Are we repeating the telecoms crash with AI datacenters?
martinalderson.com
martinalderson.com
I doubt it.
And what if the technology to locally run these systems without reliance on the cloud becomes commonplace, as it now is with open source models? The expensive part is in the training of these models more than the inference.
Absolutely, not only are most AI services free but even the paid portion is coming from executives mandating that their employees use AI services. It's a heavily distorted market.
And a majority of those workers do not reveal their AI usage, so they either take credit for the faster work or use the extra time for other activities, which further confounds the impact of AI.
This is also distorting the market, but in other ways.
People are missing the forest for the trees here. Being the go to consumer Gen AI is a trillion+ dollar business. How many 10s of billions you waste on building unnecessary data centers is a rounding error. The important number is your odds of becoming that default provider in the minds of consumers.
So voice assistants backed by very large LLMs over the network are going to win even if we solve the (substantial) battery usage issue.
I haven't seen any evidence that any Gen AI provider will be able to build a moat that allows for this.
Some are better than others at certain things over certain time periods, but they are all relatively interchangeable for most practical uses and the small differences are becoming less pronounced, not more.
I use LLMs fairly frequently now and I just bounce around between them to stay within their free tiers. Short of some actual large breakthrough I never need to commit to one, and I can take advantage of their own massive spends and wait it out a couple of years until I'm running a local model self-hosted with a cloudflare tunnel if I need to access it on my phone.
And yes, most people won't do that, but there will be a lot of opportunity for cheap providers to offer that as a service with some data center spend, but nowhere near the massive amounts OpenAI, Google, Meta, et al are burning now.
LLMs complete text. Every query they answer is giving away the secret ingredient in the shape of tokens.
As a regular user, it becomes increasingly frustrating to have to remind each new chat “I’m working on this problem and here’s the relevant context”.
GenAI providers will solve this, and it will make the UX much, much smoother. Then they will make it very hard to export that memory/context.
If you’re using a free tier I assume you’re not using reasoning models extensively, so you wouldn’t necessarily see how big of a benefit this could be.
I used ChatGPT for every day stuff, but in my experience their responses got worse and I had to wait much longer to get them. I switched to Gemini and their answers were better and were much faster.
I don’t have any loyalty to Gemini though. If it gets slow or another provider gives better answers, I’ll change. They all have the same UI and they all work the same (from a user’s perspective).
There is no moat for consumer genAI. And did I mention I’m not paying for any of it?
It’s like quick commerce, sure it’s easy to get users by offering them something expensive off of VC money. The second they raise prices or offer degraded experience to make the service profitable, the users will leave for another alternative.
I do wonder, if you (and the commenter you replied to) think this is a good thing, will you be OK with a data center springing up in your neighbourhood, driving up water or power prices, emitting CO2? Then if SOTA LLMs become efficient enough to run on a smartphone will you be OK with a data center bailout coming from your tax dollars?
[0]: https://www.mckinsey.com/industries/technology-media-and-tel...
I use LLM’s all day and a highly doubt this. I’d love to hear your argument for how this plays out.
Markets that have a default provider are basically outliers (desktop OS, mobile OS, search, social networks, etc).
All other industries don't have a single dominant supplier who is the default provider.
I am skeptical that this market is going to be one of the outliers.
All the other markets with a default provider basically rely on network effects to become and remain the default provider.
There is nothing here (in this market) that relies on network effects.
The local open source argument doesn't hold water for me -- why does anyone buy Windows, Dropbox, etc when there's free alternatives?
Installing an OS is seen as a hard/technical task still. Installing a local program, not so much. I suspect people install LLM programs from app stores without knowing if they are calling out to the internet or running locally.
No one buys Windows - it comes with the PC.
If people were shipped blank computers and told to order the OS separately, they wouldn't be buying Windows at the current price point.
You sometimes see this with real live humans who have lived in multiple counties.
Pay no attention to those fopheads from Kent. We speak proper British English here in Essex
Also very common with... most Canadians. We officially use an English closer to British English (Zed not zee, honour not honor) however geographically and culturally we're very close to the US.
