Market Overview October 14, 2025
Summary
The U.S. government shutdown is likely to drag on because ACA subsidies are a two-party deadlock with “absolutely no common ground,” while political polarization has severed the old mechanism by which centrist voters forced compromise. The speaker traces the deeper problem to the expansion of executive-order power and Congress’s loss of trust in the White House; his asymmetric call is that under policy pressure, Trump may become tougher domestically but relatively more willing to compromise abroad, while both parties’ bases may reward confrontation.
The 100% tariff threat on China looks more like a bargaining move ahead of the Seoul meeting, and the speaker expects the two sides to reach a deal eventually. China’s real choke point is not just roughly 70% of rare-earth production, but the refining stage, which is highly polluting and difficult to relocate in the short term; it will be hard to replace China within five years. He acknowledged misjudging Trump’s level of toughness in April and now sees both sides’ actions as “just putting their chips on the table.”
The shutdown has not pushed long-end rates low enough to make a steepener a “no-brainer”; around 4.2% remains the speaker’s reasonable year-end level. Government spending and Treasury supply will make it difficult for the long end to fall sustainably, regardless of how many times rates are cut; only if long-end rates return to their previous lows would he “go in no matter what” on a wider curve trade. For now, it is merely tradeable.
The trading opportunity in Q3 earnings comes from low expectations: S&P 500 earnings growth is expected at roughly 6%, below Q2’s 11%, leaving room for the AI complex to deliver concentrated beats. The market expects capex growth at large technology companies to be around 15%, while the actual pace was still above 30% at the time; on that basis, he sees a 500-basis-point beat, or roughly 5% above expectations, as entirely plausible, with substantial upside-surprise potential.
The global fiscal cycle is not over; inflation, not interest rates, is the real brake, and the risk window is more likely to open after 2026. After the November tax refunds, the expected positive fiscal feedback from BBB should emerge around the middle of next year’s first quarter and thereafter; Japan, China, and Germany are still expanding fiscally, while France’s political disputes will ultimately return to the question of how much to spend. If financial deregulation releases $4T to $5T of risk exposure, its liquidity impact could even exceed that of Fed rate cuts. Deregulation, AI development, and severe deflation in China are important offsets.
The short-term AI bubble peak is more likely to arrive when OpenAI formally goes To C or goes public, rather than when investors simply decide it has peaked on PE. Compute shortages have fully priced in expectations, but from GPT-2 through GPT-5 and Sora, demand is very high when models first launch and then decays; if consumer adoption falls short, the “left foot stepping on the right foot” loop of NVIDIA investing in OpenAI while OpenAI uses its compute could be exposed first. The U.S. 12-month PE is in the top 10% of its historical range, but ROE is also in the top 5% to 10%, so he does not view the current market as a bubble on the scale of 2000.
China’s September import rebound only shows that some investment demand has returned; it does not yet prove that domestic demand or deflation has turned around. Exports benefited from external orders, while a meaningful share of the increase in imports came from high-tech products and aircraft; consumption and CPI remain weak. The speaker’s conclusion is that “if ordinary people are not consuming, the government has to,” leaving the overall fundamental picture poor.
The crypto flash crash was a chain reaction of deleveraging after correlation instantly went to 1, not a single-cause story about Binance “misbehaving.” A basket of 50 assets, or market-making across 100 tokens at once, can fall together in an extreme market; some market makers’ unified accounts triggered stops and liquidated everything, potentially setting off a cascade of leveraged-staking accounts that borrowed USDT and bought USDe. USDe briefly traded around 0.66 by his recollection, while Ethereum’s theoretical TPS of just 60, fee bidding, and a brief suspension of withdrawals amplified the squeeze. “The bridge was too crowded; everyone could not get through,” much like negative oil prices: extreme prints in forced trading are not fundamental prices.
Deep dive
1. The Shutdown Stalemate Will Drag On, but Treasuries Have Not Reached No-Brainer Steepener Levels
ACA subsidies are the immediate deadlock behind this shutdown: Democrats view the Affordable Care Act as a governing legacy and want an extension of the subsidies written into the bill; Republicans and the Trump camp see them as a failed system that has been abused. The speaker’s view is that the two sides have “absolutely no common ground” here, with almost nothing mutually exchangeable.
