***** denotes well-worth reading in full at source (even if excerpted extensively here)
Economic Fare:
It’s hard to read the room when you don’t speak the language of the markets. Consequently, the Fed is tightening into slowing earnings growth, like ‘00 and ‘18.
..................... We lean toward a 2027 equity recovery, but the tech pullback alone is not enough. We need a Fed pivot and bottoming earnings momentum, with stabilization unlikely before 2Q27 at the earliest.
.......... We believe backward-looking data are unlikely to dissuade the Fed from near-term hikes. A sharper tightening in financial conditions, including credit stress, may be needed; sentiment and positioning remain the immediate potential brakes on the selloff.
Final Thoughts: The Massive Mistake
· We continue to regard the Fed’s hawkish pivot and unbalanced rate and balance-sheet policy as a major mistake, with capital markets signaling risks that lagging employment and inflation data have yet to reflect. ................
Why the Yield Curve, Corporate Borrowing Costs, and Equity Markets May Be Signaling a Turning Point in the U.S. Economy
........................................................................................ The resulting index has a particularly useful economic interpretation. A positive reading indicates that financial conditions are exerting downward pressure on real GDP growth over the following year, while a negative reading indicates that financial conditions are providing support. An index reading of positive one corresponds to an estimated drag of approximately one percentage point on real GDP growth over the subsequent year, while a reading of negative one corresponds to an estimated boost of approximately one percentage point. These estimates are model-based approximations rather than unconditional forecasts of GDP growth, since financial variables themselves respond to developments in the economy and the index does not incorporate every possible financing constraint, such as changes in bank lending standards. Nevertheless, the methodology provides a direct link between financial-market developments and their potential macroeconomic consequences.
The historical behavior of the FCI-G is particularly revealing. The index captures the deterioration in financial conditions associated with the recessions of 1990–91, 2001, 2008–09 and 2020, but it also identifies episodes of monetary tightening that did not immediately produce a recession. Most importantly for our discussion, it shows the exceptional tightening that followed the Federal Reserve’s monetary policy reversal in 2022. The increase in the federal funds rate was accompanied by rising mortgage rates and corporate bond yields, declining equity valuations and a substantial appreciation of the dollar. The resulting tightening was therefore considerably broader than the inversion of the Treasury yield curve. It represented a simultaneous deterioration in several channels through which financial conditions influence private-sector spending and investment, with the cumulative effects continuing to weigh on economic activity after the initial rise in interest rates.
The data through September 2026 provide an interesting perspective. The FCI-G stood at approximately −0.78 percentage points, suggesting that the cumulative effects of financial conditions were still providing support to real GDP growth over the following year. However, the decomposition of the index reveals an important concentration of that support. Equity prices accounted for approximately −0.65 percentage points, representing the overwhelming majority of the estimated positive impulse, while the dollar provided additional accommodation and mortgage rates exerted a modest restraining influence. The aggregate reading therefore concealed considerable differences among the underlying financial variables. This distinction is critical because an index that appears accommodative primarily as a consequence of elevated equity valuations may be vulnerable to a sustained reversal in the stock market, even if other financing conditions remain relatively stable.
..................
Market Fare:
Meaningful market analysis is not about complicated theories of rotation or ephemeral lines in the sand.
Layering on more indicators, more model components or more charts usually creates more noise and more opportunities for behavioral biases to overwhelm an honest assessment of reality.
.................. How investors feel about the market often influences how they are positioned in the market. Right now there is ample room for increasingly cautious investors to shift away from near-record exposure to equities.
......... We are at 55 days and counting of more new lows than new highs, the third longest continuous stretch in over a decade. As long as new lows are more numerous than new highs, it is hard to construct a bullish participation argument, especially in the absence of a breadth thrust regime.
Kelly, JPM: The Technology Threat
.......... as the events of last week reminded us rather dramatically, the extraordinary advance of technology brings with it, not just the promise of improvement but the threat of disaster. Indeed, in broad terms, the greatest risk facing investors today may be the juxtaposition of ever more powerful technology, across multiple dimensions, with the wholly inadequate political and regulatory systems needed to control it. For voters and those they elect, this speaks to the need for a serious reassessment of the importance of regulation. For investors, it underscores the need to build portfolios which, if they can’t be expected to survive “apocalypse now”, should at least be constructed to weather “something really bad soon”. ...........
.......... as the events of last week reminded us rather dramatically, the extraordinary advance of technology brings with it, not just the promise of improvement but the threat of disaster. Indeed, in broad terms, the greatest risk facing investors today may be the juxtaposition of ever more powerful technology, across multiple dimensions, with the wholly inadequate political and regulatory systems needed to control it. For voters and those they elect, this speaks to the need for a serious reassessment of the importance of regulation. For investors, it underscores the need to build portfolios which, if they can’t be expected to survive “apocalypse now”, should at least be constructed to weather “something really bad soon”. ...........
I’ve been really dreading this. Global de-risking/deleveraging turned much more serious this week. The AI/tech implosion took a turn for the worse, while cross-asset position liquidations intensified. The leveraged speculating community has suddenly found itself under the most intense pressure since the pandemic. In short, Crisis Dynamics have been unleashed – and I doubt they will be easily contained.
From the deflating historic global Bubbles analytical perspective, this week appeared decisive. It was an especially ominous end to a troubling week.
But at least Wall Street got its beloved Anthropic IPO sold – a suitably grand punctuation mark for history’s greatest era of excess and wealth redistribution. It was also a quick $91 billion liquidity suck for a deleveraging backdrop facing mounting liquidity risk.
