Where Do You Hide From an AI Bubble?

CoreWeave

Across the shows: 57 speakers in other episodes, 15 bullish and 55 bearish calls, past 30 days

• Chris Clothier · CG Asset ManagementBearish

  CoreWeave's debt cost near 15% makes continued data center investment unlikely to make sense.

  CoreWeave is a neocloud that only provides compute to AI labs, many of its businesses were repurposed crypto miners. Over the last couple of weeks its bonds have fallen in price, pushing its cost of debt up to around 15%. Chris thinks it is highly unlikely that it makes sense to keep investing at that cost of capital, so he sees neoclouds as the weakest links in the AI build-out.

  “the cost of its is bonds falling in price, and therefore, the cost of its debt rising up to sort of in the region of 15%. And I think it's highly unlikely that at that cost of capital that it can make any sense for them to continue investing.”at 12:24

Oracle

Across the shows: 24 speakers in other episodes, 5 bullish and 21 bearish calls, past 30 days

• Chris Clothier · CG Asset ManagementBearish

  Oracle is showing stress as the pivot point, given its high indebtedness and weaker core franchise.

  Chris sees Oracle as the pivot point of the AI debt cycle. Unlike the hyperscalers, it lacks their core franchise and has always run with much higher indebtedness. He says it is already starting to show signs of stress, which he treats as an early sign that more established AI-linked businesses are coming under pressure.

  “I think Oracle is the kind of... Probably the kind of the pivot point in all of this. Doesn't have the same kind of core franchise that the hyperscalers have. It has always run its business with a much higher level of indebtedness.”at 13:04

3i Infrastructure

• Chris Clothier · CG Asset ManagementBullish

  Listed infrastructure trust 3i Infrastructure should deliver about 9% nominal, or 5-6% real, returns.

  Chris says the fund holds listed infrastructure trusts including 3i Infrastructure, which it sees as a good home because equity markets look stretched. On reasonable assumptions these trusts offer nominal returns of about 9% a year with good inflation protection, so roughly 5-6% real.

  “we think that these things offer nominal returns of the order of 9% per annum, and they've got pretty good inflation protection. So they're gonna deliver something like five to 6% real”at 30:47

International Public Partnerships

• Chris Clothier · CG Asset ManagementBullish

  Listed infrastructure trust INPP should deliver about 9% nominal, or 5-6% real, returns.

  Chris says the fund holds listed infrastructure trusts, including INPP, which it sees as a safe place to hide from stretched equity markets and the AI boom. On reasonable assumptions these trusts offer nominal returns of about 9% a year with good inflation protection, meaning roughly 5-6% real.

  “we think that these things offer nominal returns of the order of 9% per annum, and they've got pretty good inflation protection. So they're gonna deliver something like five to 6% real”at 30:47

HICL Infrastructure

• Chris Clothier · CG Asset ManagementBullish

  Listed infrastructure trusts like HICL should deliver about 9% nominal, or 5-6% real, returns.

  Chris says the fund is a holder of listed infrastructure trusts including HICL, INPP and 3i Infrastructure, which it sees as a good home because equity markets look stretched and the AI boom is a concern. Using reasonable assumptions, he thinks these trusts offer nominal returns of about 9% a year and have good inflation protection, so they should deliver about 5-6% real returns.

  “we think that these things offer nominal returns of the order of 9% per annum, and they've got pretty good inflation protection. So they're gonna deliver something like five to 6% real”at 30:47

AI infrastructure & data centers

• Chris Clothier · CG Asset Management

  Hyperscaler AI capex needs about $5.5T of lab revenue by 2030, roughly 14% of GDP, which looks implausible.

