What caught my eye this week.
I am definitely not saying that the massive megacap tech deals of the past fortnight will for sure end badly. Let alone that such deals must mark the top of this bull market.
Credit with me some learning!
After more than two decades of meddling in the markets – and nearly as long making public pronouncements on this blog and elsewhere – it’d be a sackable offence for me not to have learned a bit of humility along the way.
Markets can remain irrational longer than you can believe what you’re reading on social media, as Keynes almost said.
Markets also have a way of turning today’s ‘irrationality’ into tomorrow’s ‘crucial staging point that any fool could have identified’ if you wait long enough for the proper perspective.
So yes, Nvidia investing $100bn in its major client OpenAI – or Oracle leveraging up to build out the data centre capacity required to fulfil the staggering $300bn compute deal it signed with OpenAI last week, which in turn inflated Oracle’s share price – may not be as Ponzi-ish as they look.
But these mind-bendingly big deals at the very least represent a gear shift.
Hitherto the hyperscalers (Meta, Google, Amazon et al) were just reinvesting their vast torrents of cashflow into building more data centres for paying customers. It was business as usual, albeit on steroids.
However this new phase is more self-referential. Something akin to a tech oligarch blood pact, where they are going all-in on the AI revolution and they’ll sink or swim together.
What’s My Age Again?
Sadly, I’m too old not to remember the Dotcom boom and bust at a time like this.
Not just in terms of the high valuations. (I’m thinking here of the likes of Palantir and OpenAI rather than the Magnificent Seven tech giants, most of which still don’t seem truly crazily-priced given their sales growth and margins.)
No, also in the way that Dotcom-era start-ups floated on an ocean of ultimately VC-funded advertising that paid the bills of a bunch of other start-ups, which ultimately took half of them down when somebody thought to ask “how many people are actually clicking on these things?” and pulled the plug when they got a straight answer.
I mean, doesn’t nVidia investing in OpenAI so that OpenAI can get chips from nVidia have the whiff of that to you?
Even so, you might imagine that none of this matters for your portfolio. But what if I quoted JP Morgan informing us this week that:
AI related stocks have accounted for 75% of S&P 500 returns, 80% of earnings growth, and 90% of capital spending growth since ChatGPT launched in November 2022.
This also seems like a pertinent point to remind you that US equities account for 60-70% or more of global tracker funds.
Artificial Intelligence taking all the jobs or becoming super-intelligent is one thing.
But this AI boom being exposed as productivity-sapping margin-crushing hype on a nation-state-GDP level would also cause us investors plenty of pain.
Wild Wild West
Accordingly, I’ve been worried and underweight the US at least 18 months. And boy hasn’t it made keeping within spitting distance of the global market’s return difficult.
Because once more with feeling:
AI related stocks have accounted for 75% of S&P 500 returns, 80% of earnings growth, and 90% of capital spending growth since ChatGPT launched in November 2022.
On the other hand, Wall Street’s obsession with big tech and AI has left lots of other stuff looking good value, especially outside of the US. And it’s fun to hunt around for it.
The activist manager Saba, for example, is awaiting approval for a new ETF it’s launching that will enable value-minded investors to buy a basket of UK investment trusts – specifically because so many of them are still going cheap.
Again, the parallels are obvious.
Nobody wanted to own Warren Buffett’s Berkshire Hathaway at the peak of the Dotcom Bubble. Then in the years after the bubble bust, value soared.
The set-up looks so easy, right? Yeah, too easy. We could be in 1996, say, not 1999. More importantly we’re actually in 2025, and stock market history rhymes rather than repeats.
So all I’m saying for sure is that if this is a bubble and if it does burst, then you’ll hear a lot about nVidia putting $100bn into OpenAI in every future account of it.
No Scrubs
Oh, and incidentally people keep saying the hyperscalers are spending tens of billions ‘building out the AI infrastructure’ as if they were laying down concrete.
But anyone who has ever bought a new nVidia graphics card to play the latest PC games will have found themselves confronted with jerky frame rates six months later.
These things go stale faster than you can say “whatever happened to the Metaverse?”
So if they are building out the AI infrastructure, they’re going to have to build it out again…
Have a great weekend.