At school you learn "X, Y, Zed". The toy you buy your toddler is made for the US and Canadian market and sings "X, Y, Zee" as does practically every show on TV. The dictionary says it's spelled "colour" but most of the books you read will spell it "color". Most people we communicate with are either from Canada or the US, so much of our personal communication is with US English.
But also there are a number of British shows that air here, so some particularly British phrases do sneak in to a lot of people's lexicon.
See a similar thing in the way we measure things.
We use celsius for temperature but most of our thermostats default to Fahrenheit and most cookbooks are primarily in imperial measures and units because they're from the US. The store sells everything in grams and kilograms, but most recipes are still in tablespoons/cups/etc.
Most things are sold in metric, but when you buy lumber it's sold in feet, and any construction site is going to be working primarily in feet and inches.
If anything I expect any AI-written content would be more consistent about this than I usually am.
See also how all (?) Brits pronounce Gen Z in the American way (ie zee, not zed).
Bullet points hell, a table that feels it came straight out of grok.
Some people are not from usa or England.
Past successes like Google encourage hope in this strategy. Sure, it mostly doesn't work. Most of of everything that VCs do doesn't work. Returns follow a power law, and a handful of successes in the tail drive the whole portfolio.
The key problem here doesn't lie in the fact that this strategy is being pursued. The key problem is that it is rare for first mover advantages to last with new technologies. That's why Netscape and Yahoo! aren't among the FAANGs today. The long-term wins go to whoever successfully create a sufficient moat for themselves to protect lasting excess returns. And the capabilities of each generation of AI leapfrogs the last so well that nobody has figured out how to create such a moat.
Today, 3 years after launching the first LLM chatbot, OpenAI is nowhere near as dominant as Netscape was in late 1997, 3 years after launching Netscape Navigator. I see no reason to expect that 30 years from now OpenAI will be any more dominant than Netscape is today.
Right now companies are pouring money into their candidates to win the AI race. But if the history of browsers repeats itself, the company that wins in the long-term would launch in about a year from now, focused on applications on top of AI. And its entrant into the AI wars wouldn't get launched until a decade after that! (Yes, that is the right timeline for the launch of Google, and Google's launch of Chrome.)
Investing in silicon valley is like buying a positive EV lottery ticket. An awful lot of people are going to be reminded the hard way that it is wiser to buy a lot of lottery tickets, than it is to sink a fortune into a single big one.
Incorrect. There were about 150 millions Internet users in 1998, or 3.5% of total population. The number grew 10 times by 2008 [0]. Netwcape had about 50% of the browser market at the time [1]. In other words, Netscape dominated a small base and couldn’t keep it up.
ChatGPT has about 800 millions monthly users, or already 10% of total current population. Granted, not exclusively. ChatGPT is already a household name. Outside of early internet adopters, very few people knew who Netscape or what Navigator was.
[0] https://archive.globalpolicy.org/component/content/article/1...
[1] https://www.wired.com/1999/06/microsoft-leading-browser-war/...
According to https://en.wikipedia.org/wiki/Usage_share_of_web_browsers, Netscape had 60-70% market share. According to https://firstpagesage.com/reports/top-generative-ai-chatbots..., ChatGPT currently has a 60% market share.
But I consider the enterprise market a better indicator of where things are going. As https://menlovc.com/perspective/2025-mid-year-llm-market-upd... shows, OpenAI is one of a pack of significant competitors - and Anthropic is leading the pack.
Furthermore my point that the early market leaders are seldom the lasting winners is something that you can see across a large number of past financial bubbles through history. You'll find the same thing in, for example, trains, automobiles, planes, and semiconductors. The planes example is particularly interesting. Airline companies not only don't have a good competitive moat, but the mechanics of chapter 11 mean that they keep driving each other bankrupt. It is a successful industry, and yet it has destroyed tons of investment capital!
Despite your quibbles over the early browser market, my broader point stands. It's early days. AI companies do not have a competitive moat. And it is way to premature to reliably pick a winner.
Netscape in 1997/1998 had about 90% of the target market.
