Nothing happened to them, they're still around; just consolidated into industrial operations.
The "twist" is they rot as e-waste every 18 months when newer models arrive, generating roughly 30,000 metric tonnes of eWaste annually[0] with no recycling programmes from manufacturers (like Bitmain)... which is comparable to the entire country of the Netherlands.
Turns out the decentralised currency for the people is also an environmental disaster built on planned obsolescence. Who knew.
If the power was used over the whole year (and not just one hour)
(2600 MWh / year) / (24 * 365 h/year) = 0.29 MWh = 296 kWh. Thats like hair dryer levels of power consumption (if the hair dryver was left on all the time)
Why are you even trying to argue energy consumption when the topic is eWaste due to bitcoin ASICs?
Even if we continue down this route, its something like 15% of global stock transactions going through NYSE, per transaction its extremely efficient when compared to Ethereum; but thats not the argument anyway- its that the hardware used for mining is barely useful outside of that use-case, and the shelf-life is very low to boot.
If there was a use-case, we’d have found it by now, since 30,000 Tonnes a year of it ends up in landfills, surely someone would dig it out or buy it if it had utility.
AI, obviously! A bubble doesn't mean demand vanishes overnight. There is - at current price points - much more demand than supply. That means the market can tolerate price hikes whilst keeping the accelerators busy. It seems likely that we're still just at the start of AI demand as most companies are still finding their feet with it, lots of devs still aren't using it at all, lots of business workflows that could be automated with it aren't and so on. So there is scope for raising prices a lot as the high value use cases float to the top, maybe even auctioning tokens.
Let's say tomorrow OpenAI and Anthropic have a huge down round, or whatever event people think would mark the end of the bubble. That doesn't mean suddenly nobody is using AI. It means they have to rapidly reduce burn e.g. not doing new model versions, laying off staff and reducing the comp of those that remain, hiking prices a lot, getting more serious about ads and other monetized features. They will still be selling plenty of inferencing.
In practice the action is mostly taking place out of public markets. We won't necessarily know what's happening at the most exposed companies until it's in the rear view mirror. Bubbles are a public markets phenomenon. See how "ride sharing"/taxi apps played out. Market dumping for long periods to buy market share, followed by a relatively easy transition to annual profitability without ever going public. Some investors probably got wiped along the way but we don't know who exactly or by how much.
Most likely outcome: AI bubble will deflate steadily rather than suddenly burst. Resources are diverted from training to inferencing, new features slow down, new models are weaker and more expensive than new models and the old models are turned off anyway. That sort of thing. People will call it enshittification but it'll really just be the end of aggressive dumping.
There may not be that much demand at a price that yields profit. Demand at current heavily subsidized “the first dose is always free” prices is not a great indicator unless they find some way to make themselves indispensable for a lot of tasks for a lot of people. So far, they haven’t.
Yes if/when prices rise there'll be demand destruction but I think demand will keep rising for the foreseeable future anyway even incorporating that. Lower value use cases like vibe coding hobby apps might fall by the wayside because they become uneconomic but the tokens will be soaked up by bigger enterprises that have found ways to properly integrate it at scale into their businesses. I don't mean Copilot style Office plugins but more business-specific stuff that yields competitive advantage.
You’re just repeating their predictions. Investors are starting to get nervous that there’s no real proof these things could justify burning a Mt. Everest sized pile of $100 bills to achieve.
Yes it's only a prediction based on what I'm seeing. And I'm not disagreeing with the investors that there's overinvestment right now. Prices need to rise, spending on R&D needs to fall for this stuff to make economic sense. I'm only arguing that there's plenty of demand, and assuming price rises happen smoothly over not too short of a period, any demand destruction at the lower levels will be quickly counter-balanced by demand creation at higher value-add levels.
It's also possible non-tech industries just have a collective imagination failure and can't find use cases for AI, but I doubt it.
I know there is demand — I even know a few people in high-level dev roles, one easily in the 99th percentile for pay, that were taken from their regular, important dev tasks to make agents for paying clients.
I’m not worried about the technology flourishing. I’m worried about my fucking retirement, know what I mean? The question isn’t whether there is demand, it’s how much demand there is, because they’re betting on having all the demand. I don’t have a ton of money. People getting this wrong in a way my financial advisor can’t outmaneuver is an existential threat to my ability to not live in crumbling public housing in a few decades.