From Monevator
Crypto ETNS: what you need to know – Monevator
Last chance to buy some Monevator T-shirts before we rationalise our shop – Monevator
From the archive-ator: Patient investing requires a little faith – Monevator
News
New digital ID scheme to be rolled out across UK – GOV.UK
Retail sales slump for twelfth month in a row – This Is Money
UK forecast to have highest inflation rate among the rich nations – BBC
Record number of savers and investors facing a tax bill… – Which
…while half of the population live in households that get more in benefits than they pay in tax – T.I.M.
Trump’s $100,000 H-1B fee sparks a global race for top talent – CNBC
Reeves urged to take 2p off employee NI and add it to income tax in Budget – Guardian
£2,000 savings buffer can be the turning point for financial wellbeing – Yahoo Finance
HSBC demonstrates first-known quantum algorithmic trading with IBM – HSBC
Wealthy investors from US, China, and Hong Kong apply for New Zealand’s ‘golden visa’ scheme – Guardian
Momentum is crushing value this year. Again. – Sherwood
Products and services
Is Santander’s new Edge Explorer bank account worth it? – Which
IKEA Family credit card review – Be Clever With Your Cash
Nationwide trims mortgage rates despite BoE hold – This Is Money
Get up to £1,500 cashback when you transfer your cash and/or investments to Charles Stanley Direct through this affiliate link. Terms apply – Charles Stanley
How to avoid falling victim to a ‘money mule’ scam – Guardian
Barclays cuts mortgage rates for home buyers with smaller deposits – This Is Money
Get up to £200 cashback when you open or switch to an Interactive Investor SIPP. Terms and fees apply, affiliate link. – Interactive Investor
How does home equity release work? – This Is Money
Supermarket Christmas delivery slots – Be Clever With Your Cash
Homes to rent for up to £1,800… in pictures – Guardian
Comment and opinion
How can we solve the 60% tax trap without damaging the economy? – This Is Money
Why the 5% Rule is the new 4% rule… – Of Dollars and Data
…although US retirees actually tend to follow a 2% rule – A Wealth of Common Sense
How to minimise active ETF bid/ask spreads [US but relevant] – Morningstar
If sustained growth is the aim, the UK must alleviate child poverty – Observer
Welcome to London, divorce capital of the world [Paywall] – FT
How you can profit from the coming devaluation – Simple Living in Somerset
Millions of Americans are becoming economically invisible – Bloomberg via AP
Debunking the active fund ‘persistence scorecard’ debunking – FT
Naughty corner: Active antics
Exploring capital efficiency [PDF, nerdy] – AQR
What activist managers buy in Japan – Verdad
Nope, sorry, the S&P 500 is not the new risk-free rate – FT
How pensions windfall could turbocharge UK stock market [Affiliate link] – II
The perils of concentration – Market Sentiment
American Express: an empire of plastic – Quartr
Kindle book bargains
Flash Boys by Michael Lewis – £0.99 on Kindle
Alchemy by Rory Sutherland – £0.99 on Kindle
The Green Budget Guide by Nancy Birtwhistle – £0.99 on Kindle
Techno Feudalism by Yanis Varoufakis – £0.99 on Kindle
Or pick up one of the all-time investing greats – Monevator shop
Environmental factors
China has finally pledged to cut carbon emissions – Sherwood
Demand for oil will not peak until 2030, warns BP – This Is Money
New homes may be forced to fit water-saving toilets and showers – Sky
The near-extinction of rhinos is at risk of being normalised – The Conversation
Ocean acidity crosses critical threshold for marine life – Guardian
Robot overlord roundup
The data centre blob [PDF] – JP Morgan
Chatbait is taking over the Internet – The Atlantic [h/t Abnormal Returns]
The algorithm will see you now – Works in Progress
If Anyone Builds It, Everyone Dies review – Guardian
AI generated workslop is here… – CNBC
…and it’s destroying workplace productivity – Harvard Business Review
American companies talk about AI, but can’t explain the upsides [Paywall] – FT
Not at the dinner table
The moral case against Nigel Farage – The Newsletter of (Not Quite) Everything
From low taxes to economic fragility – Klement on Investing
A left-wing version of Trump isn’t the answer – The Argument
Why is Trump bailing out Argentina? – Paul Krugman
Sometimes democracy works: on same-sex marriages – Aeon
Letter from an ICE detention facility – Bitter Southerner
Trump’s move against the media is an authoritarian classic – A.P.