OpenAI today has about 30% of the target market, maybe? (seeing as how every single Windows installation comes with copilot chat already, it's probably less. Every non-tech user I know has already used copilot because it was bundled and Windows prompted them into using it with a popup. Only one of those non-tech users even heard of OpenAI, maybe 50% of them have heard that there are alternatives to Copilot, but they still aren't using those alternatives)
I agree. Right now a lot of AI tools are underpriced to get customers hooked, then they'll jack up the prices later. The flaw is that AI does not have the ubiquitous utility internet access has, and a lot of people are not happy with the performance per dollar TODAY, much less when prices rise 80%. We already see companies like Google raising prices stating it's for "AI" and we customers can't opt out of AI and not pay the fee.
At my company we've already decided to leave Google Workspace in the spring. GW is a terrible product with no advanced features, garbage admin tools, uncompetitive pricing, and now AI shoved in everywhere and no way to granularly opt out of a lot of it. Want spell check? Guess what, you need to leave Gemini enabled! Shove off, Google.
Yea, I think this is wrong. The analogy is more like the App Store, in that there is very little to do currently other than a better Google Search with the product. The bet is that over time (short time) there are much more financially valuable use cases with a more mature ecosystem and tech.
We're in the "dial up era" of AI.
Unlike the smartphone adoption era where everything happened rather rapidly, we're in this weird place where labs have invented a bunch of model categories, but they aren't applicable to a wide variety of problems yet.
The dial up -> broadband curve took almost a decade to reach penetration and to create the SaaS market. It's kind of a fluke that Google and Amazon came out of the dial up era - that's probably what investors were hoping for by writing such large checks.
They found chat as one type of product. Image gen as another. But there's really not much "native AI" stuff going about. Everyone is bolting AI onto products and calling it a done day (or being tasked with clueless leadership to do it with even worse results).
This is not AI. This is early cycle WebVan-type exploration. The idea to use AI in a given domain or vertical might be right, but the tools just don't exist yet.
We don't need AI models with crude APIs. We need AI models we can pull off the shelf, fine tune, and adapt to novel UI/UX.
Adobe is showing everyone how they're thinking about AI in photoshop - their latest conference showed off AI-native UX. And it was really slick. Dozens of image tools (relighting, compositing, angle adjustment) that all felt fast, magical, and approachable as a beginner. Nobody else is doing that. They're just shoving a chat interface in your hands and asking you to deal with it.
We're too early. AI for every domain isn't here yet.
We're not even in the dialup era, honestly.
I'd expect the best categories of AI to invest in with actually sound financials will be tool vendors (OpenRouter, FAL, etc.) and AI-native PLG-type companies.
Enterprise is not ready. Enterprise does not know what the hell to do with these APIs.
I'm going through the process of buying a home, and the amount of help its given is incredible. Analyzing disclosures, loan estimates, etc. Our accountant charges $200 an hour to basically confirm all the same facts that ChatGPT already gave us. We can go into those meetings prepped with 3 different scenarios that ChatGPT already outlined, and all they have to do is confirm.
Its true that its not always correct, but, I've also had paid specialists like real estate agents and accountants give me incorrect information, at the cost of days of scheduling, and hundreds of dollars. They also aren't willing to answer questions at 2am in the morning.
The sellers already had an inspection done. The full report is over 100 pages
You can get “real” inspections done but they cost thousands of dollars and take a full or more day to do with multiple subject matter experts. Almost no one does this.
Waiving inspection other than for major material defect is what I’ve done for all my real estate transactions. I’m not putting in an offer to nickel and dime someone over trivial bullshit like a busted GFCI circuit. My offer simply accounts for the trivial odds and ends I’ll have to take care of. Plus I’d much rather get the work done myself since I don’t trust a seller to do anything but the bare minimum.
Every one of my friends who have had five figures or more of surprise repair work on homes they purchased all had an inspection done. None of those could have found the various hidden damages for those buildings short of destructive stuff like pulling drywall out or lifting up shingles from a roof. Don’t worry though, the inspectors found stuff like a bathroom faucet with a crack in the knob.
If you look at the standard offer document for waiving inspection it's pretty easy to walk it back. You're simply waiving a contingency - you can typically still inspect the property itself. I'm sure if you get way off the beaten path you are correct, but almost no one is engaging in totally non-standard contracts where I'm at.