  Chris estimates about $1T a year of hyperscaler capex over five years, so about $5T of kit in the ground, depreciated over roughly seven years (GPUs about four, other kit about ten). With hyperscaler EBITDA margins of about 55% and a 13% pretax return on invested capital, he says the build needs about $2.75T of annual revenue by 2030. Adding frontier-lab gross margins of about 50%, he gets roughly $5.5T of lab revenue, about 14% of projected 2030 US nominal GDP of $39T. He calls that a bit less than Americans spend on food today, and says the open-ended investment runway is economically insane.

  “So what is the revenues that as it might be the frontier labs, the anthropics, the open AIs need to be generating? ... at a 50% gross margin, well, that points to $5,500,000,000,000 of revenue.”at 7:42

• Chris Clothier · CG Asset Management

  Data center build-out could slow from political pushback or tighter credit, not from weak demand.

  Chris sees two possible brakes on the AI boom: political resistance to building data centers, and the availability of credit. He cites a credit trader who reported Google and Amazon issuing several billion in sterling and euro bonds this year, and banks sounding out large London fund managers about a single €100bn deal for Meta. If credit dries up the build-out slows, but he thinks the fallout would be relatively limited because the compute already deployed is valuable.

  “partly it's the political resistance to the build out of these data centers, which may be the thing that causes this whole thing to slow down. And then on the other hand, it's the availability of credit to do this.”at 17:12

• Chris Clothier · CG Asset Management

  Hyperscaler AI spending is rational given huge demand backlogs, though unlimited investment is not.

  Chris says the hyperscalers' quarterly earnings calls show CFOs describing huge backlogs of demand for AI compute. He concludes that the investment going on today is entirely rational because demand is very high and deployed AI compute is likely to stay valuable. He still finds an open-ended runway of investment economically insane.

  “Whenever you listen to the quarterly earnings reports of Microsoft, of Google, their CFOs are saying, look, we've got these huge backlogs of demand. And so all of the investment that's going on today is entirely rational.”at 18:39

US Treasuries & yields

• Chris Clothier · CG Asset Management

  Inflation-linked bonds have beaten nominal bonds by about 1.2% a year since 2000, so they are sensible.

  About 48% of the fund is in inflation-linked bonds, split roughly 50/50 but tilted towards US TIPS and UK index-linked gilts. Since TIPS came into existence in the late 1990s they have outperformed nominal Treasuries by about 1.2% a year, a cumulative outperformance of over 90%. Chris argues that this happened in an era when inflation was well behaved, so TIPS don't need a deeply inflationary environment to be a sensible place for money.

  “they've outperformed their nominal counterparts by 1.2 per year. So that's cumulative outperformance of over 90%. ... we don't necessarily even need for it to be this deeply inflationary environment for them to be a sensible place to have some of your money.”at 28:33

• Chris Clothier · CG Asset Management

  Long-dated bonds are in a secular bear market, and he won't buy them until a bond crisis.

  Because of fiscal pressures the fund keeps portfolio duration very short and has been reluctant to own long-dated bonds. Chris thinks the only thing that will change the fiscal mood is a bond market crisis, so the fund is waiting for one before buying. He calls it a secular bear market for long-dated bonds until that happens, while accepting that yields may fall in a slowdown.

  “we're not interested in owning long dated bonds until the fiscal zeitgeist changes. And we think that the only thing that is going to bring about a change in the fiscal zeitgeist is a bond market crisis.”at 32:40

Gold & precious metals

• Chris Clothier · CG Asset Management

  Gold was behaving like a speculative asset, so the fund sold its small holding around $5,100.

  Chris says the fund sold its roughly 1% gold position around February at about $5,100 because gold was starting to behave like a speculative asset. That felt weird and fundamentally wrong to the team. He admits they should have owned much more gold on the way into the mini bubble, but they now own none.

  “we saw that gold was starting to behave like a speculative asset. That just felt weird and and fundamentally wrong to us.”at 29:57

• Chris Clothier · CG Asset Management

  Clothier dislikes gold as an asset, but sees no reason central banks will abandon it.