From a market instability perspective, it was the worst week since (at least) “liberation day.” The major difference being this week’s turbulence was not triggered by ill-conceived tariff policies, with the President enjoying the ability to easily reverse market sentiment (“Trump put”).
The MAG7 index dropped 5.2% Friday and sank 10.3% for the week, slightly less than the 10.4% “liberation day” weekly loss. Friday losses included Nvidia 14.0%, Tesla 7.6%, Meta 5.9%, and Amazon.com 4.1%. Losses for the week: Nvidia 22.9% (“largest loss since 2008”), Meta 12.6%, Tesla 11.4%, Alphabet 10.3%, Microsoft 7.5%, Amazon.com 7.4%, and Apple 3.9%. Nvidia CDS surged an unprecedented 22 Friday to a record 145 bps.
............. Importantly, fear was anything but confined to technology stocks. Ominously, the KBW Bank Index sank 9.5% this week, and the Broker/Dealer Index was hammered 10.4%. Friday losses included Morgan Stanley (5.3%), Citigroup (4.0%), Bank of America (3.2%) and JPMorgan (3.0%). For the week, Morgan Stanley dropped 13.1%, Goldman Sachs 13.7%, Citigroup 10.8%, and JPMorgan 10.3%. Apollo lost 10%, KKR 9.7%, Ares Management 8.4%, and Blackstone 7.8%. Robinhood sank 20.2%, and Interactive Brokers fell 13.0%.
The European STOXX 600 Bank Index dropped 7.3% this week. Japan’s TOPIX-Banks Index sank 11.0%.
Friday echoed the ‘risk-off’ days of years past. Ten-year Treasury yields dropped 10 bps to 4.94%, with yields down 21 bps for the week. .........................
Also indicative of aggressive speculating community deleveraging, the precious metals (and commodities) were under heavy liquidation. Gold slumped 8.9%, Silver sank 15.6%, Platinum fell 12.8%, and Copper dropped 9.9%. .................
In summary, market behavior is indicative of mounting fragilities for a highly indebted and leveraged world. ..................
I refer to “crisis dynamics.” The current environment is almost surreal. I don’t recall crisis dynamics ever feeling so palpable. ..................
History’s greatest Bubble was always going to end badly. It just went on for so long and to such outrageous extremes. Reckless monetary inflation; egregious speculation and leverage; epic malinvestment; deplorable wealth inequality and corruption. And we’ve reached the point where trust in government and our institutions is at such a low ebb. I can’t imagine a more precarious environment for managing financial, economic, social and geopolitical crises.
The unfolding crisis will present a major test for the “Fed put.” Speculative deleveraging will require another major inflation in the Federal Reserve balance sheet. They don’t expect to have to do it; they seemingly don’t have a plan to do it; but they will have no choice.
I come back to the same two questions: Will they move swiftly and in significant size to stabilize markets and restore confidence before things spiral out of control? Understandably, their current focus is to raise rates and get inflation under control. And they face the uncertainty of a highly unstable Treasury market ...........................
Thursday was supposed to be the day the AI trade got its groove back.
Nvidia had beaten and raised even without Jensen, Anthropic had priced the biggest IPO in history at $83, and hedge funds were sitting 11 points underweight tech, which is why Goldman's Rich Privorotsky told clients in his Thursday morning note that "today feels important" and that it "doesn't take much good news to squeeze." Instead, the S&P swung from a gain of more than 1% to a 2% loss, its sixth straight decline and longest losing streak since April 2024, and closed on its lows, back below its 200-day moving average. Only 38 S&P constituents went up. 464 went down.
So what went wrong? In his Friday morning note ("China Restrictions. 2018 Playbook. FCI Tight.", available to pro subs), Goldman's head of EMEA Delta One trading admits he got blindsided too:
Sharply lower from the US open yesterday. A lot of confusion about what exactly went wrong. For many, myself included, the catalysts were meant to be supportive… instead we got a pretty violent liquidation. I think objectively there are a few things, both macro and micro, that are challenging the market.
Those "few things" turn out to be: a far more aggressive China chip regime, a Fed that wants to "shock" people, and real yields sitting near their highs just as breakevens collapse. Put them together and Privorotsky sees a familiar movie, and readers who were around in late 2018 will remember how that one went before the happy ending. ......................
........... And the chip curbs only stack on top of the four things that were already crushing AI multiples, which Privorotsky lists as a) cost of capital (further challenged by the change of leadership at NVDA), b) DC restrictions, c) oversupply/efficiency unlocks for inference and d) lack of revenue to support capex. Point (a) is one we have been pounding the table on since October 2025, when AI quietly overtook the banks to become the largest debt sector in the market. Point (d), well... the market just valued a model lab at $2 trillion on its opening print, so maybe that one is still a work in progress.
............ What really bothers Privorotsky, however, is what didn't work:
Perhaps more disturbing was the failure of the defensive bid… utilities and staples sold off alongside everything else, despite some relief in Treasury yields. They still outperformed relatively, but the protection in absolute terms disappeared. What had been a rolling rotation increasingly looks like outright equity liquidation. ....................
In hindsight, the November/December 2018 parallels feel increasingly close. The Fed risks looking in the rear view mirror at inflation and the AI trade, effectively overtightening into a market already showing considerable stress. 2yr inflation breakevens have collapsed towards their lows, leaving 2yr real yields effectively near the relative highs. That's an extraordinary tightening in real policy conditions at precisely the point the equity market is signaling deteriorating growth and liquidity. .......................
As for a bottom, the contrarian case is there, for what it's worth: "the sheer depth of current pessimism and one sided de-grossing suggests the market may be approaching a tactical local low." ...............