He partly traces the institutional roots to the Supreme Court, which in several rulings has supported presidents bypassing Congress through executive orders, making the U.S. increasingly resemble a “government of emergency,” with presidential power approaching the level seen under Roosevelt. The result is not only greater room for Trump to act, but a continued collapse of trust between Congress and the White House.
The 2013 and 2019 shutdowns were ultimately constrained by public pressure: the side that initiated the standoff was blamed and forced to compromise. This time, left- and right-wing voters first support their own side, while “the centrist voters actually blame both sides.” When creating a deadlock can even increase support among the base, public opinion no longer provides a clear exit ramp.
Markets have not repeated the classic safe-haven script: the shutdown has raised the probability of further rate cuts but has not driven TLT and long-end yields down enough. Given year-end government spending and Treasury supply, the speaker sees “4.2% as the long-end rate for me”; only if the long end returns to its previous lows would a steepener be a “no-brainer,” so current positioning should be more restrained.
2. Rare Earths Turn the China Escalation into Leverage Ahead of the Seoul Meeting
The sequence this time was China imposing extraterritorial controls on rare earths, corresponding to a possible new round of semiconductor restrictions from the U.S. House, followed by Trump’s threat of an additional 100% tariff. The speaker does not deny that both sides have the ability to choke the other off, but sees the moves as escalation ahead of negotiations.
China’s advantage is not limited to roughly 70% of rare-earth production. The key is refining: it is a highly polluting stage, and the refining portion of the supply chain is heavily concentrated in China. U.S. environmental constraints make it difficult to move the industry back directly, while relocating it elsewhere would require years of cultivation. His time horizon is that “China will still be very difficult to replace within five years.”
His revision is worth preserving: “When the tariff issue happened in April, we had misjudged Trump’s policy to some extent.” At the time, he expected Trump to be tougher and to cause a larger economic shock; his current framework is that Trump may be tougher domestically but relatively more willing to compromise abroad. Ahead of the Seoul meeting, both sides are “just putting their chips on the table.”
3. Low Expectations Leave Broad Room for AI Earnings Beats in Q3
The speaker’s baseline is that S&P 500 earnings growth is expected at roughly 6% in the third quarter, well below Q2’s 11%. Expectations are not demanding, which means earnings season is more likely to produce a dense run of beats, with the AI sector especially positioned to deliver upside surprises.
The clearest mismatch is capex: the market expects investment growth at large technology companies of around 15%, but the speaker had not seen the actual trend slow to that level; the relevant growth rate was still above 30%. A “500-basis-point beat” or a “5% beat” is therefore possible, leaving substantial upside-surprise potential in the near term.
But an earnings beat answers only one quarter, not when the cycle ends. The speaker breaks the current market into two wheels: the global fiscal supercycle and the AI compute-demand supercycle. The former ends with inflation; the latter’s short-term endpoint will not be clear until OpenAI’s consumer demand is tested in the market.
4. Global Fiscal Policy Is a “Fast Bull Market,” and Deregulation Could Amplify Liquidity
The positive fiscal feedback from the U.S. BBB bill is expected to start appearing around the middle of next year’s first quarter and thereafter: the shutdown and budget standoff will eventually be resolved, while funds will need some time to enter the real economy after tax refunds begin in November. Short-term political noise has therefore not changed the direction of the fiscal impulse in early next year.
This is not just about the U.S. Japan’s new prime minister is inclined to continue Abenomics, China will keep spending, subsidizing consumption, repairing roads, and building infrastructure, Germany has a 500-billion plan, and France’s political debate will ultimately come down to how much money to spend. The speaker’s overall judgment is that “the global fiscal supercycle is actually not very stoppable.”
The fiscal cycle differs from the monetary cycle after 2008: monetary liquidity transmits slowly, while fiscal spending goes directly into goods, consumption, and services. The pandemic already demonstrated its inflationary impact. “This is not a slow bull market; it is a fast bull market.” His baseline remains that “2026 is fine,” but inflation may force the cycle to stop after 2026; deregulation, AI development, and Chinese deflation are several offsets.
Financial deregulation is an easily overlooked amplifier. If constraints on Tier One Capital relative to total exposure under Basel III are relaxed, and Treasuries are removed from the risk-exposure denominator, banks could expand their exposure by $4T to $5T without adding capital and use 10x to 100x leverage in Treasury arbitrage. In the speaker’s view, that “could have a bigger effect than your Fed cutting rates.”
