Market Overview: January 20, 2026

Summary

• 2026’s base case: a bull market year for U.S. equities, but a “volatile bull market.” This is a year of U.S. decoupling growth—the U.S. outperforms the rest of the world, unlike the 2025 setup of global recoupling; China is less certain: “SAFE has been selling stocks these past few days—that’s the market they control.” The 3 macro themes are unchanged: disinflation, soft employment, and moderate growth. The Fed will skew dovish, and after Powell steps down, “whether it’s Kevin or whoever takes over, I think we’ll get one more rate cut”—not unlimited easing, just one additional cut.
• Trump is a dotted line: “Everything he does is theater.” Immigration is no longer a viable political achievement after the Minnesota shooting incident, reportedly involving ICE; all he can tout now is reshaping trade and reshaping industries. Drug-price and credit-card-rate pressure amount to industries giving him “a face-saving concession.” He has no Congress: tariffs rely on emergency legislation, the Supreme Court’s IEP ruling could come “any minute,” and Trump is more likely to lose on Polymarket. The second fiscal package could be filibustered past the midterms, making passage unlikely. This year’s volatility is Trump drama, with very little substantive impact.
• The K-shape is a danger signal; I expect the scissors spread to close. Nvidia has the most certain earnings story—growth of roughly 70-80% this year, with the rate still accelerating—but the worst narrative and a falling stock price. The winners are optical modules and memory, where the narrative is strong but earnings have yet to show up. There are only 2 ways to resolve this: everything gets cut together, with Nvidia seeing only a small decline; or the right stocks rise and the wrong ones fall—chasing optical modules and memory “can cut you in half.” We will probably see that this year, even if shortages drive a recovery by year-end. The simplest test: can MU deliver 75-80% growth this year? “Your stock price has already risen more than your earnings have.”
• Compute is the “eternal bottleneck for the next several years”: chronic shortages. In semiconductors, “believe anyone who says there’s a shortage—believe it early.” OpenAI’s compute rose from 0.2 GW in 2023 to 1.9 GW in 2025, a 13x increase, while revenue grew 10x—efficiency is improving. The end buyers, OpenAI and Anthropic, are not short of capital; Anthropic appears to have just raised at a $350B valuation. TSMC is the core constraint. From TSMC to possibly Nvidia—the original quote was “Avidian”—the entire chain is a seller’s market: whatever gets produced gets sold.
• AI productivity is showing up in the numbers; this is not “semiconductors hyping themselves.” JPMorgan, BofA, Citi, and Wells Fargo each invest 10% of revenue in technology annually. Excluding JPMorgan, the major banks have reduced headcount by 5% in a year, and AI-driven operating leverage accounts for a large part of that improvement. His own investment-banking experience: “Ninety-nine percent of what I used to do can now be generated with one click.” Efficiency gains in high-paying sectors such as law, accounting, and consulting are “just beginning to flower”—buy companies using AI to improve their own productivity.
• Tail hedge: use a 10-year Treasury swaption as a lottery ticket. The hedge is not for a 5% decline, but for a 15-20% market drawdown. At 4.2%, the 10-year yield looks relatively safe and close to a ceiling: changes to bank capital requirements could unlock $4-5T of balance-sheet expansion, passage of the second fiscal package is uncertain, and a government shutdown at the end of January would all push rates lower. The key is that vol is very low—buy an out-of-the-money swaption, where a 20-basis-point decline in rates could generate roughly 50x returns, using 0.5%-1% of the portfolio to “buy a lottery ticket.” He would be leaning into long duration.
• Positioning discipline: do not go to cash. “You won’t be able to stay in cash; by year-end it will have risen so much you’ll go crazy,” then chase the highs and get caught in the correction. Hold durable, profitable names—Nvidia is delivering “real growth in the high-80% range.” On a pullback, do not buy optical modules: “The narrative blows wherever the wind takes it.” Buy CPUs where the trend has yet to emerge: AMD and Intel. Intel also offers Trump’s only tangible political achievement to tout. The dollar looks to be forming a base; my personal view is that this year and next will bring a return to “U.S. exceptionalism.”