Who’s getting rich off your attention? – Kyla Scanlon
Things are really bad folks – Freddie deBoer
Off our beat
Why Putin can’t afford to let Ukraine prosper – WSJ
National Railway Museum reopens after £11m refit – Guardian
25 interesting ideas from 2025 – Derek Thompson
The king of coffee nerds – Financial Times [h/t Abnormal Returns]
Is the Golden Age of TV officially over? – Stat Significant
Perspective on life, money, and success from rock star Billy Corgan [Video] – via X
Scammed into scamming – Reuters
Mid-20th Century culture is getting erased – The Honest Broker
Why every country needs to master the electric tech stack – Noahpinion
‘Very mean’ squirrel has sent two people to ER in Californian city – Associated Press
And finally…
“Greed is good because it makes things predictable. No need to coerce or enforce or foist any delusions when you have people volunteering to do the labor of self-persuasion.”
– Carrie Sun, Private Equity: A Memoir
Like these links? Subscribe to get them every Saturday. Note this article includes affiliate links, such as from Amazon and Interactive Investor.
“…while half of the population live in households that get more in benefits than they pay in tax”
If you choose – misleadingly, in my view – t0 categorise the State Retirement Pension as a “benefit” that result seems almost inevitable. But if you used a better definition, is it still true?
Yes, with the US market dominated by a very few big companies all at high PEs and all operating in roughly the same space (or at least whose fortunes are somewhat tied to “tech”) and that US market dominating global trackers it’s very hard to feel enthusiastic or even indifferent to just adding to your global trackers holding every month no matter how much of a passive investor I want to credit myself as being (and hard to feel enthusiastic about the US anyway with His Donaldness running the show). So much so I do wonder if I should be looking at putting 50% into a global tracker and the other 50% into “other” (world ex US tracker or Europe / UK / Asia trackers as suits your prejudice).
Not the first person to wonder / worry about this of course (may even be the last person to actually start to think about doing something about it!).
@David 14 – I very much share your discomfort. That discomfort is amplified by being essentially at my FI number. More and more, I keep thinking of mantras like ‘be fearful when others are greedy’ and ‘if you’ve won the game then don’t swing for the fences’. It would be great to get a Monevator article about CAPE ratios, esp as it pertains to those on the cusp of FI (though not necessarily RE)
Imo there is a at one level a fairly normal cyclical AI bubble that should burst but leave behind a few winners (imo Google looks good, Meta etc can keep their AI video slop). Within that you have the normal slightly dodgy behavior – and Nvidia has always funded its customers aided by its high valuation and has a few red flags running through the accounts, but in these bubbles that is normal.
But if you step back with a US focus you also have imo high house prices (see UK also), incredible household debt (lesser in UK), high earnings multiples across the board in the US (opposite to UK) and a Private Equity bubble to layer in (lesser in UK). Plus whatever crypto is (lesser in UK).
On Private Equity, you have the likes of HubSpot that never really have earnings but get through the PE-cycle and list to now be worth $27bn while always losing money, now as a fully mature company. That in turn creates confusion over what value is – is it earnings or revenue? And that seeps out, so you have Palantir and Tesla valued on the idea of something vague in the future. That imo is not a
part of the AI bubble, but results more from the PE bubble and from the debt bubble (eventually borrowing to speculate). And it’s quite confusing having multiple bubbles in play at once.
In the end it feels to me like there is a broader growing volcano fed by the belief that all this can keep expanding. But it’s hard to read because it’s like a multiheaded snake.
@Andy/David : I’m also in the “if you’ve won, quit playing so much” space – have toned down equities in my portfolio. I’ve even bought a chunk of long long term index-linked gilts, which in theory provide a guaranteed over-inflation return. Somehow, with the state of the UK I can’t quite feel fully confident in that however.
Whilst nothing is a crystal ball, we are at CAPE levels only seen really in the dotcom crisis. They could of course go above and beyond that in theory, but I think there’s no shame in taking this time to review whether one’s asset allocation is appropriate.
@Part-Time Analyst
“…is it earnings or revenue?”
Did you mean “profit or revenue”? I thought earnings and revenue are the same thing. Excellent comment, by the way.
Only just saw also that Jonathan Clements from Humble Dollar passed away this week too, sad news. He was an excellent writer.
Expensive US markets, highly concentrated, cause concern to many investors, it has caused me to mentally flip flop from sticking to plan and becoming more cautious , with some portfolio changes made, with poor conviction if I am honest.
