Ryan C. Smith
@rcsmitheco.bsky.social
350 followers 810 following 680 posts
Author of, "The Real Oil Shock: How Oil Transformed Money, Debt, and Finance", host of "A History of Capitalism" podcast. Economist & historian, currently #amquerying "The Emerging Energy Revolution" on how renewables+batteries are transforming the world.
Posts Media Videos Starter Packs
rcsmitheco.bsky.social
The really neat thing about the growth in renewable energy over the past year is how end-user adoption has been a key driver. The shift to renewables is inevitable in part because renewables don't have the same steep capital requirements as fossil fuels & are much easier to adopt.
#ClimateSky
Reposted by Ryan C. Smith
islandpress.bsky.social
WHAT SHOULD A CLEVER MOOSE EAT? is a delightful exploration of the North Woods and what its natural history teaches us. WHITE PINE is a journey through the natural and human history of the white pine.

This month only get both for $40 (+ free shipping!) islandpress.org/2for40
rcsmitheco.bsky.social
Imagine telling everyone you know absolutely nothing about economics like this.
#EconSky
carlquintanilla.bsky.social
ANDREESSEN: Even if AI ends up destroying all the jobs, “the result would be hyper-deflation of prices, which is the thing that people miss. .. Things that today cost a lot of money will all of a sudden be cheap or free.”

@fortune.com
fortune.com/2025/10/08/b...
rcsmitheco.bsky.social
Then that has to be part of any such calculations and discussions.

Starting with asking whether the conditions that enable bubbles are also detrimental to democracy.
rcsmitheco.bsky.social
You just KNOW this will be in a future history book.
dannykpolitics.bsky.social
WATCH: ‘Portland Frog’ pepper balled by ICE agents
Reposted by Ryan C. Smith
billmckibben.bsky.social
America is in crisis.

So it's a small relief that the rest of the world is actually getting remarkable stuff done with energy. We just passed a huge milestone globally!

Now back to work on America.
billmckibben.substack.com/p/something-...
Something extraordinary just happened
But of course there's a but. Or two.
billmckibben.substack.com
rcsmitheco.bsky.social
Especially because all that would do is allow the usual suspects to walk away from the wreckage. There's just no amount of money that's going to make this tech sustainable, if there was we'd have a glimmer of it by now.
rcsmitheco.bsky.social
That's the real kicker there. They've created a deeply unstable situation where everyone is propping everyone else up and there just *isn't* the money there to bail it all out.

I suppose the US government *could* do 2008 again but the social consequences of that are probably not good.
rcsmitheco.bsky.social
It is *very* common for bubbles to experience escalating surges as they get closer to bursting. Watch the markets for further unrealistic exuberance right up until someone runs for the door.
#EconSky #AI
rcsmitheco.bsky.social
Also:

The money to make this work just isn't there, check out @edzitron.com for more. 11/10
edzitron.com
In fact, reporting and projections suggest that OpenAI has committed to over $450 billion in compute alone in the next five years. Their other costs are likely to be in the tens of billions, with $30bn in salaries over the next few years.
www.wheresyoured.at/openai-onetr...
Compute Costs - At least $456.9 Billion Over Five Years
Let’s run through this one more time: 

OpenAI has committed to spending $300 billion on compute through 2030 with Oracle.
OpenAI has said it plans to spend an additional $100 billion on backup compute over the next four years.
It’s unclear whether this figure includes the commitments stated previously, or are something distinct. Based on the wording of the coverage I’ve read, I’m inclined to think it’s a separate commitment — which is not good! 
OpenAI has agreed to buy $11.9 billion, $4 billion and $6.5bn in compute from CoreWeave over the next five years.
Per The Information, OpenAI expects to spend at least $13 billion in 2027 and $28 billion in 2028 on compute with Microsoft.
This also matches up with The Information’s reporting — though I actually think it’s going to be higher, as OpenAI fudges its numbers and “rounds” them (see the diagram).

