https://www.brookings.edu/wp-content/uploads/2026/09/4c_Van-Nieuwerburgh.pdfEconomic footprint of AI's buildout, from specialized chips and electricity to purpose-built data centers. (Two-thirds of a data center's costs are IT equipment, with one-third going to real estate and its associated power infrastructure.) "At an average of 3.3 percent of GDP per year, the projected buildout would be larger relative to the economy than the major U.S. canal, railroad, electrification, highway, and telecommunications investment booms."The calculations come from Stijn van NieuwerburghJust as the rail and telecoms expansions led to notable bubbles and busts, Van Nieuwerburgh wrote that the extent of the buildout, the still-untested revenue streams, and the intricate financing structure emerging around AI mean it could be primed for a fall. "This is freaking complicated," he said in a briefing with reporters of the arrangements emerging between AI firms, major tech hyperscalers, banks, private credit lenders, real estate firms, and a host of other players involved in building what he conservatively estimated at 183 gigawatts worth of new data-center capacity over the next seven years, compared with about 57 gigawatts currently installed.... >The investment underway already has outstripped what the major players can fund from their own cash flows. "These developments do not imply that financial distress is imminent. Strong growth in AI applications, high utilization, and continued improvements in model capability could support the projected infrastructure and generate stable cash flows," he wrote.As an example, he wrote that the AI industry will need to be earning $3.3 trillion in annual revenue by 2033 to achieve the expected return on the investment, and "given current estimates of annual combined revenues of OpenAI and Anthropic of around $100 billion, revenues would need to grow at roughly 80% per year."
>>543400168>3.3%fake and gay
>>543400168>buildoutI use AI today to solve nearly all my problems
>>543400288it sounds pretty legit thoughGDP is roughly 30 trillion I thinkroughly 3% of that is 1 trillion monies
>>543400466uh huh and why did they specify .3%? because theyre fagmasons who put 33 everywhere
>>543400553yeah that may be but the numbers themselves reflect the GPD and the costs that actually is now being pumped into AI datacenters, its largely public info on how much they have announced to put money in what project
>>543401095its fake and gayeven if the numbers were true (they probably are not) they would have increased the amount past what was needed or justified just to match that number
The potential overspend on datacenters feels like an unnecessary problem that they are creatingLike just build the datacenters slower...? Wait until you have enough money to pay for them
>>543401269and yet, they announce new datacenter projects every day, on many countries, without having properly even one large datacenter running in 2026, everything is predicted to happen later in 2027Finland:Seinäjoki city:massive 550MW AI campus (SJK01) in July 2026, with an initial investment over €1.5 billion(not yet operational)Google (Hamina, Kajaani, Muhos, Vaala): Announced a landmark €13 billion investment in September 2026 to expand its Hamina campus and build new data infrastructure across Kajaani, Muhos, and Vaala through 2028Nebius (Lappeenranta): Announced a €8.5 billion AI data center project in March 2026 with construction slated to come online in phases starting in 2027Microsoft (Kirkkonummi / Vaasa): Pre-construction and earthworks advanced significantly through mid-2026(not yet operational though)
>>543400347Does it solve the fact you're gunna go bankrupt and are running into a "force majeure" deficit that your shit autocomplete AI didn't see coming?Hoo boy can't wait for when it starts hacking nuclear secrets, I bet it'll be worth trillions then!
>>543400168Why would we build out this broke shit that we have now?
>>543401507My new favorite thing is having AI confirm that I'm not stupid and that it is in fact the giant trillion dollar companies who are behaving irrationally.I think most people who look at this from the outside agrees that it doesn't make any sense. AI is already present, available, and growing. We can already use it, and the labs can continue to develop it. If they want to spend a couple hundred billion on data centers every year, then there is no problem. But these very smart people all want to spend a trilly and they're going to make a mess of everything for no good reason
>>543400168Ai has already taken over the meme industry and the online shilling industry. Its pure head in the sand ludditte behavior to think ai is going away.
>>543401781no one is saying AI is going awaybut we just may have to settle with models we already haveas it becomes financially impossible to make them better at this pointit would require a new approach: actually optimizing the current models instead of just creating more of AI processing devices (Nvidia GPUs) to create even more powerful AI by pure number crunching
>>543401862That's not even the constraint !They can continue to spin up improved models with the current infrastructure, with longer training intervals.We haven't hit any kind of wall in AI development. These nerds just want to scale to RSI as soon as possible and the adults in the room that should be telling them NO are too busy going to AIPAC dinners. Sure we can go a little faster if we slurp every last dollar from the global financial system (and more) but that cost isn't worth the benefit. Almost everyone agrees on this except the guys signing the checks :D
It's interesting to think about the amount of initial training data needed to "prime" AI so that it recursively improves and you get super intelligence. There's the technical infrastructure needed, massive amounts of people generating the data over hundreds or thousands of years, the human systems that make it all possible, and a bunch of other stuff.and now in this moment it seems like we're on the cusp of being able to do it or not. Odds seem pretty slim for all this...
>>543402214if/when we hit RSI/superintelligence we will still need to build more and more datacenters to feed the beast so that it can keep getting smarter.This is even more reason that the new construction should be paced in a way that correlates with the amount of money that the companies are taking in... we should be planning to build compute over the next several decades at a reasonable rate. Instead it seems that they are all in a sprint to build more than the debt market can take by 2030, with electrical generation as the only "bottleneck" (note that they aren't taking the electrical buildout as seriously as they should be either)
>>543401591>it's just shitty autocomplete>it's going to hack nuclear secretsYou ever wonder if the anti-AI people are really just against any kind of intelligence?
A lot of these ai companies are faking it so there’s a bubble in that.>>543403034They’re segmented this way. Politicians wanted a farmable, bot-like person for reliable votes. Social Media algorithms feed them AI bad content so they parrot it.
>>543400168/new/ memes, good times