Our take: the tactical low may well be close, and with funds this underweight a bounce could be violent. But the 2018 playbook had a crucial second act, the Powell pivot, and nobody at the Warsh Fed is reading from that page yet. Until someone does, every rally is probably one to sell into, and the bond market, which tends to figure these things out first, is already pricing the ending. Back in 2018 stocks didn't bottom until Christmas Eve. With December FOMC on the 9th and CPI not long after, we may not have to wait quite that long... or we may get a December that rhymes uncomfortably well.
***** QTR: The Best Is Yet To Come Undone
.................................... Because if there’s one thing I’ve been trying to communicate to readers over the last four months, it’s that the unraveling of the AI bubble was never going to be a single event, a single bad earnings report, or a single afternoon of panic selling. It was going to be a chain reaction, one in which deteriorating sentiment, increasingly questionable financing arrangements, political uncertainty, and scrutiny of the companies at the center of the boom could begin feeding on one another.
And as much as Wall Street has wanted to declare the worst over at virtually every opportunity over the last two weeks, my argument has consistently been that the next shoe probably hasn’t even dropped yet.
.............. More specifically, I was concerned about the enormous amount of capital being committed to AI infrastructure, the questionable economics underpinning some of that spending, and the extent to which global financial markets had become dependent on a handful of technology companies continuing to deliver perfection.
One of my central arguments was that investors had confused the extraordinary potential of artificial intelligence with the idea that every dollar being spent to develop it would necessarily generate an attractive return.
.............. I also focused specifically on Nvidia and the increasingly circular financing arrangements supporting the AI ecosystem. My concern was that the company had become so central to the market, and its chips so central to the investment plans of other companies, that any serious shock to Nvidia could produce consequences far beyond its own shareholders. Here’s what some of that web looks like, courtesy of Zero Hedge:
.................... The election didn’t need to cause the bubble to burst on its own. It simply needed to introduce another reason for investors to reassess the extraordinary assumptions embedded in stock prices.
............. The snowball isn’t sitting at the top of the hill anymore. It’s rolling downhill, gathering mass, and picking up speed.
.............. And when I look at the S&P 500 and Nasdaq right now, I don’t see anything that convinces me we’ve established a durable bottom. A violent decline isn’t a valuation argument, and a reflexive bounce isn’t evidence that the problems responsible for the decline have been resolved.
....................... If Nvidia’s problems cause lenders to reconsider the economics of the AI buildout, or if falling equity valuations make it harder for these companies to raise fresh capital, the consequences could become self-reinforcing.
............. I wouldn’t underestimate the political incentives to calm markets if the selling accelerates. Any reassuring statement might produce a sharp rally without necessarily resolving the underlying issues.
................... The potential interaction between falling stock portfolios, weakening consumer confidence, tighter credit, and an already strained housing market is something I think investors would be foolish to ignore. The real danger is not necessarily that every asset class collapses simultaneously. It’s that stress migrates from one part of the financial system to another, with each successive problem making the next one harder to absorb.
Retail buying dips like it's Dec of 2021…
Summary: We have increasing conviction that we’re in the early stages of a cyclical top in US equities. Importantly, tops are processes, not events. And we expect this one to play out over the next 3–6 months. That means expect a lot of chop and vol, failed breakouts and breakdowns, as the tape swings wildly in a broad sideways pattern before rolling over into a cyclical bear market sometime in the latter half of 2027.
This view is informed by the key indicators we track, a few of which we share below. Our TLS remains at -3. Breadth is weak. We’re not even close to capitulation according to our Trend Fragility measure. And Schwab’s STAX shows retail aggressively bought the dip last month, sending the index to its highest level since Dec ’21.
We also increasingly believe the US is going to significantly escalate the conflict with Iran before year-end. The DXY is breaking out, led by widening rate spreads. Our inflation leads are inflecting higher. Small caps are outperforming large caps. We cover all of that, plus a short commodity trade and more, below… .............
........................................... The more interesting part of Snider's report is where the money went. In Q3, funds cut TMT and added to cyclicals and defensives alike. Relative to the Russell 3000, net tilts rose most in Financials (+254bp), Consumer Discretionary (+237bp) and Industrials (+190bp), while funds also added to Staples (+154bp) and Utilities (+120bp). The Info Tech "underweight" grew by another 619bp to a record -2,267bp (much of that reflects tech's huge 35% weight in the index).
................................................... The pros dumped the AI trade for cyclicals. History says they may have just traded one crowded trade for another.
......................... Less clarity than ever. He said he cannot remember a period when he had less clarity on where markets will be in twelve months. For a man who built his career on picturing the world 18 to 24 months ahead, that is quite a confession.
Defence first. “We are trying to not lose money as opposed to make money.”
........................... 95 to 65
Thirty positions gone in a single quarter. When his conviction drops, he does not argue with it. He cuts.
................... The Soros lesson. “It’s not whether you’re right or wrong, it’s how much you make when you’re right and how much you lose when you’re wrong.”
................ Contrarianism is overrated. “I think contrarianism is overrated.” Soros taught him that the crowd is right most of the time, and that the skill lies in not getting caught in the rest. A crowded trade does not bother him if the thesis is right and the trend is with him. Being different and being right are two separate things.
............. Forgive the mistake. For about fifteen years he felt like an imposter, half-convinced his record was luck. His advice to good managers in a drawdown: “Just get over it and move on.”
AI credit default swaps have blown out, exposing the same mispricing of credit risk Wall Street ignored before 2008
......................... The AI buildout already has an equivalent price signal. It shows what the ABX showed in 2007: a fundamental mispricing of credit risk.
II. The Swaps Have Blown Out
Much of the AI build sits off-balance sheet, in leased data center campuses that rarely trade and have no public price. What does trade is the credit of the companies standing behind them.