OpenAI’s Other Costs Are Deeply Questionable - I Estimate It Will Spend At Least $30 Billion On Salaries Alone In The Next 4 Years, And Other Costs (Storage, Data, Etc) Are Likely To Be In The Billions
So, OpenAI’s supposed $115 billion burn, leaked to The Information on September 6 2025, does not really make much sense.

Assuming we trust these numbers, we’re meant to believe that OpenAI will only lose $17 billion in 2026, on (supposed) revenues of $30 billion, meaning it’s burning $47 billion in 2026.

Assuming, again, that we trust the leaked numbers, OpenAI will spend $40 billion on compute in 2026, leaving it with $7 billion of leftover burn rate.


Now, The Information’s “cash burn” projections are extremely kind, based on their own reporting around OpenAI’s 2024 costs:


The Information reports that some of these costs are the upfront cost of training that has been amortized over a few years, but if that’s the case, why are there other training costs too? Does OpenAI own its own hardware? What does that billion dollars represent? Who knows! What’s a billion between friends?

In any case, I also think The Information is lowballing the cost of employees that make a median of $1.37 million a year, though stock-based compensation may play into that too. But even if we assume they make $500,000 a year in cash on average, I refuse to believe that the 3,500 or so people that worked there in 2024 only cost the company $700 million. If we multiply those two figures, we’re left with nearly $1.75 billion. Some estimates say over 6,000 people work at OpenAI, and with that same math, you’d be looking at $3 billion, and that’s before you add overheads like benefits and taxes.

Assuming that OpenAI grows its headcount by 25% a year, that means it’ll be at 7,500 people in 2026, 9,375 by 2027, 11,718 by 2028, 14,648 by 2029, and 18,310 by 2030, meaning that assuming its average compensation remains the same (which is unlikely, given the ongoing demand for AI talent), it’ll pay $3.75 billion in 2026, $4.688 billion in 2027, $5.859 billion in 2028, $7.324 billion in 2029, and $9.155 billion in 2030.

And this is for a very conservative estimate of the cost of its staff, and one that doesn’t include stock based compensation, or any of those other ancillary costs that I mentioned earlier (like benefits and payroll taxes).. 

Other costs like data, hosting, sales, and marketing are hard to gauge, but they are, without a doubt, going to increase. Nevertheless, they amounted to $1.8 billion in costs in 2024, and if we assume that they increase by, say, 20% each year (another conservative estimate,) this would mean that they’d be $2.16 billion in 2025, $2.592 billion in 2026, $3.11 billion in 2027, $3.7 billion in 2028, $4.47 billion in 2029, and $5.3 billion in 2030. These are, I want to be clear, very likely very low.
rcsmitheco.bsky.social
At this point, responsible economists should really be asking two questions:

1) How do we stop this pattern of increasingly large, unstable economic bubbles?

2) How can we salvage what we can from this bubble & repair the damage while we can? 10/10
#EconSky #ClimateSky #AcademicSky
rcsmitheco.bsky.social
This same spending boom is now probably the largest bubble in history. Said bubble is directly contributing to rising prices, choking out investment in other sectors of the economy, & appears to be the only thing holding up the US economy.

All of which was also true before the gov't shutdown. 9/x
carlquintanilla.bsky.social
“.. His conclusion is very stark: not just that an economy already at stall speed will fall into recession as both the data-center and wealth effects plateau, but that they’ll reverse, just as in the dot-com bubble did ..”