Deep dive

1. The Macro Setup Is Unchanged: Disinflation + Weak Employment + Moderate Growth = Bull-Market Backdrop

• He opened with 4 questions: What state is the market in, what should you own, what should you buy on a pullback, and how do you hedge tail risk in a volatile bull market? The macro answer is “no change.” Last week’s CPI confirmed the 3 U.S. themes: disinflation, soft employment, and moderate growth. “Inflation risk is declining, employment is relatively sluggish, but economic growth is still okay”—a bull-market setup.
• The Fed can only get softer. After Powell steps down and Trump appoints his successor, “whether the final compromise is Kevin or whoever takes over, I think we’ll get one more rate cut”—with a clear boundary: not Trump’s version of cutting rates to extremely low levels, just one additional cut.
• The global call is a year of U.S. decoupling growth: the U.S. outperforms the rest of the world, unlike the 2025 setup in which markets recoupled with the dollar and EM outperformed the U.S. China is uncertain because “SAFE has been selling stocks these past few days—that’s the market they control.” But with the midterms ahead, this is also “a year of volatile gains, a volatile bull market.”

2. Trump Is a Dotted Line: Volatility Comes From Drama, Not Fundamentals

• He divides the macro picture into a solid line—economic fundamentals—and a dotted line—what the president wants to do. The conclusion on the dotted line is unambiguous: “Everything he does is theater.” On drug prices, “everyone gives you a concession… everyone plays along with the show”; on pressuring industries, “everyone knows you don’t have the ability to legislate.” The avatar is therefore drama: this year’s volatility is Trump drama.
• Immigration, one of the 3 pillars of his agenda, has become difficult to sell after the Minnesota shooting incident, reportedly involving ICE: “It’s hard to tout that as a political achievement anymore.” That leaves reshaping trade relations and reshaping industries. The only industry with a tangible achievement to show is Intel: “He invested money, but actually did nothing… Intel was already about to take off, then he bought the dip and claimed he made billions of dollars for the American people.”
• The 2 “strange moves” last week were the software blowout on Thursday and Friday, and volatility in financials despite clean quarterly reports and solid earnings. The financial-sector volatility was driven by Trump’s pressure campaign: he said he wanted credit-card rates capped at 10%. “How could credit cards possibly be 10%?… At 10%, you can’t even cover your costs.”
• The hard constraints are Congress and the Supreme Court. Tariffs rely on emergency legislation; the IEP ruling could come “any minute,” and Trump is more likely to lose on Polymarket. His sense is that the longer the process drags on, the more of the tariffs could survive. The second fiscal package and affordability payments both require Congress and could be filibustered past the midterms, while the odds of Republicans taking Congress in the midterms are “very, very low.” On tariffs against Greenland or Europe, Goldman estimates a 10-20 basis-point impact—“no one in the market is calculating these things right now” because everyone knows “this guy could change his mind tomorrow; he may just be trading the stock.”

3. The Micro Theme: Compute Is Perpetually Scarce, and AI Is New Productivity Replacing Old Productivity

• Compute—the balance between token demand and supply—is “an eternal contradiction for the next several years: chronic shortages.” His rule of thumb is simple: “When anyone in semiconductors says there’s a shortage, believe it… believe it early.” The key point in OpenAI’s presentation was that compute rose from 0.2 GW in 2023 to 1.9 GW in 2025, a 13x increase, while revenue grew 10x. OpenAI and Anthropic can both raise capital—Anthropic appears to have just raised at a $350B valuation—and efficiency is improving.
• His answer to the AI-bubble argument is that Wall Street keeps asking where the railroad-scale new productivity is, “but that’s not what it is. AI is fundamentally new productivity replacing old productivity. It doesn’t necessarily create something new; it eliminates what you used to do.” Last week’s software selloff reflected that fear. He cited CIM, as stated, as “a dead end”—it will either be killed by Microsoft, directly by OpenAI, or by Microsoft using OpenAI’s interface—but that process takes time. Software earnings should remain reliable over the next 1-2 years.
• The core supply-side constraint is TSMC: “However many chips TSMC can produce… you can assume they will all be sold.” Allocation goes mainly to Nvidia, with Broadcom possibly second—the original audio was “阿巴狗”—despite a valuation far above Nvidia’s. From TSMC to possibly Nvidia—the original quote was “Avidian”—the entire chain is a seller’s market.