5. OpenAI To C Could Mark the Short-Term Bubble Peak, but This Is Not a 2000 Replay
The speaker identifies the potential inflection point as OpenAI formally going To C or formally going public. Repeated claims that “not enough tokens means not enough compute” have caused the market to fully price in the potential of Sora and other consumer applications; once they are actually opened up, whether the products can meet expectations will shift from imagination to a testable outcome.
From GPT-2, GPT-3, GPT-4, and GPT-5 through Sora, the speaker has observed that compute demand is very high when a new model launches and everyone uses it, but gradually decays after some time. Consumer applications will struggle to immediately match the market expectations formed during a period of rapid growth, so insufficiently fast adoption or problems with the models could mark “the short-term top of the AI bubble.”
The skepticism is concentrated on user applications: NVIDIA invests in OpenAI, and OpenAI then purchases its compute, which does carry the suspicion of “the left foot stepping on the right foot.” Digital media offers a counterexample, however: Meta and Google are using AI to improve ad revenue, margins, and gross margin, with returns on investment “visible to the naked eye,” showing that there are already clear use cases.
The U.S. 12-month PE is around the 90th percentile of the past 20 years, but ROE is also in the top 5% to 10%. A large share of the gains in U.S. technology stocks has come from earnings growth, which the speaker attributes mainly to higher ROE or returns on investment. He therefore believes the current bubble has not reached 2000 levels; the major bubble has not peaked, but the short cycle could still peak first.
6. China’s Import Rebound Signals Investment Recovery, but Low PE Is Offset by Low ROE
September imports and exports both increased, but they send different signals: stronger exports can be explained by continued orders from external economies such as the U.S.; a meaningful share of the increase in imports came from high-tech products and aircraft, showing that some investment demand has returned but not proving a recovery in household consumption. The speaker cites building compute centers and buying aircraft as examples of recovering investment demand.
The speaker’s conclusion remains cautious: consumption has not recovered, and CPI does not show that deflation has ended. Economic improvement is driven mainly by government investment. “If ordinary people are not consuming, the government has to,” so road repairs, infrastructure, aggregate social financing, and consumption subsidies will continue. The fundamentals have not truly escaped their difficulties.
Valuation cannot be separated from returns on capital. Chinese companies trade at around 13x PE, which looks cheap, but ROE is in the worst 10% of its historical range; high U.S. PE corresponds to ROE in the top historical tier. Splitting stock-price contributions between earnings and multiples, the global market is roughly 50/50, while emerging markets, including China, have a larger contribution from bubble expansion. Low PE alone is not enough.
7. The Crypto Flash Crash Was Deleveraging with Correlation at 1, Amplified by Ethereum Congestion
After Trump announced the tariff measures, U.S. equities plunged and Bitcoin itself fell only modestly, but large numbers of altcoins suffered “liquidation-style declines.” The speaker rejects attributing the abnormal prices directly to Binance: negative oil prices were not caused by the CME “doing evil,” but by leverage being unwound simultaneously, when “the bridge was too crowded and everyone could not get through.”
Market makers in small coins typically buy into declines and sell into rallies, using position limits instead of stop-losses to prevent inventory losses from expanding quadratically. A market maker may hold a basket of 50 assets or make markets in 100 tokens at once; these strategies depend on low or negative correlation between assets. In an extreme market, correlation becomes 1, and even if every token falls only 5% at the same time, the portfolio may become unmanageable.
After some market makers were forced to stop out, their unified accounts triggered one-click liquidation, potentially selling USDe as well. At the same time, circular-staking accounts that borrowed USDT and bought USDe for yield continued to be liquidated. Accounts using VIP loans did not liquidate in the same way, while those using margin loans had to accept stop-losses.
Binance used on-exchange trades as its reference price. USDe, according to the speaker’s recollection, was briefly marked down to around 0.66, causing many unified accounts holding USDe to liquidate together. Subsequent prices cannot be treated as rational prices; they were extreme prints caused by sudden deleveraging.
The final layer of amplification came from Ethereum: its theoretical TPS is only 60, and concentrated liquidations created on-chain congestion and fee bidding. Transfer costs briefly became uneconomic, and Binance was also briefly unable to process withdrawals and transfers. The abnormal prices then rebounded, showing that this was first and foremost an event-driven deleveraging episode, not the rational instantaneous repricing of altcoins to zero.