OpenAI, Anthropic and the neoclouds could not finance campuses on their own credit, so the debt was made bankable by renting a stronger name: Oracle as lessee of record on the Stargate campuses, NVIDIA guaranteeing OpenAI’s lease in Ohio, Google backstopping Fluidstack’s leases for Anthropic, Broadcom standing behind Anthropic’s chip vehicles, and Meta writing residual value guarantees on its own off-balance-sheet campuses. Like the mortgage bond insurers, these companies rent their credit to obligations that could not earn it on their own.
Broadcom, NVIDIA and Google did it to sell more chips, Oracle to buy its way into the AI boom, and Meta to keep its own build off its balance sheet. This bit of financial engineering turned speculative frontier-lab demand into investment-grade debt. ...............................
.........................................
.................................... Over the past year NVIDIA became the AI build’s lender of last resort. When a neocloud could not borrow against its customers’ credit, or OpenAI could not get a landlord to build on its own credit, NVIDIA stood in. Its guarantees did for the AI build what the bond insurers’ wraps did for subprime: they made a weaker credit financeable by lending it a stronger one.
But every guarantee NVIDIA writes adds to its contingent liabilities and widens its swap, and every widening lowers the value of the guarantees it has already written. At 129bp, NVIDIA’s swap prices the bank as BBB-, on the edge of high yield; it touched 145bp when Jensen Huang’s resignation was announced. The bank of NVIDIA sits on the edge of high yield, and every guarantee it writes pushes it closer.
A lender of last resort is only as good as its own credit, and the bank of NVIDIA has lost much of its ability to manufacture demand by guaranteeing its customers’ leases.
............................ Paulson’s case had three parts: weak underwriting, excessive leverage and a fundamental mispricing of credit risk. The AI build has all three. The underwriting took the agency rating as the credit and the guarantor’s name as the guarantee. The leverage reached 85% to 95% of cost on buildings that have yet to be finished. And the mispricing is now visible: the AI build was financed as a single-A credit and trades as double-B.
Credit crises are built when credit risk is mispriced and assets are financed with more debt than their true risk can support. They break when the market reprices that risk, usually well before most of the defaults arrive.
The swap market has already repriced the AI build. The agencies and the equity market have only started to follow.
Mark to Swap.
Vid Fare:
Material Fare:
............ Gold's reputation as a refuge rests on long holding periods. Harvey said gold is highly reliable "over the extreme long term, which means centuries," but not over the short term.
Its annualized volatility, above 15%, is similar to the S&P 500's. Its correlation with stocks is close to zero over three- to five-year windows, but he cautioned this is an average and short periods can deviate sharply. His research shows that upside inflation surprises are usually associated with weak markets, and gold can fall in those episodes too.
Suki Cooper, Standard Chartered's global head of commodities research, said gold's role "has really evolved." Investors increasingly value it as a high-quality liquid asset they can tap in extreme scenarios. When sharp equity pullbacks trigger margin calls, she said, gold is a prime candidate to sell, especially after it had been scaling record highs until January.
Over longer holding periods, Cooper said, structural drivers remain in place. Those include record central-bank buying since 2022 and concerns about global debt, de-dollarization and currency debasement. ........
................................... Separately, UBS analyst Aditi Samajpati pointed out that sliding copper prices might present an attractive entry point into the metal trade given copper's underlying fundamentals are still rather poor .............
- More than 50 major U.S. oil and gas companies earned over $130 billion in profits between April and September 2026.
- Refiners and oil producers have benefited from elevated energy prices and disruptions to Middle Eastern supplies.
- American consumers have faced billions in additional fuel costs as the war continues to disrupt global energy markets
................. El Niño, the warm phase of the El Niño-Southern Oscillation (ENSO), is nearing historic strength. The Relative Oceanic Niño Index reached 2°C in August–October 2026, meeting the threshold for a very strong event. The October 2026 outlook of the U.S. National Oceanic and Atmospheric Administration (NOAA) put the chance of a very strong El Niño at 96 percent for November–January, and the chance that it becomes the strongest since 1950 at 83 percent for October–December. The World Bank Group’s food price forecasts assume an average effect for an episode of this strength. Worse-than-expected drought and heat in South and Southeast Asia and Australia could curtail rice, edible oil and sugar output, and heavy rains could damage crops in Brazil, East Africa, and the southern United States, pushing prices above forecasts.
Bubble Fare:
Marcus: Game over, man?
Sure looks like the bubble has finally popped
........... Ben Thompson of Stratechery just wrote an “obituary”, declaring that “The demise of Jensen Huang was the end of a great many things, including Nvidia’s long-term prospects, the Al bubble, and the exceptional way in which the U.S. viewed itself in the world.” .............................
One big issue is that the “bank” that was financing it all, Nvidia, is unlikely to continue to doing so. In the words of a Wall Street Journal article yesterday, “Nvidia’s Days as an Open Checkbook for the AI Economy Are Numbered: Interim CEO Colette Kress is expected to rein in deals after a stretch in which Nvidia aggressively financed computing deals and startup investments.” ....................
The other big issue, which I have harped on endlessly, is that the LLM companies still lack a technical moat and hence are still struggling to be profitable. That was true in 2023 when I first wrote about it, and remains true.
From the jump, that has always meant that Nvidia would not be able to sustain its phenomenal growth indefinitely, because you can’t sell chips endlessly if the companies buying those chips aren’t making profits from those chips. And mostly they weren’t. ...............
Anthropic, which dropped steadily after its opening yesterday, is set to drop nearly 10% more at the opening bell.
................ It is very much looking like the death spiral that this newsletter has long warned about has begun.
