuel. That, dearest readers, is today’s AI industry.
Novel; glamorous; brilliant; and one supply shock away from rolling back into the garage with a very confused expression. This is the real tension of the moment; not whether AI will replace us, but whether its own infrastructure can survive long enough to even try.
Now, in order for this to not be true, I think that any country that wants to even maintain the current level of AI power would have to do the following:
- Hoard every single new GPU that comes out while they still do.
- Build up data centers, AI factories, and Hyperscale Data Centers as fast as possible in order to survive future breakdowns in the hardware side supply chains.
- Governments have to invest in this together with the private sector.
- The private sector has to use us, the public, to train AI in all imaginable ways- audio, video, coding, writing, agentic activities- we use it, we pay for the privilege, they get feedback and training data. Once the supply chain is shaken, they can take most of what we have achieved out of our hands and keep the best of AI strictly for government and corporate use, if cutting edge hardware becomes scarce for years to come. Good for them, not for us.
Guess what, dearest readers? If these points sound familiar, it is precisely because they are doing this right now! Here we are, sitting and thinking they cannot justify spending all this money on AI physical infrastructure at all based on the current revenue and pricing. But, if you factor in what we now know about the hardware side supply chain fragility and a supply that is not long for this world, what they are doing makes perfect sense. Also, it makes perfect sense for companies to tinker with smaller, more efficient or task/sector oriented LLM's as well as better RAG (look it up babes) approaches.
Bottom line, when the supply chain likely breaks, do not expect it will be cheap or even easy for you to use the latest version of Nano Banana Pro or the latest Kling AI video to generate low brow clickbait easily for your YouTube Shorts channel side hustle. They will make it more expensive or lock it down completely unless they develop ways to make it less resource intensive. They will need that power for the office, the government and the military as we might find ourselves in the world of cutting edge hardware scarcity. So, I will make a prediction that, if you care about AI and use it now for text, image or video generation, you may want to build or buy a PC in the used or new market, whatever you can afford, that can at least run a decent LLM now. I was curious so I looked around, and honestly most GPU's that would do the trick are closer to 1,000-1,200 CAD here in Canada (obviously ahha). So, you may be looking at a custom PC upwards of 2,000 CAD when all is said and done.
Or, the most fragile supply chain ecosystem ever does not get sabotaged, harmed or destroyed even if we go into World War 3, right? Aaand, everything will be just peachy, right? ;)
Sigh...
What do you think? There is plenty to consider here, and not just whether to do calls or puts on NVIDIA with a long expiration date...