I'm curious what liability you think would apply for an inspection that misses whatever it may be that ends up in dispute after the transaction closes - since the whole point in the inspection is finding that beforehand? If I find a material defect in the foundation after I close - it won't matter if I had an inspection or not. Unless I can prove the seller knew about it and failed to disclose.
And if I ever sell any properties - I will be pretty loath to sell to anyone demanding an inspection contingency. They are almost always used for nickel and dime BS that I really don't have time for. If you walk the place, get your inspector to do so too, and come up with a punch list and still want to make an offer, discount it appropriately and fix it yourself after you close. It's nearly always either pointless or used as a negotiation tool after the fact due to the fact buyers can expect a seller to not want to walk away from the middle of a transaction (sunk cost/time). I'd much rather take an offer at 5% less up-front than deal with someone wasting 30-45 days on the market and my time dealing with trivial items.
Waiving inspections means even if they do, they’re not on the hook anymore.
Which is why you get an inspection.
I agree. Wait until it's $249 a month. You'll feel differently.
Competition and local SLMs will also provide a counter to massive price increase
Or much like what is going to happen with Alexa, it just dies because the cost of the service is never going to align with “what the market can bear”. Even at $75/mo, the average person will probably stop being lazy and just go back to doing 10 minutes worth of searching to find answers to basic questions.
In any given day I never have no access to free LLM help.
Since all the models are converging onto the same level of performance, I mostly can't even tell responses from ChatGPT and Claude apart.
> Right now a lot of AI tools are underpriced to get customers hooked, then they'll jack up the prices later.
Good luck with that. I mean it.
The ChatAI TAM is now so saturated with free offerings that the first supplier to blink will go out of business before they are done blinking.
I see people (like sibling reply to parent) boasting about the amount of value they get from the $20/m subscription, but I don't see how that is $20 better than just using the free ChatAIs.
The only way out of the red for ChatAI products is to add in advertising slowly; they have to boil the frog. A subscription may have made sense when ChatGPT was the only decent game in town. Subscriptions don't make sense now - I can get 90% of the value of a ChatAI for 0% of the cost.
I don't. This is simply the "drug dealer" model where the first hit is free. They know that once people are addicted, they will keep coming back.
The question of course is, will they keep coming back? I think they very much will. There are indications that GenAI adoption is already increasing labor producitivity labor improvements at a national scale, which is quite astounding for a technology just 3 years old: https://news.ycombinator.com/item?id=46061369
Imagine a magic box where you put in some money and get more productivity back. There is no chance Capitalism (with a capital "C") is going to let such a powerful growth machine wither on the vine. This mad AI rush is all about that.
IMHO the investors are betting on a winner-takes-it-all market and that some magic AGI will be coming out of OpenAI or Anthropic.
The questions are:
How much money can they make by integrating advertising and/or selling user profiles?
What is the model competition going to be?
What is the future AI hardware going to be - TPUs, ASICs?
Will more people have powerful laptops/desktops to run a mid-sized models locally and be happy with it?
The internet didn't stop after the dotcom crash and the AI wont stop either should there be a market correction.
By itself, this doesn't tell us much.
The more interesting metric would be token use comparison across free users, paid users, API use, and Azure/Bedrock.
I'm not sure if these numbers are available anywhere. It's very possible B2B use could be a much bigger market than direct B2C (and the free users are currently providing value in terms of training data).
But currently, aside from generating "creative media", I'd say I'm pretty much opposed to LLM tools. They have yet to demonstrate any value to me at work or with respect to the areas of research I am interested in, and given the kind of statistical mechanism that they are, I do not believe they are capable of doing so.
Interesting take, because I'm the opposite of it. My biggest use case is getting into a completely new topic, as it's the most frictionless starting point for most of the queries. Then I look around based on the rudimentary knowledge that I can gather from LLMs. However, I'm completely opposed to any sort of creative media created by LLMs and try to avoid it as much as I can (music, images, and etc.).
Also, it has become the natural workflow for me to throw bunch secondary priority work stuff to Claude and let it do its things, while I focus on the important stuff.
My point is, everyone finds a way to use it. Some are opposed to specific things, others are using other parts.