It would be fair to say that the magnificent 7 are very expensive, the rest of the US market has an elevated CAPE ( say 33).
Looking at the All World , the 40% non US holdings are probably fairly valued, CAPE about 20 , the remaining US market is 60% of the whole, is expensive with around ⅓ of that being very expensive.
On aggregate The All World is expensive but does that have sufficient predicative power to be more than a little cautious ?
I am reminded of the Peter Lynch quote
“Far more money has been lost by investors preparing for corrections, or trying to anticipate corrections, than has been lost in the corrections themselves.”
dearieme – State Pension is a welfare benefit for almost all recipients, as their working life NI contributions fall way short of the value of the state pension they received – both in strict monetary terms, as well as the fact its zero risk, adjusted for inflation etc.
@JDW — Indeed, very sad. I’ve included a few of Jonathan’s posts since he had his diagnosis but for some reason I felt uncomfortable including his final post in the links. I can’t explain exactly why… I think it was because it was written almost as a personal letter to US readers who have known him for decade, and it felt a bit inauthentic? A bit silly in retrospect if s.
Here it is for those who enjoyed his writing:
https://humbledollar.com/forum/farewell-friends/
@TI – thanks for adding that link to Jonathan Clements, sad news, I always enjoyed his posts. Interesting to see the back story, I hadn’t picked up that he was British.
I have been underweight in US for years, even before the LLM AI appearance. I have no regrets. Even if/when this all shakes out I don’t see the need to take huge region concentration risks – if Japan was 70% of the world index, I wouldn’t be holding a single world index either, it’s not just an anti-AI protection move – at what point is a world index not diversified? ~70% in any one region must already be getting close.
Although, if, like me, you are intentionally down on US but higher percentages elsewhere you probably saw fairly large japan and (asia-japan) index gains just recently. For the first time in a long time my rebalance involved selling quite a lot of japan and asia-japan.
At the level of an individual investor, a 70% US weight in a world index leads to huge losses if the AI bubble bursts. Add to this political risk associated with US…
Ironically, nvidia still has a good business if AI bust happens, as they were doing the exact same work/research pre-AI, just the gpu’s and underlying code was bought by end users for playing pc games instead. They will need to pay folk off and readjust back to normal growth etc, but all the efforts in improving their gpu offerings for both crypto and now AI will be just as useful in the future for what was once their core business. They aren’t holding huge inventory, so if AI goes pop they go back to putting complex mathematics in GPUs for games. Those companies that now have factories full of racks of nvidia gpus at high expense, they have real problems if AI takes a back seat.
The sad thing about AI so far is if you put aside mediocre chat bots and summarising documents (with some hallucinations thrown in just to make the current output completely unusable if its important work or has accountability requirements), the only improved use case for this generation of AI is unmanned military drones etc, which is quite sad. The other uses of AI being thrown about like medical imaging and such is not new and been happening for decades, everyone is just throwing AI in to funding applications to win funding or to improve their “impact” scores.
Thanks for bringing the ‘state of the rhino’ report to the attention of the readership.
@TI – very timely.
Last week I shuffled my defensive assets. I sold some BHMG to buy Pershing Square Holdings (I’m hoping at least one of them can repeat 2008 and hit it out of the park if SHTF). To counter the increase in US in PSH, I sold some HSBC FTSE All World and I’ll buy a All-World ex-US to keep US as is.
Not very passivista, but there you go.
I was a great follower of Jonathan Clements -he helped a lot of people and will be greatly missed
He was of course a believer and promoter of “ indexing ” long before it became so fashionable
Global indexing in equities and bonds have been a success with funding my 23 years of retirement -so far
I don’t really feel any urgent need to change and certainly am constantly unable to second guess the market
One global index fund for equities and bonds respectively certainly makes investing life cheap ,very easy to understand and manage
Very boring of course especially if your interest is the stockmarket and its various shenanigans
Some very special investors can beat the market return-sadly I realised fairly early into my investing career that I was not amongst the lucky few
I suppose it was always true that only a very few companies actually move the stockmarket-5%?-also these successful companies tend to be different every year etc etc
Sitting tight here
xxd09
@TI:
The “Millions of Americans are becoming economically invisible” Bloomberg article seems to be the antithesis of the theme of last weeks weekend reading.
The picture the article paints sounds, at best, somewhat unstable to me. Any thoughts?