  Clothier says he hates gold as an asset, citing no income, difficulty valuing it, and a tendency to bubbles. He warns that silver was once as good as gold and then fell roughly 75%, which shows the danger of such assets. He is not predicting a collapse, but he sees no reason central banks would give up gold, partly because doing so would bankrupt them.

  “I just I hate it as an asset. ... I see no reason why, central banks will give up on gold, not least because it would have the effect of of bankrupting them, I suppose.”at 39:41

Central banks outside the US

• Chris Clothier · CG Asset Management

  UK real gilt yields are very attractive versus history and long-term growth; five-year gilts are comfortable holdings.

  Real yields on government bonds are, in Clothier's view, fundamentally very attractive compared with history and with long-term growth rates. He is very concerned about a fiscal crisis and its effect on the long end of the curve, but expects short rates to be cut if a fiscal crisis or AI slowdown occurs. That supports owning five-year nominal or index-linked gilts with a lot of comfort, and he is constructive on the yield curve out about five to six years.

  “real yields in government bond markets are fundamentally very attractive compared to history, compared to long term growth rates.”at 37:46

• Chris Clothier · CG Asset Management

  Four UK rate hikes priced over the next year look overdone; short rates likely to be cut in a slowdown.

  Markets price about four UK rate hikes over the next twelve months. Even after a recent elevated UK inflation print, Clothier thinks that pricing is overdone. He argues that if a fiscal crisis or an AI slowdown hits, short-term interest rates will be cut. That limits how far the yield curve can steepen and makes five- to six-year nominal and index-linked gilts a comfortable holding, which his fund is constructive on.

  “Even with the recent elevated inflation print that we got in The UK, I still think that looks overdone. ... if there is a fiscal crisis and or if there is the AI slowdown ... is that short term interest rates will get cut.”at 38:26

US growth & recession risk

• Chris Clothier · CG Asset Management

  An AI-led equity sell-off could trigger a nasty US recession, though not a financial crisis.

  If AI earnings disappoint and the equity market falls, Chris argues the US economy is more tied to the stock market than ever, so wealth effects would feed through quickly. He says the savings rate is low, the federal government is running a 6% deficit, and corporates are spending at a scale not seen since the railway boom. That leaves few fiscal stabilisers, so a nasty recession is entirely possible even if there is no financial crisis.

  “perhaps we don't have a financial crisis, but it's entirely possible that we have a nasty recession.”at 20:00

AI chips & semiconductors

• Chris Clothier · CG Asset Management

  Chip companies' earnings are a big share of S&P profits, so an AI capex slowdown would hit them hard.

  Chris and his team estimated about $400B of profit pool from $1T of annual capex, because chip companies have very high gross margins. Chip companies were about 10% of S&P earnings this year and are estimated at 17% next year. A credit-driven slowdown in AI spending would therefore hit S&P earnings first, and the banks' earnings, which rely on trading and fixed income origination, would also suffer second.

  “We came up with a number of about 400,000,000,000 because of the fact that particularly the chip companies have such high gross margins. Chip companies proportion of S and P earnings this year is 10% and their estimated share of S and P earnings next year is 17%.”at 20:09

US deficit & government debt

• Chris Clothier · CG Asset Management

  With the federal deficit at 6% of GDP, government cannot step in with more spending in a downturn.

  Chris says the federal government is running a 6% budget deficit, while corporates and households are also in spending mode. He argues the classic fiscal stabilisers cannot come from a government that is already running this deficit, so it cannot step in with more spending if the AI boom unwinds.

  “So obviously, the federal government is running a 6% budget deficit. ... I can't see that the federal government can start spending even more.”at 21:22

US consumer

• Chris Clothier · CG Asset Management

  US savings rate is far below its long-run average, so it could jump in an earnings slowdown.

  Chris notes the BLS revised the savings ratio, but says the savings rate is still well below its long-term average because the rich are richer than thought and tend to save. He argues the savings rate could rise very dramatically if there is a major earnings slowdown in the stock market, which would weigh on the consumer and the economy.