Jensen Huang effectively admitted Nvidia circular financing was hitting its limitis in seeking a $500 billion financing which has gone nowhere
Given that Jensen defenders have been depicting Jensen Huang as essential to Nvidia’s so-called circular financing, the “Bank of Nvidia” needs to be rebranded as the “Bank of Jensen.”
A new Wall Street Journal story provides much more detail on a scheme we covered briefly in our earlier Nvidia coverage. Nvidia had solicited Wall Street heavyweights like Blackstone, KKR and Brookfield to provide $500 billion in funding to Nvidia customers. An October Reuters report described how the program was not getting off the ground due to an unbridgeable gap between the level of financial support the money mavens wanted and what Nvidia provided.
The new Journal story has a lot more insider-y detail than the October Reuters account did. While that makes for a more entertaining read, the focus on the gossip (admittedly, reporter finds) has the effect of obscuring the bigger story architecture. As we pointed out earlier, Nvidia was suffering a big drop in market share that had started in 2023,1 hence the eagerness to sell GPUs to Chinese buyers. The article describes how Jensen was concerned about the fall in Nvidia’s multiple and wanted to gin up more financing from outside parties to goose it back up. That is a tacit admission that he was seeking more leverage to create the appearance of better performance. Leverage both amplifies profits and losses.
It may also be that Jensen realized he was approaching the limits of what he could do with his circular financing gimmicks. Even very deep tech pockets like Amazon were open about increasingly needing to use debt to fund AI capex.2 It is hardly a secret that across the industry, planned expenditures even once cash-rich tech giants exceed what they could fund out of their own cash flow and funds on hand, since AI is still a very long way from being a sensible economic proposition. .................
.................. As we have said before, a lot of plates are in the air and they look to be wobbling. But the flip side is that one of the oldest sayings in equity-land is that stocks climb a wall of worry. But with AI growth now unquestionably dependent on credit and that continuing to get more costly and scarce, AI growth seems destined to slow soon, and likely considerably too. Stay tuned.
For investors, the magic of the technology might be wearing off, with US stocks on track for one of the worst years in two decades
The AI fever in markets is breaking. US stocks are on track for one of their worst years in two decades and suddenly the whole topic is on big investors’ lists of “things to worry about” for 2027 rather than a reason to be endlessly cheerful.
It is hard to say what exactly has spoilt the mood, but you can trace it back from the memory chip crash out in Asian markets over the summer. Since then, a wide range of important groups of stakeholders have said “no”. Corporate bond investors are sick of being force-fed often questionable tech debt. The general public in the US is sick of supersized data centres in their neighbourhoods. Fund managers are sick of, let’s say, flaky corporate governance standards at Nvidia, which would matter less were it not for the fact that in October, this company hit a peak market capitalisation of about $5.7tn — more than twice the size of the entire FTSE 100 index in the UK.
And now this week’s spectacular stock market listing of AI lab Anthropic feels less like a moment of triumph (aside from the chosen few now counting their billions) and more like a last hurrah. Fizzy market debuts are a classic sign of a cyclical top in markets, and a $1.5tn market capitalisation at listing certainly fits the fizzy description. It is time to move into a new gear.
Just a quick thought on the day that the Anthropic public offering goes live …
There is a lot of chatter and excitement in the financial world surrounding artificial intelligence, which, if you are a true believer, is about to change absolutely everything.
Which is nonsense, but I will get to that in a moment.
For now, I want to focus only on the concerns lurking in the background about the possibility that the current froth might just be that … froth … and not much more. The question that keeps coming up concerns the chance that we are stuck somewhere in a financial bubble. Are the expectations and valuations out of whack with the likely reality? If so, when does that reality catch up with the dream world currently, apparently, befuddling what we are all told our “perfect markets”?
One contrast being drawn is that last really big bubble. Setting aside the catastrophe of the great financial crisis when our financiers drove us all over a cliff, the last real “bubble” was the dot.com version that burst at the end of the 1990s and into early 2000.
Back then, the bubble was obvious: stick prices were so far out of kilter with earnings that it took no analysis at all to suspect that something was up. Act one point, for instance, Cisco was trading at a P/E ratio of 221x. Absurd by any measure. But, by the standards of the all-knowing market of those heady days, somehow normal.
If we accept simple measures such as P/E ratios to give us insight into the possible presence of a bubble, we are driven to the conclusion today that, no, there is no bubble. Ratios are relatively benign for most of the top businesses today. I say relatively carefully. Some, like Tesla, are absurd, but Tesla is part of a Musk cult and not a properly valued business.
My perspective on this is somewhat different: we should not be looking at the P in P/E but the E.
Corporate earnings are at, or near, all-time highs, especially with respect to the economy as a whole:
................ As I said, what we need to examine is not the price in the P/E ratios, but the earnings. It is not unusual in history for businesses to be able to establish what appear to be impregnable sources of extraordinary profit (rents?) during a period of great technological transition. Having said that, just how many instances of “great transition” have we truly experienced during our era of large corporate dominance? Two or three? Maybe a couple more, but not a sufficient enough number to know whether what we are experiencing is a normal cycle that will be suppressed by subsequent competition and/or regulatory action. Are we due another great round of “trust busting”?
Those earnings suggest we are.
Back, briefly, to artificial intelligence.
Is it going to “change everything”?
In a word, no, but that’s only because we have gone through the great transition into the industrial world already. When everything, absolutely everything, was changed. .........................
A.I. Fare:
As Anthropic approached this week’s public stock offering, the company persisted in its famously paradoxical messaging—warning about the dangers of rapidly advancing artificial intelligence while telling us how rapidly its artificial intelligence is advancing.