That's the only AI I use anywhere near weekly. I have tried claude a few times, it was useless at helping me with my questions. I haven't really been back.
I would say if executed well the revenue per user could be at least an order of magnitude more than Google search ads as the ads could be much more convincing and the information density is higher in chat.
But the AI providers are betting, correctly in my opinion, that many companies will find uses for LLM’s which are in the trillions of tokens per day.
Think less of “a bunch of people want to get recipe ideas.”
Think more of “a pharma lab wants to explore all possible interactions for a particular drug” or “an airline wants its front-line customer service fully managed by LLM.”
It’s unusual that individuals and industry get access to basically similar tools at the same time, but we should think of tools like ChatGPT and similar as “foot in the door” products which create appetite and room to explore exponentially larger token use in industry.
Let's estimate 200 million office workers globally as TAM running an average of 250k tokens. That's 50 trillion tokens DAILY. Not sure what model provider profit per token is, but let's say it's .001 cents.
Thats $500M per day in profit.
Anthropic expects to break even in 2028. They’re all unprofitable now.
Are they unprofitable because they don't profit on inference, or because they reinvest all of the profit into scaling up?
Remember how long Amazon was unprofitable, by choice.
They are scaling up using VC money, not revenue. As far as profit on inference goes, it's hard to separate it out from training: they cannot, at any given time, simply stop training because that would kill any advantage they have 6 months down the line.
For all practical purposes, you can't look at their inference costs independent of the training cost; they need to keep spending on both if they want to continue doing inference.
> Remember how long Amazon was unprofitable, by choice.
That was a very different scenario - AMZ was not spending their revenue on land-grabbing, they were spending their revenue on long-lived infra, while AI companies now are spending VC investment, not revenue, on land-grabbing.
The difference between spending your revenue on short-lived infra (training a new model, acquiring GPUs) and long-lived infra is that with long-lived infra, at any time, even after 10+ years, you can stop expanding your infra and keep the resulting revenue as profit.
With short-lived infra (models, GPUs), you can't simply stop infra spending and collect profit from the revenue, because the infra reached end-of-life and needs to be replaced anyway.
But I do think the important thing to look forward to is AI work which is totally detached from human intervention.
I find it irreplaceable.
+
>Not sure what model provider profit per token is, but let's say it's .001 cents.
So you'd be willing to pay thousands for a new feature, right?
Pharma does not trust OpenAI with their data, and they don't work on tokens for any of the protein or chemical modeling.
There will undoubtedly be tons of deep nets used by pharma, with many $1-10k buys replacing more expensive physical assays, but it won't be through OpenAI, and it won't be as big as a consumer business.
Of course there may be other new markets opened up but current pharma is not big enough to move the needle in a major way for a company with an OpenAI valuation.
But my bigger claim is that ~half the Fortune 500 will be able to profitably deploy AI with spends in the tens or hundreds of millions per year quite soon. Not that pharma itself is a major contributor to that effect.
Those all seem possible, but I wouldn't assign greater than a 50% probability to any of them, and the valuations seem to imply near-certainty.
This has been experimented on before by many companies over the recent years, most notably Klarna which was among the earliest guinea pigs for it and had to later on backtrack on this "novel" idea when the results came out.
I could see things like "nitrate" and "nitrite" possibly being a stumbling block for an LLM.
Since I’m not a scientific researcher, I have no idea if he’s just blowing smoke but I think it’s reasonable to think of a purpose-built system which has an LLM component being used by a team to do something useful.
Sam Altman will say anything if he thinks it will increase people's investment in his company.
Don't believe a word that slimeball speaks
How would an LLM be any good at this?
Even if all of the GPUs inside burn out and you want to put something else entirely inside of the building, that's all still ready to go.
Although there is the possibility they all become dilapidated buildings, like abandoned factories
Of the most valuable part is quickly depreciating and goes unused within the first few years, it won't have a chance for long term value like fiber. If data centers become, I don't know, battery grid storage, it will be very very expensive grid storage.
Which is to say that while there was an early salivation for fiber that was eventually useful, overallocation of capital to GPUs goes to pure waste.
Maybe it's cheaper if we measure by dollars or something, but at the same time we lack the political will to actually do it without something like AI on the horizon.