  “savings rate is well below its long term average. And so you could entirely see that the savings rate would rise very dramatically if there were a major earnings slowdown in the stock market.”at 21:47

US inflation

• Chris Clothier · CG Asset Management

  Energy shocks, war, deglobalisation and rising emerging wages make inflation likely to persist longer term.

  Since the Second World War, every major inflation surge has come from an energy price shock or a war, and both are in play now. Deglobalisation is taking hold, and combined with rising wages in emerging economies it makes the goods deflation of the past twenty-five years hard to sustain. War also destroys capital and realigns trade flows, all of which is inflationary.

  “If you go back to the second world war, since then, every major surge in inflation has come from some combination of an energy price shock or a war. And presently, we've got both of those situations in play.”at 23:05

Defense

• Chris Clothier · CG Asset Management

  Defense spending is economically nonproductive, so diverting money to it adds to inflationary pressure.

  Chris says defence spending is clearly desirable but does not increase the productive capacity of the economy, and it consumes resources. Money diverted towards it is therefore not productive, and he adds this to the inflationary backdrop he has already described.

  “the point about defense spending is that it is economically nonproductive. Now it's clearly desirable, but it doesn't actually do anything useful. It doesn't increase the productive capacity of the economy.”at 24:17

Europe's economy

• Chris Clothier · CG Asset Management

  UK gilts offer no appeal; US yields drive UK yields and few need long-dated gilts.

  Chris says nothing the chancellor could do would make him want thirty-year gilts. The UK is a small open economy, few people need to own long-dated gilts because defined benefit pension funds are closed and in surplus, and the biggest driver of UK yields is what happens to US yields. He also expects gilts would not be a safe haven in a French bond crisis, since these markets trade in sympathy with each other.

  “there's nothing that the chancellor could do that would make me want to own thirty year gilts. And the reason for that is that The UK is, as as we know, a small open economy and increasingly few people need to own long dated gilts.”at 33:44

AI bubble & tech valuations

• Chris Clothier · CG Asset Management

  Low global savings rates could limit the capital funding the AI-driven boom, halting it.

  Clothier points out that the savings rate across the world as a whole is pretty low at the moment. Savings are the source of capital for the boom he is discussing, so he thinks it could all come to a shuttering halt. He admits the timing is impossible to predict. The context is partly cut off, so the exact boom referred to is only inferred from the surrounding AI discussion.

  “the savings rate, you know, at least across the world as a whole is pretty low at the moment. And so, yeah, I think this could all come to a shuttering halt.”at 37:20

Credit markets

• Chris Clothier · CG Asset Management

  Credit spreads are very tight and a reckoning is coming; short-duration credit limits price risk.

  Clothier says credit spreads are very tight and he thinks there is a reckoning to come in credit markets. If spreads widen, the price risk on relatively short-duration credit is modest, so holding it is a perfectly sensible place to be. The bearish call on spreads is the core point and the short-duration positioning is his mitigation.

  “credit spreads are very tight at the moment, and I think that there is a reckoning to come in in credit markets. ... if you're buying relatively short duration credit, the price risk ... if spreads widen is relatively modest.”at 38:55

Earnings & corporate profits

• Chris Clothier · CG Asset Management

  Large government deficits inflate corporate profits; a fiscal retrenchment would hit equities hard.

  Clothier cites the Kalecki equation, which says large government deficits show up in corporate profits. On that logic, corporate profits are artificially elevated by government spending. He expects that spending to continue for a long time, but says that if it stops, it will be very nasty for equity markets. Western governments would first need a nasty slowdown to get a grip on their fiscal situation, which he sees as the precondition for a more sustainable footing.

  “large government deficits show up in corporate profits. So corporate profits are artificially elevated by the fact that essentially where we have all of this, government spending.”at 41:16