A week ago, in an amendment to its S-1 filing with the SEC, Anthropic said it had now reached a threshold described in its safety policy as a “dramatic acceleration” of AI progress attributable to the growing role of AI itself in creating new models. In the original filing, it had warned that such acceleration could “cause the capabilities of our models to advance more rapidly than our ability to develop, validate, and implement corresponding safety, alignment, and security measures.” Also: Anthropic may not be able to identify “a significant acceleration in our rate of development until after it has occurred.”
Anthropic made this revision to its S-1 a week after METR, a highly regarded NGO that assesses the capabilities of AI models, issued a report about Anthropic that noticeably raised the freakout level within the AI safety community. The report found that Anthropic’s best models have over the past year advanced roughly twice as fast as during the preceding year and a half. This is the kind of steepening slope of progress that is said to precede “recursive self-improvement”—the point where the AI is designing and building new generations of itself, with humans relegated to a marginal role if any.
Paul Christiano, a former head of alignment at OpenAI who has made fundamental contributions to AI research, said on a podcast that the METR data was “exactly the data you would expect to see 6-12 months before producing superintelligence”—which means, he said, that there is a “reasonable chance” that “in six months, there’s a disaster that kills very large numbers of people, or where the government is overthrown.”
Those last five words may sound crazy, but this kind of talk is not uncommon in AI circles. In fact, in 2023 Anthropic CEO Dario Amodei, appearing on the same podcast, said that someday we’d have powerful agentic models and “if such a model wanted to wreak havoc and destroy humanity or whatever, I think we have basically no ability to stop it.” We “will reach the point where [this is] true, as we scale the models... That definitely seems the case.”
Well, now that we do have powerful agentic models, why does Amodei keep scaling them? Why doesn’t Anthropic slow down?
One answer can be found in a letter posted by Amodei two months ago, a letter that emphasized the value of slowing down—and that some people took as a promise that his company would slow down. Maybe these people didn’t read the fine print. More than a dozen paragraphs after declaring that “we must slow the pace at which we improve the capabilities of AI models,” Amodei added that American AI labs shouldn’t slow down enough to endanger “the lead that US companies have over authoritarian regimes, chiefly the Chinese Communist Party.”
Amodei’s concern with winning the AI race against China is shared by many in Silicon Valley and Washington DC. That’s one reason for the near-consensus that any significant slowdown, such as a pause in the large training runs that create new generations of AIs, must be “internationally coordinated.”
But for political and geopolitical reasons, this prospect seems remote—all the more so now that the scandal which put Nvidia CEO Jensen Huang out of a job has also put wind in the sails of US China hawks and raised the political cost of cooperative overtures toward Beijing by President Trump. ........................
The continued debate over whether Jensen Huang’s lover Alice Huang, the controlling shareholder of a Nvidia Singapore customer, Megaspeed, was a Chinese spy is a diversion from evidence ofserious abuses at Nvidia.1 The Wall Street Journal has just published important evidence of Nvidia’s and in particular Megaspeed’s chip laundering operations in Southeast Asia ..........
We will also start to look at the seemingly overwhelming topic of Nvidia’s circular financing. We suspect that most have been deterred by the MEGO-inducing yarn diagram of inter-connections among Nvidia customers and AI infrastructure players, such as data center funders. One way to pull on these strings is to look at underlying frauds. One is channel-stuffing, as in accelerating the recognition of sales revenue by pushing more product onto buyers than they want or need (as in beyond what one might consider to be prudent inventorying). That has the effect of moving sales that would otherwise occur in future quarters to the current one. While not legally classified as a stand-alone fraud, it has been successfully prosecuted, with Enron and Bristol-Myers as examples.
Another may be whether Nvidia has so heavily subsidized some of its nominal sales that they cannot be classified as true sales. ............
The Securities and Exchange Commission's probe of Nvidia is examining potential accounting issues at the chipmaker
........................
.............................................................. In the long term, there is the possibility that without Jensen at the helm, Nvidia will operate differently, perhaps pulling back on financing of customers and easing the relentless marketing that has made them the unstoppable force in the AI business. The long-term effect on value will depend on how much revenue growth rates and margins decline from the base case: .......................
.................................. As you wait for these uncertainties to resolve, you cannot sit still, since the market will likely continue to try to adjust based on expectations. I do think that if the market continues to mark down the stock, there will be a point at which you will find the stock to be a buy, notwithstanding unresolved uncertainties. For me, that value is about 30% below my estimated value of $4.8 trillion, yielding a breakeven value per share of around $140 per share, and I have a limit buy waiting at that price. If it happens, I will have Nvidia back in my portfolio, and if it does not, there are other stocks to buy!
........................... If you were using the Alibaba Qwen 3.8 Flash model, these 75 different 20-page papers would cost you 47 cents. To be clear, that’s 47 cents in total, not per paper.
This is the big problem for Anthropic and OpenAI. They have frontier models where they charge as much as $40 for a million output tokens. If they can sell large amounts of this AI, they could have a profitable business model. But the Chinese AI is likely to be good enough for the vast majority of users, and even for most uses by the most sophisticated users. The demand for this frontier AI is likely to be very limited.
................. For this reason, the best way to describe the plunge in the AI-related stocks over the last few weeks is: a good beginning.
Charts:
1:
(not just) for the ESG crowd:
Hansen: Horse Race Update
The UN Climate Change Conference 2026 (COP31) in Ankara, Turkey is the “worst Cop ever”, climate activists said in a packed final press conference towards the end of the UN climate summit. Delegates are locked in division over the words of the final statement. All mention of getting rid of fossil fuels has been removed from the draft communique. And the draft also fails to show how developing countries will receive financial help to deal with climate crisis. “It has all the hallmarks of a deal to let carbon capture through in exchange for an electrification target that does not even require renewables-based electricity or ensures access for the people who need it most. None of this is defensible.” ............