For example, many data center operators are pushing for nuclear power: https://www.ehn.org/why-microsoft-s-move-to-reopen-three-mil...
That's one example among many.
So I'm hesitant to believe that "electricity is a small cost" of the whole thing, when they are pushing for something as controversial as nuclear.
Also the 2 are not mutually exclusive. Chip fabs are energy intensive. https://www.tomshardware.com/tech-industry/semiconductors/ts...
AI companies are saying they are trying to build nuclear because it makes them sound serious. But they are not going to build nuclear, solar and storage is cheaper more flexible and faster to build. The only real nuclear commitment is Microsoft reopening an old nuclear reactor that had become uneconomic to operate. Building anything new would be a five+ year endeavor, if we were in a place with high construction productivity like China. In the US, new nuclear is 10 years away.
But as soon as Microsoft restarted an old reactor, all their competitors felt like they had to sound as serious, so they did showy things that won't result in solving their immediate needs. Everybody's renewable commitments dwarf their nuclear commitments.
AI companies can flaunt expensive electricity at high cost for high investor impact precisely because electricity is a small cost component of their inputs. It's a hugely necessary input, and the limiting factor for most of their plans, but the dollar amount for the electricity is small. The current valuations of AI assume that a kWh put towards AI will generate far far more value than the average kWh on the grid.
Many of legacy systems still running today are IBM or Solaris servers at 20, 30 year old. No reason to believe GPU won’t still be in use in some capacity (e.g. interference) a decade from now.
VS plurality of AI investment, i.e. trillions are going towards fast deprecating components where we can say with relative confidence will likely be net negative stranded assets in terms of amoritization costs if current semi manufacturing trends continues.
Keeping some mission critical legacy systems around is different than having trillions that makes no financial sense to keep on the books, i.e. post bubble new gen hardware will likely not have scarcity pricing or better compute efficiency (better apex and opex), there is no reason to believe companies will legacy GPUs around at scale if every rack loses them money relative to new hardware. And depending on actual commercialization compute demand, it can simply make more economic sense to retire them than keep them going.
TLDR old durable infra tends to retains positive residual value because they're not easy to replace economically/frequently, old compute has negative residual value because they are easy to replace economically/frequently.
- OverUtilized/UnderCharged: doesn't matter because...
- Lead Time vs. TCO vs. IRS Asset Deprecation: The moment you get it fully built, it's already obsolete. Thus from a CapEx point of view, if you can lease your compute (including GPU) and optimize the rest of the inputs for similar then your CapEx overall is much lower and tied to the real estate - not the technology. The rest is cost of doing business and deductible in and of itself.
- The "X" factor: Someone mentioned TPU/ASIC but then there is the DeepSeek factor - what if we figure out a better way of doing the work that can shortcut the workflow?
- AGI partnerships: Right now, you see a lot of Mega X giving billions to Mega Y because all of them are trying to get their version of Linux or Apache or whatever at parity with the rest. Once AGI is settled and confirmed, then most all of these partnerships will be severed because it then becomes which company is going to get their AI model into that high prestige Montessori school and into the right ivy league schools - like any other rich parent would for their "bot" offspring.
So what will it look like when it crashes? A bunch of bland empty "warehouses" with mobile PDU's once filling all their parking lot space gone. Whatever "paradise" that was there may come back... once you bulldoze all that concrete and steel. The money will do something else like a Don McLean song.
You're not quite thinking things through there man. Once the elites who built these follies have gone, the mob will go shopping for building materials. I wouldn't be surprised even if people end up living in these datacentres once they become derelict. They have AC after all.
This is still pretty far from a guarantee
On Amazon, buying a 5090 costs $3000 [2]
That's a payback time of 212 days. And Runpod is one of the cheaper cloud providers; for the GPUs I compared, EC2 was twice the price for an on-demand instance.
Rental prices for GPUs are pretty darn high.
[1] https://www.runpod.io/pricing [2] https://www.amazon.com/GIGABYTE-Graphics-WINDFORCE-GV-N5090G...