We’re talking about this needing to meaningfully move towards making whole digit percentages of the US GDP, soon. Not only are these initiatives largely unprofitable, they’re increasing their expenses based on hopes and vibes. I think a whole lot of people are so focused on short-term gains and being king of the hill that sustainability is a distant afterthought — just like it was in the .com era. I have zero faith in the current cohort of tech leaders to get this right.
Anyone who regularly tries to rent GPUs on VPS providers knows that they often sell out. This isn't a market with lots of capacity nobody needs. In the dot.com bubble there was lots of dark fiber nobody was using. In this bubble, almost every high-end GPU is being used fully by someone.
We can use the GPUs for research (64-bit scientific compute), 3d graphics, a few other things. We programmers will reconfigure them to something useful.
At least, the GPUs that are currently plugged in. A lot of this bullshit bubble crap is because most of those GPUs (and RAM) is sitting unplugged in a warehouse, because we don't even have enough power to turn all of them on.
So if your question is how to use a GPU... I got plenty of useful non-AI related ideas. But only if we can plug them in.
I wouldn't be surprised if many of those GPUs are just e-waste, never to turn on due to lack of power.
> I wouldn't be surprised if many of those GPUs are just e-waste, never to turn on due to lack of power.
That's my fear.
The problem is these GPUs are specifically made for datacenters, So it's not like your average consumer is going to grab one to put into their gaming PCs.
I also worry about what the pop ends up doing to consumer electronics. We'll have a bunch of manufacturers that have a bunch of capacity that they can no longer use to create products which people want to buy and a huge backstop of second hand goods that these liquidated AI companies will want to unload. That will put chip manufactures in a place where they'll need to get their money primarily from consumers if they want to stay in business. That's not the business model that they've operated on up until this point.
We are looking at a situation where we have a bunch of oil derricks ready to pump, but shut off because it's too expensive to run the equipment making it not worth the energy.
> As it turns out Nvidia's H100, a card that costs over $30,000 performs worse than integrated GPUs in such benchmarks as 3DMark and Red Dead Redemption 2
I predict there's going to be a niche opening up for companies to recycle the expensive parts of all these compute hardware that AI companies are currently buying and will probably be obsolete/depreciated/replaced in the next 2-5 years. The easiest example is RAM chips. There will be people desoldering those ICs and putting them on DDR5 sticks to resell to the general consumer market.
A technological arms race just occurred in front of your eyes for the past 5 years and you think they're going to let the stockpile fall into civilian hands?
In 2 years the next generation chips will be released and th se chips will be obsolete.
That's truly e-waste. Now in practice, we programmers find uses of 10+ year old hardware as cheap webhosta, compiler/build boxes, Bamboo, unit tests, fuzzers and whatever. So as long as we can turn them on we programmers can and will find a use.
But because we are power constrained, when the more efficient 1.8nm or 1.5nm chips get released (and when those chips use 30% or less power), no one will give a shit about the obsolete stockpile.
In what sense? Not competitive for chat bot providers to use? Is that a metric that matters?
> when the more efficient 1.8nm or 1.5nm chips get released
What if they don't get released? You don't have a broad and competitive set of players providing products in this realm. How hard would it be to stop this?
> no one will give a shit about the obsolete stockpile.
You have lived your life with ready access to cutting edge resources. You ever wonder how long that trend could _possibly_ last?
As in: the 1.5nm or 1.8nm GPUs will use less power and therefore can actually be plugged in.
We are power constrained. The GPUs of this generation can't even be plugged in yet because of these power constraints.
When power is a problem, getting lower power GPUs in is a priority. The 1.8nm and 1.5nm next generation is already in production, and will likely launch before these massive GPU stockpiles are used.
And then what? Why plug in last generations crap when the next generation is shipping?
--------
Todays GPUs have to actually launch and be deployed while they are useful. Otherwise they could fully be obsolete and lose significant value.
I assume even really out of date cards and racks will readily find some use, when the present-day alternative costs ~$100k for a single card. Just have to run them on a low-enough basis that power use is not a significant portion of the overall cost of ownership.
It’ll be interesting to see what people come up with to get conventional scientific computing workloads to work on 16 bit or smaller data types. I think there’s some hope but it will require work.