Thanks for all the other interesting links too.
Those subheadings 😉 I’m partying like it’s 1999 in my head!
@Curlew – Earnings was perhaps a bit careless on my part to use as it can have a dual meaning, but I meant in terms of profits. So earnings as in the E in EBIT or EBTIDA.
Now is the LLM winter of our discontent, made glorious ASI summer by this Son of Neural Nets.
LLMs predict language. They don’t predict for the World. They’re the perfect info mirror.
But there’s nothing new in the mirror. Just a reflection looking back at us.
Without true novelty there can’t be the real breakthroughs needed for additional growth, less still a tech / econ singularity to solve the many ills ailing the world.
Aside from the fundamental foundational flaw (another ’empty box’ as SBF put it for crypto), there’s also the technical barrier of current approaches: exponential increases needed for linear improvement using compute / parameter scaling.
LLMs alone can’t be the solution.
Maybe they are part of the solution.
Maybe they’re a distraction and dead end.
But investment into AGI based on LLMs alone *will* fail that objective.
This emphatically doesn’t mean though that the search for AGI ends.
We know it is possible empirically from nature – Darwinism produced us.
We know it is possible from our well tested foundational model of physics.
Generalised intelligence (and autonomous super intelligence) is/will be just emergent complexity from quarks and fields.
Like everything else. In truth, all just Atoms and the Void, as Democritus put it.
Echoing the caution in this thread about LLM (mis-sold as AI) driven market froth, I’ve tracked similar skepticism for the past year via 100+ comments now on Monevator’s May 2024 “First they came for the call centres” piece over here:
https://monevator.com/weekend-reading-first-they-came-for-the-call-centres/
TL:DR: Starting with barbell strategies tilting to momentum ETFs like SPYO for AI upside, while hedging with ex US small cap value for mean reversion, my views evolved: following LLM critics like Gary Marcus / Ed Zitron, doubting scaling to AGI, and likening it to 1990s nanotech hype constrained by physics. And in the @TI above linked “Queasy about US Valuations” piece (also 2024), this also extended to broader market unease, advocating modest shifts (e.g., 20% to bonds) for asymmetric protection, toy models show 2:1 upside/downside ratios favoring small tweaks, plus return stacking via WTEF ETF to cut S&P exposure from 65% to 42% while boosting total allocation to the equivalent of 130% with gilts and ex US equities.
We need to know our history on this.
It’s not just a rerun of Dot.com from 1996-99 or the Nifty Fifty from 1970-72.
There’s a longer term pattern here.
Natural Language Processing’s post-1958 timeline (Mark 1 Perceptron, 1st artificial neural net) amplifies the current AI winter risks: ELIZA (1966) ignited hype, but AI winters still struck hatd (1974-80, 1987-93) post-over promises. Revivals via 1980s stat based NLPs, early attempts at World models, and development of what became the Internet led to hiatus until 2010s deep learning (Word2Vec 2013), Transformers (2017), and multimodal advances (GPT-4o 2024, Grok-4 2025).
Now hallucinations and inevitable near term ROI shortfalls (on the truly staggering over investment into increasingly rapidly obsolescent intensive GPU / TPU/ NPU data centre architectures) all loudly signal pending disillusionment.
Yet, this trough could yet still seed ASI breakthroughs via much neglected hybrid neural symbolic approaches.
Portfolio fallout? An AI winter craters Magnificent 7 like Dot.com’s 75% plunge.
So I’m staying underweight US tech, diversifying globally: history cycles, rewarding patient investors eyeing the thaw.
Enjoyed (if that’s the right word) the ‘who’s getting rich off your attention’ article.
Great context to the broader theme of omni-slop AI content, while being suitably bleak for a grey Sunday morning.
Two interesting articles in the Economist, one relevant to this.
Apparently the use of AI in internet searches is destroying the value of ads on the internet (rather than going to the usual items with ads in them, internet users are going to the AI search to get their answer). If true, advertisers will surely notice and down go the huge ad revenues. Hoist by one’s own petard comes to mind.
Unrelated, the Economist also did a lucid piece on a looming UK Government debt crisis. It cites Labour’s ideological unwillingness to do the one thing which can avoid the bond market making life impossible ie cutting the welfare bill (15% of working age people on benefits etc).