> You can already use Claude Code for non engineering tasks in professional services and get very impressive results without any industry specific modifications
After clicking on the link, and finding that Claude Code failed to accurately answer the single example tax question given, very impressive results! After all, why pay a professional to get something right when you can use Claude Code to get it wrong?
Giant telecoms bought big regional telecoms which came about from local telecoms merging and acquiring other local telecoms. A whole bunch of them were construction companies that rode the wave, put in resources to run dark fiber all over the place. Local energy companies and the like sometimes participated.
There were no standard ways of documenting runs, and it was beneficial to keep things relatively secret, since if you could provide fiber capabilities in a key region, but your competition was rolling out DSL and investing lots of money, you could pounce and make them waste resources, and so on. This led to enormous waste and fraud, and we're now on the outer edge of usability for most of the fiber that was laid - 29-30 years after it was run, most of it will never be used, or ever have been used.
The 90s and early 2000's were nuts.
At the local level, there is generally a cable provider with existing rights of way. To get a fiber provider, there’s 4 possible outcomes: universal service with subsidy (funded by direct subsidy), cherry-picked service (they install where convenient), universal service (capitalized by the telco) and “fuck you”, where they refuse to operate. (ie. Verizon in urban areas)
The private capitalized card was played out by cable operators in the 80s (they were innovators then, and AT&T was just broken up and in chaos). They have franchise agreements whose exclusivity was used as loan collateral.
Forget about San Diego, there are neighborhoods in Manhattan with the highest population density in the country where Verizon claims it’s unprofitable to operate.
I served on a city commission where the mayor and county were very interested in getting our city wired, especially as legacy telco services are on the way out and cable costs are escalating and will accelerate as the merger agreement that formed Spectrum expires. The idea was to capitalize last mile with public funds and create an authority that operated both the urban network and the rural broadband in the county funded by the Federal legislation. With the capital raised with grants and low cost bonding (public authority bonds are cheap and backed by revenue and other assets), it would raise a moderate amount of income in <10 years.
We had the ability to get the financing in place, but we would have needed legislation passed to get access to rights of way. Utilities have lots of ancient rights and laws that make disruption difficult. The politicians behind it turned over before that could be changed.
I stumbled on old maps that showed a complete coverage of fiber in my municipality, paperwork from a company that was acquired, and which in turn merged, then was bought out by one of the big 5 ISPs. When local officials requested information regarding existing fiber, this ISP refused and said any such information was proprietary. They later bid on and won contracts to run new fiber (parallel to existing lines which they owned, which still had more than a decade of service life left in them at that point).
I estimated that only around 10-15% of the funding went toward actual labor and materials, the remainder was pure profit. The local government considered it a major victory, money well spent.
Consensus theory: If AGI then superintelligence.
AI CapEx plans are not ROI based. Rather, they are the cost of "how do I remain competitive in the race to attain AGI" coupled with conveniently deep pockets. The money is being spent because the spenders can afford it and they see it as an existential risk as much as a profit opportunity.
Maybe OP's conclusion about the headline question blunts some political opposition to data centers, but that's not the salient issue.
The issue is this: America is betting a meaningful chunk of GDP that AGI is possible. This is The Manhattan Project 2.0.
In reality, if you have a dryer outlet, you have a good fraction of 10 kW available.
If someone can afford an 8 GPU server, they should be able to afford some #6 wire, a 50A 2P breaker, and a 50A receptacle. It has the same exact power requirements as an L2 EV charger.
You would need a neutral if it was a 208/120V three-phase service.
Neutrals and grounds are sized per the NEC, neutrals are the same size as the line conductors and equipment grounds are sized off of a table.
#6 conductors and #10 ground is what the NEC calls for.
I live the NYC 208V life doing mostly resi, though.
Quick search of the spec for that is 6 power supplies, 2 of which are redundant. Looks to use a neutral to me. Says it uses C19/C20 connectors
edit: wait most ranges use 14-50R outlets and need a neutral ran. I am calling your statement into question. Surely harmonics and 120V internal draws cause non-zero neutral current. And I'm sure GPUs have harmonics being semiconductor flavored.
My bad, NEMA 14-50R is a 4-wire receptacle with a neutral.
I learned something new today, 200% neutrals are not required by the NEC but can help with non-linear loads, and certain transformers that mitigate harmonics need 200% neutrals.