The Antalya draft’s targets need two years of world copper mine output, almost nine record years of building in five, and at least US$14.8 trillion, 49 times the US$300 billion a year it commits
On Monday, 16 November, the COP31 Presidency in Antalya circulated a draft that orders the world to build, by 2030, power plants, batteries and grids costing at least US$14.8 trillion. The order cannot be filled: it needs two years of the world’s copper mine output and almost nine record years of building in five. It passed with barely a ripple. No report on the draft has counted what it costs. ................................
......................................... The mines are not growing fast enough to meet it. World mine output rose from 19.2 million tonnes in 2015 to 23 million tonnes in 2024, about 0.4 million tonnes a year, so the machine’s extra 1.9 million tonnes equals almost five years of that growth, needed at once. Recycled copper supplied less than 17% of refined demand in 2024, down from 18% in 2015.
The IEA’s Global Critical Minerals Outlook 2026 projects that primary copper supply from existing and announced mines falls about 25% short of requirements in 2035 under today’s policy settings. That is narrower than the 30% it projected a year earlier, as new mines advance in the Democratic Republic of the Congo and Zambia, but the same report says that “the short- and medium-term supply outlook appears to have worsened considerably”. Disruptions at the Grasberg mine in Indonesia and at Kamoa-Kakula in the Congo tipped the refined copper market into deficit in 2025. ............................................
.......................... The world has the energy, 1% of supply, but not the metal to build the machine
Energy is counted in joules, and an exajoule (EJ) is a billion billion joules. The world supplied more than 600 EJ of energy in 2025, according to the Energy Institute, and end users, the factories, vehicles, homes and offices that burn fuel and draw power, consumed 453 EJ in 2024, according to the IEA.
Making the steel, iron, aluminium, copper, glass and concrete for the machine takes about 29 EJ. That is the embodied energy of each material, the energy used to mine, smelt and form it, taken per kilogram from the University of Bath’s Inventory of Carbon and Energy. Across five years, 29 EJ is about 1% of the energy the world supplies.
Once built, the machine delivers 12,188 TWh a year. A petajoule is a thousandth of an exajoule, and each TWh holds 3.6 petajoules, so 12,188 TWh is 43.9 EJ, or 9.7% of all the energy the world’s end users consumed in 2024. The world has the energy to build the machine. What it lacks is spare copper, silicon, graphite and lithium: every tonne the machine takes is a tonne the world’s mines already sell to another buyer. .................................
The era of 1.50°C is gone. Say hello to our new little friend, 2.0°C
If you haven’t been paying attention, the just released 2026 UN Emissions Gap Report stated that staying below the Paris Agreement limit of 1.5°C above the 1850-1900 pre-industrial baseline is now an impossibility. However, to the delight of techno-optimists worldwide, the report also provided a narrow window of opportunity to limit the anomaly to 1.7°C-1.8°C. The formula? Just reduce global greenhouse emissions 45% from 2025 levels by 2035, increase renewable electricity capacity by a factor of 3x over 2025 by 2035, reduce coal use 55% over 2025 levels by 2035, and suck 320 gigatons of CO₂ out of the atmosphere by 2100.
Easy-peasy, right?
Or totally and completely and unequivocally delusional. ...............
Private money and strategic rivalry are racing ahead of governance on who can alter the atmosphere and who bears the fallout
.................. Another very recent report, Carnegie Europe’s Planetary Power: Geoengineering and the New Geopolitics of Climate Intervention, points to an additional dimension. Its analysis of Big Tech identifies the growing involvement of technology companies, venture capital and philanthropic networks in climate intervention.
These actors may see geoengineering as a way to protect existing economic systems, influence future carbon markets and sustain technology-intensive growth. The danger is a technological “fix later” narrative that weakens incentives for emissions reduction ...................
SAI is contentious, geopolitically difficult and not without risk. The Highest Possible Ambition pathway ignores it, but what about everything else? What difference could we make if we did it all?
.................. Much of the conversation about deploying cooling technologies assumes it would come from stratospheric aerosol injection (SAI), which would spray tiny reflective particles high in the atmosphere to dim sunlight slightly and cool the planet. That’s for good reason, since it’s the one option that could help with nearly all of the risks of overshoot at once, and it could plausibly start within the next decade. But what if, for a variety of reasons, we decide not to do it? Maybe we can’t get over the hurdles around governance and decision-making, or the geopolitical risks look too high, or somebody starts it and then we decide to stop, or it has consequences that the models didn’t show. What then?
The stabilisation field does however have many more potential tools than SAI. We wanted to play out how effective those tools would be in managing overshoot without the global cooling benefits from SAI. In this article we compare two futures in 2050, one that follows the UN path alone, and one that decides against SAI but uses every other stabilisation and adaptation option it can, and look at what each would mean for five domains around the world. ......................................
We don’t know with enough confidence today that SAI is safe to deploy. But we can say with enough confidence that it’s too hot and that we need to find ways to bring the temperature down. If the world decides against SAI, it has to find another path, and that path is far thinner than most people assume. So all of the work in Future B is needed whether or not the world ever decides to use SAI. .............................
Earth history consists of long periods of boredom punctuated by short periods of terror. Changes do not take place gradually but as sporadic bursts, as a series of minor catastrophes(opens in new tab).