@Hospitaller
Re the Economist article: “… It cites Labour’s ideological unwillingness to do the one thing which can avoid the bond market making life impossible ie cutting the welfare bill”
The article clearly stated two things not one (“…it is abundantly clear that the fiscal adjustment should start with pensions and the welfare budget.“) before expanding on these two things.
The article is available at https://www.economist.com/leaders/2025/09/25/britain-is-slowly-going-bust though it’s paywalled (but if you set up create a free account they let you read a few items gratis).
RC #12 DH #19 – the thing is, the ability to summarise something automatically is pretty useful, summarising is probably selling the technique a little short, being able to wrangle information into the way you want to consume it is almost a superpower. The WWW provided a means to access information, through a delivery mechanism and search facility. It sounds dull but it turned out to be revolutionary. The ability to curate the worlds knowledge into a format that suits you, automagically and within seconds, to aid its consumption may well be equally revolutionary?
That said, being revolutionary doesn’t preclude huge busts en-route as per 1999.
The previous AI winters were really because the theory was great but none of it actually did anything useful. That isn’t now the case, we’ve seen huge step changes in performance of AI models across multiple pardigms, CV (CNNs), RNNs (inc. LLMs), RL etc. the difference being some new, novel architectures but critically, the compute to support it. Chat-bots and drones doesn’t really cover it.
You don’t need AGI for all this to be game-changeingly useful. Novelty isn’t where 99.9% of the work unfolds.
RC #12, Rhino #23: Current monetised LLMs (OpenAI, Anthropic, Mistral, Grok, etc) together generates in the low billions of revenue (and larger losses) annually. That’s a rounding error against even one hyperscaler’s merely current yearly Capex (~$60–100 bn).
McKinsey boldly est ~$6.7 Tn data centre Capex to 2030, roughly $5.2 Tn of which targets frontier model grade compute.
*If* that happens, then Nvidia likely becomes a $800-$1,100/share ($20-27 Tn cap) stock (~4-7x from now).
But No Guests = No Party.
Sure sunk + fully legally committed investments raise floor, as site builds already started (land, power, cooling) & multi yrs supply contracts mean they’ll be some extra capacity led min. level of additional demand persistence.
However, will several trillions+ of GPU Capex actually show up in just the next few yrs?
It all doesn’t add up to much unless and until enterprise + individual LLM specific profits – not revenues – are nearing $1 Tn in 2030.
My sense on that number is that it’s a LoL one.
And, while hyperscalers can repurpose racks from one workload to another (cloud/HPC, enterprise inference, simulation, gov/sovereign projects), all this alleged future demand ultimately depends upon LLMs being sufficiently commercially transformative (to the accounting bottom line, and not just with BS generation) that staff can be replaced at scale (many millions laid off), thereby justifying businesses spending the well over $1 Tn in 2030 (likely very substantially more) which will be necessary to generate ~$1 Tn in profits for model providers that year to keep feeding the insaitable Capex monster.
Seems one hell of stretch from here just now. The only scaling I see are in OpenAI’s losses. This could be Metaverse times 100 or WeWork times 1,000.
LLM’s are the reason for all the recent public AI hype and investment, and these are where most the problems lie, not AI in the general sense – every LLM tool has a disclaimer along the lines “results may be innacurate”. The most useful areas where modern inference approaches will contribute anything to society or even in business is in areas where work has been progressing for decades long before chat GPT i.e. not LLM’s.
In terms of investing, the useful inference work would still be happening at maybe only a reduced rate without LLM’s – the funding and progress has increased as researchers are gaming the funding applications process by putting things under AI and with more hardware etc, so we do have some benefits, and some good research is happening. However, as someone who has formally researched in statistical inference and spent a career in the application of statistical inference, I stand by my opinion that chat GPT spurred huge investments in technology that will struggle to ever lose the disclaimers about innacuracies – to me as someone who cares about correctness, this is a problem. Whether this lack of rigour leads to stock market implosions is not something I know, but I choose where I invest my money, and the AI hype is being driven by biased, hallucinating, LLMs tools. I am not that interested to be investing in these tools despite my entire and current career being in areas concerned with the methodologies.
I might be proved wrong in a decade, but that is fine, I make my money from building useful models, not from investing in inference tools that are… mostly ok, probably close to what your training data says, but might be wrong in places sometimes, give wrong sources, and plain make stuff up in others?