* The GPUs in use in data centers typically aren’t built for consumer workloads, power systems, or enclosures.
* Data Centers often shred their hardware for security purposes, to ensure any residual data is definitively destroyed
* Tax incentives and corporate structures make it cheaper/more profitable to write-off the kit entirely via disposal than attempt to sell it after the fact or run it at a discount to recoup some costs
* The Hyperscalers will have use for the kit inside even if AI goes bust, especially the CPUs, memory, and storage for added capacity
That’s my read, anyway. They learned a lot from the telecoms crash and adjusted business models accordingly to protect themselves in the event of a bubble crash.
We will not benefit from this failure, but they will benefit regardless of its success.
if there's ever a glut in GPUs that formula might change but it sure hasn't happened yet. Also, people deeply underestimate how long it would take a competing technology to displace them. It took GPUs nearly a decade and the fortunate occurrence of the AI boom to displace CPUs in the first place despite bountiful evidence in HPC that they were already a big deal.
What about the possibility of improvements in training and inference algorithms? Or do we know we won't get any better than grad descent/hessians/etc ?
This is a kind of risk that finance people are completely blind to. Open AI won't tell them because it keeps capital cheap. Startups that must take a chance on hardware capability remaining centralized won't even bother analyzing the possibility. With so many actors incentivized to not know or not bother asking the question, there's the biggest systematic risk.
The real whiplash will come from extrapolation. If an algorithm advance shows up promising to halve hardware requirements, finance heads will reason that we haven't hit the floor yet. A lot of capital will eventually re-deploy, but in the meantime, a great deal of it will slow down, stop, or reverse gears and get un-deployed.
This is not the case for AI data centers at all. The compute is a major cost in the whole build budget. Having that installed and not used is financial ruin.
And having it used consumes so much energy that it's everyone else's ruin.
There are compounding incentives for this. I see this as the most likely outcome, though it will likely be stepwise rather than gradual.
What do you think LLM tuned GPUs or TPUs are going to be used for that is completely different and not AI related?
My main point in arguing that now isn’t like 2000 is that unlike in 2000 we have actual hardware and physical assets underpinning this bubble. In 2000 the assets were literally just imaginary. Yes there is speculation now but it is underpinned by silicon that will still be worth decent money even after LLMs are exposed as a hallucinatory mirage.
The key dynamic: X were Y while A was merely B. While C needed to be built, there was enormous overbuilding that D ...
Why Forecasting Is Nearly Impossible
Here's where I think the comparison to telecoms becomes both interesting and concerning.
[lists exactly three difficulties with forecasting, the first two of which consist of exactly three bullet points]
...
What About a Short-Term Correction?
Could there still be a short-term crash? Absolutely.
Scenarios that could trigger a correction:
1. Agent adoption hits a wall ...
[continues to list exactly three "scenarios"]
The Key Difference From S:
Even if there's a correction, the underlying dynamics are different. E did F, then watched G. The result: H.
If we do I and only get J, that's not K - that's just L.
A correction might mean M, N, and O as P. But that's fundamentally different from Q while R. ...
The key insight people miss ...
If it's not AI slop, it's a human who doesn't know what they're talking about: "enormous strides were made on the optical transceivers, allowing the same fibre to carry 100,000x more traffic over the following decade. Just one example is WDM multiplexing..." when in fact wavelength division multiplexing multiplexing is the entirety of those enormous strides.
Although it constantly uses the "rule of three" and the "negative parallelisms" I've quoted above, it completely avoids most of the overused AI words (other than "key", which occurs six times in only 2257 words, all six times as adjectival puffery), and it substitutes single hyphens for em dashes even when em dashes were obviously meant (in 20 separate places—more often than even I use em dashes), so I think it's been run through a simple filter to conceal its origin.
Other than that I'd rather choose a comprehensive article than a summary.
On topic: It is always quite easy to be the cynical skeptic, but a better question in my view: Is the current AI boom closer to telecoms in 2000 or to video hosting in 2005? Because parallels are strong to both, and the outcomes vastly different (Cisco barely recovered by now compared to 1999 while youtube is printing money).