4. The K-Shape Is Unsustainable: The Most Certain Earnings Are Falling While Narrative Stocks Rise

• The market has split in 2: story versus earnings. “The companies with the most certain earnings and the highest certainty of real growth have the worst narratives, rise the least, or are even falling. The only representative is Nvidia.” Last year, growth was 60% while the stock rose only 40%; this year, growth is expected at 70-80% and is still accelerating, yet the stock is falling. Microsoft sits in the middle: relatively certain earnings and a PE in the low-20s, “but it has no story.” The winners are those with “great narratives whose earnings you have not yet seen”—optical modules and memory.
• This scissors spread cannot persist in semiconductors because optical modules, memory, and CPUs are all suppliers to Nvidia’s ecosystem: “When the skin is gone, what can the fur cling to?” If Nvidia’s earnings falter, total compute demand is in trouble. TSMC capex is 40, 40, 40, while token demand is 3x, 3x, 3x; once the overall direction changes, memory will not hold up either.
• There are only 2 exits: everything gets marked down together, with valuations recovering where they should—Nvidia’s decline would be limited because earnings are so certain—or the scissors spread closes, with the right stocks rising and the wrong ones falling. Chasing optical modules and memory “can cut you in half… you are very likely to see these 2 industries cut in half this year.” But that is not necessarily fatal; by year-end they could still recover because the shortage is real. The test is MU: “Can it deliver 80% growth this year?… Your stock price has already risen more than your earnings have.”

5. The Evidence for AI Productivity: Major Banks Except JPM Cut Headcount 5% in a Year

• In response to the charge that “semiconductor companies are just hyping themselves,” he pointed to end-user economics. JPMorgan, BofA, Citibank, and Wells Fargo each spend 10% of revenue investing in technology, with AI accounting for an important share. Excluding JPMorgan, which is expanding, several major banks reduced headcount by 5% in a year. That is a major source of the operating leverage investors had been rewarding in banks.
• Firsthand testimony carries the most weight: “We used to work 100 hours every week writing PPTs… 99% of what I did in investment banking in my first and second years can now be generated with one click.” Analysts used to write meeting notes and build presentations; “now a computer generates them in 1 second—or 1 minute if 1 second isn’t enough.”
• The runway is only beginning. The top 4 banks invest 10x as much in technology as the bottom 2. In law firms, accounting firms, consulting, and other high-paying U.S. sectors, the efficiency gains from replacement “are only just beginning to flower.” That leads directly to the stock-picking framework: first look for companies using AI to improve their own efficiency. Banks are rising because they are doing exactly that.

6. Hedging and Positioning: Swaptions as Lottery Tickets, CPUs on Pullbacks, No Cash

• The tail-risk hedge is for a “two-standard-deviations-out” event: a 15-20% market correction. “I don’t know why it would happen; I don’t see a macro reason for it right now, and positioning isn’t crowded, but you still worry that it might happen.” The instrument is the 10-year Treasury, with 4.2% “basically a ceiling.” The logic has reversed: in 2024-2025, Treasuries were “like Sisyphus pushing the rock uphill”—whatever came down would rise back up—because supply was enormous. This year, liquidity is abundant: changes to bank capital requirements could unlock $4-5T of balance-sheet expansion, the second fiscal package may not pass, and a possible government shutdown at the end of January would all push rates lower.
• The real sweet spot is volatility. Low vol makes swaptions cheap: “If Treasuries move down 20 basis points, you could probably make around 50x.” Use 0.5%-1% of the portfolio to “buy a lottery ticket,” producing multiples of the stake during a sharp market selloff. The historical footnote is Quantum Fund in 2000: after losing heavily in the tech bubble, it recovered the money through long Treasuries. If he had to put on a position, “I would be leaning into long duration in Treasuries.”
• The positioning discipline is blunt: “I strongly advise against going to cash… you won’t be able to stay in cash; by year-end it will have risen so much you’ll go crazy,” then chase at the highs and get caught in the correction. Hold “durable, profitable companies.” Nvidia’s earnings are delivering “real growth in the high-80% range”; after last year’s mismatch—60% growth versus a 40% stock gain—there is substantial room for valuation repair.
• What to buy on a pullback: not optical modules—“the narrative blows wherever the wind takes it”—but CPUs where the trend has yet to emerge: AMD and Intel. He stressed that this was his personal view. Intel also has Trump’s only tangible political achievement to tout. On the dollar, the index fell 7% last year, but he sees a basing process this year with limited room for further downside. The dollar is “not entirely a petrodollar; it is a U.S. equity dollar.” His personal view is that this year and next, at least through the first half, will bring a return to “U.S. exceptionalism.”