That is my position. If chat GPT didn’t get released, the AI investments of recent years would not have happened. So the decision to invest to support it is not about general inference and AI (which has solid foundations and happening for decades), but whether you think LLM’s justify the hype.
@RC #25: it’s so hard to know/guess what to do.
Never bet against America says Buffett. But, the US underperformed (for example) EMs horrendously in 2000-2008.
As Soros puts it (paraphrasing heavily), every bubble has a rational foundation, and an irrational response to it.
Businesses will be built on, or adjacent to, LLMs, but which will reward investors?
Which have even sustainable business models?
Does ~50x for Nvidia make sense on a 30 year ‘risk free’ rate of 5%?
Example (recalling point (6) in #88 (21 June 25) in the “First they came for the call centres” thread): when the car was first fully commercialised there were hundreds of listed manufacturers. Most went bust.
But even if you’d had the luck or foresight to hold a successful one like Ford, then during 1980 to 2020 you ‘only’ made 23x.
However, of all the companies out there, it was Walmart which benefited most from the car, as shoppers could visit massive suburban stores off the highway (a 1,600x return over that same period).
This feels like one of the more difficult times to be an investor.
Not because of deglobalisation, populist nonsense, or the sinister shadows of staflation; but owing instead to the recent, seeming emergent, exogenous shock of a ‘new’ technology which is so hyped, but still just might be a progenitor of sorts to the supposed solution of all our problems (or the harbinger of misaligned doom).
Maybe diversification is the only rational response:
https://www.telegraph.co.uk/business/2025/09/25/history-shows-investors-that-it-pays-to-diversify/
But how much (and how to*) diversify away from the US?
What is too much (or too little) risk to, in effect, ‘go short’ the US tech darlings (or at least the semis) by under weighting them vis a vie their cap weights?
*Tom Stevenson has a rose tinted view of the attributes of active managed funds IMO; but for DM and EM micro and nano cap or Frontier Markets you do have to fall back on specialist ITs and open ended funds because there just aren’t passive alternatives.
@Kamae — Forgot to say, glad someone noticed and got it! 😉
Letter from an ICE detention facility – Bitter Southerner
An illegal immigrant who managed to get away with it through what seems to be an administrative error (maybe deliberate) in 2012 and couldn’t keep a low profile. What an idiot, he has a family.
@CC #28: What are you talking about? There’s no mention of 2012 or of an administrative error in the article. It says that he entered the US lawfully in 2004, and is entitled to permanent residence (read to the article’s end including the endnote). If that is correct (and there’s no other information given us) then how exactly is this unfortunate individual an ‘illegal immigrant’? If the facts as presented in the piece are correct, then the opposite would appear to be true. In any event, his appalling treatment is exactly what I’d expect from regime of the fascist clown in the White House. I’ve never wanted to visit the US and, frankly, even in more normal times than these its bland and banal plastic culture repels me. I’d hope that other UK citizens will consider boycotting it now that its true and unpleasant character is coming to the fore. At some point silence becomes complicity. Trump has not made America great, but rather, instead, has made it both the laughing stock of the world and a terrible warning to it. It’s sad that our late Queen and now also our King has to entertain him.
@Delta With certain the non financial articles its best to check for further info.
Well wiki says this: “Guevara was ordered to leave the United States in 2012.The case was appealed and subsequently closed, with him being allowed to remain in the country. Guevara eventually obtained a work visa, he has a pending green card application.” If that’s correct, then:
1. He was allowed to remain, for whatever reason.
2. He obtained a work visa.
3. He was still working when detained.
4. He had applied for a Green Card whilst covered by a work visa.
Perhaps he wasn’t actually entitled to a work visa. But that’s the very point of having the rule of law, with independent and impartial courts to thoroughly, fairly and (hopefully) expeditiously investigate in order to find the relevant facts, and then apply the law to those facts, as found.
In the meantime, the inhumanity of treatment and disregard for judicial authority over executive action is more than troubling.
We all need to show some elementary compassion.
How would you feel if you were in his shoes?
If it ever had it, then the US isn’t showing much of it right now.
It’s the Friday after this weekend reading now. 3rd Oct.
WRL & S&P500 at an all time high (again) today.
After this weekend reading I’d been mulling over taking a little bit of profit this week.
Discussed with the other half & we decided to sell off a v small portion of our ISAs today to pay off a chunk of mortgage.
Gains seen in the last 3 weeks have been enough to knock 3 years off our mortgage (a good amount in ISAs vs v low mortgage payements).