Something similar happens with ideas. Ways of understanding the world that persist for centuries or millennia are occasionally challenged and replaced. Copernicus showed that Earth revolves around the Sun, fundamentally changing how humans understood their place in the universe. James Hutton and Charles Lyell demonstrated that Earth was far older than the Bible suggested. Darwin’s theory of evolution was so revolutionary that some people still reject it.
These transformations are obvious in retrospect. They are much harder to recognize while they are happening. Established ideas rarely give way easily, even when evidence suggests they should.
War Fare:
Rickards: Hard Truths On Iran And Ukraine
Geopolitical Fare:
Steve Witkoff is pushing Kiev and the Europeans toward the settlement Moscow has long sought
..................... There is not much good to say about either Steve Witkoff or Jared Kushner, the New York landlords who proceed incompetently along as President Trump’s “special envoys.” While both seem to do well maximizing profit in the Manhattan real estate market, neither displays any fitness in matters to do with diplomacy or statecraft. I cannot tell what Kushner does other than speak for the Zionist state inside the Trump White House’s councils. As to Witkoff, he does seem to have one habit that now proves a virtue. Like the man for whom he fronts, he tends to agree with the last person who has spoken to him. This can be a good thing when the last person who has spoken to him has something useful to say.
So it now seems as negotiations resume in a multi-sided effort to reach an agreement that ends the Ukraine war and—the main thing—produces the enduring settlement between Russia and the West that Moscow has long sought. ..................................
Moscow’s stipulations are what President Putin and Sergei Lavrov, the Russian president’s FM, call “the root causes” of the Ukraine war. The New York Times, in a piece published at the weekend, referred to these as “hard-line demands that stand little chance of gaining acceptance by Ukraine and its European allies.” Yes, yes, Russia’s interests have long been marked down as “non-starters,” that clumsy English-ism favored during the Obama and Biden years. But The Times, as so often, is once again behind the curve. “Hard-line demands” is yesterday’s rhetoric. As of Abu Dhabi, Moscow’s legitimate security interests are hard line only in the Western press, and negotiations that address them at last appear to be inevitable. ................................
“When American-led talks to end the war in Ukraine resumed last week in Abu Dhabi,” The New York Times reported last Sunday, “the overriding question was whether Russia intended to give any ground after a nine-month lull in negotiations.”
Completely upside-down. Post–Abu Dhabi, the overriding question was and remains how long it will take for the Kiev regime and its European sponsors to accept that the cause is lost and proceed accordingly toward a settlement they have heretofore done everything in their power to prevent. .............................
I do not trust the perfidious Europeans any more than Moscow appears to do.
This will prove a telling moment for Europe, whether or not it assumes a place at the table when talks on Ukraine resume. One way or the other, what “Europe” means, what it has made of itself, and how its future is likely to turn out will become clear. I do not expect good to come of this, having hoped forlornly for many years that the Continent would make itself an independent pole of power with a voice others would want to hear. ........................
What gets done when is never certain when the Trumpster holds forth. But I take his strikingly specific mention of new territorial arrangements as another mirror—this one reflecting Steve Witkoff’s apparent progress toward a final settlement with the Russians that Ukraine and the Europeans will eventually have to accept.
................. At face value, these commitments mean nothing. The meeting itself, however, has already stoked concerns that even a partial ‘Yalta’ could reshape geopolitics, geoeconomics, and markets. Some are referring to it as “Yalta 2.0”: this time, with the US, Russia, and China at the helm of the evolving world order, but with Europe, or at least Ukraine, perhaps still on the chopping block. .............
Other Fare:
.................................. We need an ideology for the masses which respect the law of alliances. Again, no separate peace. No quitting till everyone gets what they need, and no stopping until the power of our enemies, the rich and powerful, is broken.
To win, we must be as relentless as our enemies, but unlike our enemies we must fight for everyone except our oppressors. Everyone should have a good life. Everyone should have a fair share of what society produces. At the end, we only have each other, and if we don’t have that, if we insist on fighting each other as fragmented identity groups, all we do is empower our greatest enemies: our own leaders.
Simplicius: Spate of Odd Cyber-Outages Strikes US as Nations Worldwide Warn of “Mystery” Threat
................ As a technical matter, both the Microsoft Outlook outage and the FAA’s nationwide ground stop were attributed to “cascading failures” following technological upgrades. I like that term because it captures the domino-like phenomenon where a single glitch in the system can quickly spiral into an incapacitating social and economic crisis. ..............
For the past few weeks, Western countries have begun playing up some unknown disaster or “event”, with UK in particular issuing breathless warnings to its citizens to stockpile goods and make emergency plans in case of some mysterious “disruptions” ............
................ Granted, we must not forget the eternal Hanlon’s Law: “Never attribute to malice that which is adequately explained by stupidity.”
And there is plenty of that—or at least incompetence and obsolescence—when it comes to the US power grid and general infrastructure. Shockingly, it was revealed that the FAA still runs floppy disks on much of its critical “computer” system networks: .............
Major outages affecting Microsoft Outlook and Air Traffic Control ground normal life to a halt, revealing the risk of single points of failure. Twiddling our thumbs until the next Chernobyl.
Radagast: If it happens, it was supposed to happen
Johnstone: Capitalism Is Easy If You Are Rich And Evil
Canada Fare:
A former central banker is playing a role usually reserved for royalty
Canadians for Justice and Peace in the Middle East (CJPME) has confirmed that the Tim Hortons app is sharing sensitive user data with Forter, an Israeli cyber-tech company founded by alumni of the Israeli military’s infamous Unit 8200. ............
From the controversial training of Ukrainian neo-Nazi formations to NATO’s military expansion in Latvia: Canadian military readiness, strategic priorities and institutional shortcomings


















