Chief of Staff @SIEPR. PhD'ing @UW. I research energy markets during the transition and consumer sentiment. Views my own.
Palo Alto, CA
- Tweets 5,709
- Following 1,950
- Followers 3,075
- Likes 37,207
Ryan Cummings retweeted
FWIW here is Claude's attempts to compare a) Van Nieuwerburgh b) Goldman's c) Jared Bernstein & Ryan Cummings d) mine. Not huge variation honestly
New BPEA paper by @SVNieuwerburgh argues that AI buildout will cost $10.3T over next 8 years. Furthermore, to get a 10% return on this, those investing will need annual revenues of $3.7T by 2032! (Approx 9% of GDP!)
brookings.edu/wp-content/upl…
2
5
25
26,267
Ryan Cummings retweeted
In our recent Stripe Economics piece, we conclude that aggregate productivity growth is so far decent-not-miraculous & there's reason to be cautious that even that recent firming is about AI microproductivity gains. 1/2
stripeeconomics.com/p/ai-and…
7
31
177
39,275
👀
Me+@econJaredB are out w/ a new piece examining how much revenue is needed to justify hyperscaler AI capex. We estimate that incremental AI revenue is currently b/n $86-$188.1B for the hyperscalers, but they need $13.1-$18.7 *trillion* over the next decade for that to pencil out.
2
1
11
7,970
Ryan Cummings retweeted
In July we argued that the AI ecosystem required $2-3trn annual revenues in perpetuity to deliver a payback on AI capex. Yesterday Goldman published a similar report with extremely similar results
12
37
214
20,141
Ryan Cummings retweeted
ill read more🫡 but i also suspect there'll be a lot of future fights over these corps "over"attributing a certain amount of earnings growth to AI in order to juice ROI estimates and that itll be messy and unfalsifiable in many cases
1
1
2
355
Two additional things to stress:
1) Even if you make the most aggressive possible assumptions about incremental AI revenue, use the lowest discount rate possible, and assume all capex after 26 is canceled, to just *break even* the firms need to increase revenue 3-4x *next year*
Me+@econJaredB are out w/ a new piece examining how much revenue is needed to justify hyperscaler AI capex. We estimate that incremental AI revenue is currently b/n $86-$188.1B for the hyperscalers, but they need $13.1-$18.7 *trillion* over the next decade for that to pencil out.
3
10
41
3,668
2) @PeterBerezinBCA and @econcallum have done similiar analyses. They use slightly different assumptions, but all of our estimates are roughly consistent with one another. There is no way around the fact that the incremental AI revenues need to explode incredibly fast.
1
5
286
Me+@econJaredB are out w/ a new piece examining how much revenue is needed to justify hyperscaler AI capex. We estimate that incremental AI revenue is currently b/n $86-$188.1B for the hyperscalers, but they need $13.1-$18.7 *trillion* over the next decade for that to pencil out.
14
27
126
29,932
One thing I want to emphasize: it is very, very difficult for both observers *and the firms themselves* to understand the ROI of these projects, which is a bit wild considering how much is being spent. Link here:
econjared.substack.com/p/ai-…
2
3
21
5,064
Ryan Cummings retweeted
Over the last few weeks, the AI bull discourse has shifted over to AI infrastructure and away from AI models.
Why? AI models are seeing threats to their growth, due to greater competition driving price cuts and model efficiencies. Meanwhile, GPUs and infrastructure are becoming more scarce and rental rates are rising.
I'd argue AI infrastructure is not the safe bet investors think it is. As a spend category, it's even *more* concentrated than AI spend, or any other spend category we track.
The entire bet is based around a narrow group of highly correlated buyers. And simple growth in rental rates or demand does not account for outsized concentration risk.
See for example Ramp data today in Torsten Slok's daily spark
7
8
58
22,158
A 90 day diesel export ban would truly be a storied piece of policymaking because it would quite literally be the worst of both worlds; refiners would cut runs, throw excess into storage (instead of sending to PADD 1A+PADD5), *raising* prices domestically AND internationally.
Per @politico, the admin is poised to implement a ban on diesel exports (though it remains subject to debate). politico.com/news/2026/09/23…
1
6
14
1,213
Ryan Cummings retweeted
This paper finds that 90% of the observed rise in the 10-year US Treasury since August 2020 has occurred in the three-day window around US payroll reports or speeches by top Fed officials (the chair, vice chair, or Waller), which accounts for just 24% of trading days
The paper finds those days are responsible for 81% of the rise in markets' expected average short-term rate over the next 10 years, which means this isn't about investors getting nervous about holding long-term debt and is instead about markets revising their view of short-term rates.
Anatomy of a rise: Monetary policy and the post-Covid surge in long-term interest rates
Paul Beaudry @ubcVSE @UBC, Paolo Cavallino @BIS_org, Tim Willems @timwillems85 @bankofengland
ow.ly/hVw350ZPyBH
50
135
756
308,688
Ryan Cummings retweeted
“Withdrawals from investment accounts as a share of total spending almost doubled from 3.5 percent in April 2019 to 6.8 percent by April 2026.”
The wealth effect is real.
Source: jpmorganchase.com/institute/…
8
31
132
46,723
Refinery capacity is actually slightly higher now than it was in 1980 despite the # of refineries halving, largely b/c small, inefficient refineries closed while larger, existing refineries added capacity.
Enviromental concerns had a role too, but not as big as you'd think.
the real reason is we stopped building refineries in 1977 for stupid left wing reasons and cancelled two pipelines.
5
391
1/2: To add to Garrett's excellent analysis; a diesel export ban would hammer New England (PADD 1A) in particular.
Inventories there are already at their lowest in 10 years heading into fall, and a HUGE share of homes in PADD 1A rely on diesel (No. 2 fuel oil) for home heating.
If you ban diesel exports, this is what happens:
1) Without the overseas market, diesel inventories start to accumulate domestically. This temporarily puts downward pressure on wholesale prices in the US, likely more so in the Gulf Coast than anywhere else. HOWEVER
1
4
684
2/2: B/c of a lack of pipelines, PADD 1A competes w/ (a structurally short) Europe for diesel barrels.
In the Biden Admin, after RU invaded UKR, we were worried about actual physical shortages of diesel in PADD 1A leading to folks not having heat. The situation is even worse now
1
1
136