Yes, I know all the rules about trying not to time the market, & probably the market will now continue to go up for another year making this look foolish, but on a personal level it’s a good time to lock in some profits. The eldest will start Uni in Sept 2027 & this move clears our mortgage before this, so cashflow will be improved at a time we’ll be seeing some more outgoings supporting life at Uni.
@Delta
I’d feel like an idiot who has ruined a good thing and will now be separated from family.
Came on a tourist visa and didnt go back to his country. Was asked to leave 2012, he didnt, appealed it, nothing happened either way so he carried on.
They are obviously getting rid of him because of his journalism of course as he is not a criminal (apart from illegally staying in US originally).
Re: “ruined a good thing”: fair enough as your PoV, but it seems that he originally entered the US lawfully (on a visa), so was not an illegal immigrant, then overstayed, was subject to an adverse 1st instance ruling, which he then appealed in time, and, one presumes (as, otherwise, what’s the point of appealing?), any deportation order made at 1st instance was stayed pending the appeal.
There may be an issue (if there was a stay) whether the appeal is still afoot if, as reported, ‘the case was closed’, and, therefore, whether there’s still a stay in effect on the original decision.
If there was a stay and if it is still extant, then ICE can’t just deport him.
There needs to be a judicial process to determine if the earlier undetermined appeal should continue (whether or not, as the case may be, that requires it’s reinstatement); and as to whether any stay on any order from the 1st set of proceedings should continue until the appellate court rules on the appeal.
If ICE don’t follow the rules then that’s the very definition of a lawless state.
And, candidly, not one that I’d ever want to visit myself.
Not wishing to name names – the virtue signally, condescension, dismissal of other views (and heavy handed responses) are starting to get irritating.
Around 2016 I found the website and I thought “I’d found my tribe”.
Perhaps it’s a case of “it’s not you it’s me”, but I’m not sure..
Healthy debate requires a difference of views and values, like with the motability discussion here:
https://monevator.com/weekend-reading-deckchair-rearranging-derailed/
And, by definition, politics polarises. Maybe that’s why these types of links are generally under “Not at the Dinner Table”.
Look, no one wants to silence anyone else here. I just think personally that this guy’s been treated shabbily. He’s not a criminal. He has due process rights. Perhaps given his circumstances he is (depending upon ones’ PoV) “an idiot” for doing journalism. But IMO he’s not less worthy of some basic measure of compassion.
Personally, I don’t want the UK to turn out like this, although I fear that’s the path which we’re on for after 2029.
@Delta
Apologies, illegal alien is the correct term when he overstayed in 2004.
Agreed. Indeed, an illegal alien.
My apologies too. I didn’t mean to come across (or be) hectoring or virtue signalling. I just got a bit triggered (after seeing #28) by reading this man’s story, which I appreciate is just a drop in the ocean of misfortune with what doesn’t get reported or widely read, like the wars in Sudan, the eastern DR Congo etc, and which involve incalculably greater horrors.
Thanks both for standing down amicably. These are tough times to discuss politics and goodness knows I don’t always get it right myself, but I don’t feel it should be omitted from the links. And any civil communication between opposing views has to be good.
Sorry also to have diverted the thread from the Capex theme of the W/e piece – returning to which:
– A maybe relevant thought from Oaktree’s Howard Marks, (paraphrasing) to the effect that a bubble is not related directly to price, but rather to investor psychology. Are we at the phase yet of the shoeshine boy giving tips about AI stocks? Jim Cramer doesn’t count here 😉
– Whether it’s canal mania in the 1790s, railway euphoria in the 1840s, the radio boom in the 1920s, the harddrive company craze of the 1980s, the fibre optic frenzy in the 1990s or the current LLM adjacent fever for semis, fabs and cloud compute companies; it’s generally the consumers, and not the investors, who end up with the benefits of overinvestment.
An interesting take on whether (or not) we’re in a bubble about to collapse – 5 out of 8 indicators met, so 62.5% chance, only slightly more than a coin toss:
https://youtu.be/d04KMZTUBv4?si=YoomfITMZkg4tblM
I listen to this guy a 2x speed on YouTube as it sounds better and obviously saves time. You might want to do likewise. This guy uses thumbnail click bait to get views but then delivers very down to earth, evidence based, don’t panic, substance. Fair do’s in my book.