@marketwatch.com
www.marketwatch.com/story/the-ai...
rcsmitheco.bsky.social
On top of being a major dead-end, LLM datacenters are also huge drains on power grids & are directly linked to rising electricity prices across the US. It would not be surprising if these rising prices are part of why consumer spending has consistently declined since 2025 began. 8/x
The AI Data Center Boom Is Driving Up Electricity Costs, Research Shows
The price of electricity, juiced by demand from power-hungry data centers, is being passed on to residential customers.
www.cnet.com
rcsmitheco.bsky.social
I could spend all day on the repeated broken promises, persistent inability to fix hallucinations, recommendations to eat rocks & follow the Beatles' nonexistent advice on leaving dogs in hot cars, or any number of other examples. The point is this tech has consistently failed to deliver. 7/x
Mark Zuckerberg Humiliated as AI Glasses Debut Fails in Front of Huge Crowd
Meta CEO Mark Zuckerberg's demos of new smart glasses repeatedly failed, leading to awkward stares, deafening silences, and muted laughter.
futurism.com
rcsmitheco.bsky.social
AI, by contrast, has yet to produce any profits or demonstrate there is a pathway to profitability. Reporters like @edzitron.com have been sounding the alarm on this trend & the failure of AI companies to deliver on profit-making promises for over a year. 6/x
The AI Industry Is Still Light-Years From Making a Profit, Experts Warn
Some AI investors seem to finally be waking up to a huge problem: that the industry has not yet figured out how to make a profit.
futurism.com
rcsmitheco.bsky.social
Usually, this term is used in resource economics because the main examples (the Netherlands & natural gas, the Persian Gulf & oil) were driven by fossil fuel wealth. Those sources of wealth also, unlike the crypto & then AI bubbles, actually added value, albeit mismanaged, for their societies. 5/x
rcsmitheco.bsky.social
Dutch disease, to give a short summary, refers to economies where a single sector becomes a major driver of profits & economic activity. This encourages investment to focus on the safe, sure bet at the expense of other sectors. Over time, the economy becomes dependent on that sector's success. 4/x
Dutch Disease: Wealth Managed Unwisely
When countries experience a sudden large increase in income, the consequences can be harmful
www.imf.org
rcsmitheco.bsky.social
The scale of the US economy's reliance on tech spending, generally, is also becoming better understood with increasing alarm. Economically speaking, becoming this dependent on a single sector is often referred to as the Dutch disease. This term came out of resource economics in 1977 but fits. 3/x
Without data centers, GDP growth was 0.1% in the first half of 2025, Harvard economist says | Fortune
Is a U.S. without data centers a country without GDP growth?
fortune.com
rcsmitheco.bsky.social
The business press has seen something of a creeping about-face on AI ever since reporting emerged this past spring saying that AI spending made up 40% of US GDP growth at that point in time. This shift was a product of soaring investment & plummeting consumer spending & their knock-on effects. 2/x
justinhendrix.bsky.social
"The hundreds of billions of dollars companies are investing in AI now account for an astonishing 40 per cent share of US GDP growth this year... In a way, then, America has become one big bet on AI."
America is now one big bet on AI
It’s seen as the magic fix for every threat to the US economy
www.ft.com
rcsmitheco.bsky.social
The Bank of England certainly knows a thing or two about bubbles, seeing as they've been near or at the epicenter of many of the past three centuries most well-known examples.

This one, however, might just be the worst with all the trade war & killing agriculture stuff in play. 🧵 1/x
#EconSky #AI
Bank of England Warns of an AI Bubble Burst
Valuations appear stretched especially for AI companies, the bank said.
gizmodo.com
rcsmitheco.bsky.social
This is also consistent everywhere renewable energy is employed. There is simply no competition between traditional energy sources & the potential offered by wind & solar.
#Greensky #ClimateSky #EnergySky #EconSky
zcolman.bsky.social
Trump & his Cabinet say wind & solar are making electricity more expensive.

Federal data say the opposite.

Our analysis showed states w/ > avg wind/solar generation & > avg acceleration of wind/solar since ‘21 were likelier to have < avg power prices.

www.politico.com/news/2025/10...
The government’s own data rebuts Trump’s claims about wind and solar prices
The Trump administration has targeted renewable energy for driving up electricity prices, but POLITICO’s analysis shows states that are growing their wind and solar power typically have lower power co...
www.politico.com
Reposted by Ryan C. Smith
msmarchmont.bsky.social
George Square colours,
Edinburgh