Market Overview | May 13, 2025
Summary
- The core of this rally is not a sudden reacceleration in growth, but the 90-day US-China “ceasefire” stripping out the left-tail risk premium attached to the White House “not caring about markets.” The speaker acknowledges that the agreement was far better than expected, but has moved only to “cautious optimism”: policy can reverse quickly, and the market could collapse again after 90 days, so this “is not a market to FOMO into.”
- The speaker’s fundamental view of the US economy is unchanged: growth is continuing to slow, with the probability of recession around 50%, perhaps even above 50%, well above Goldman’s revised-down estimate of 35%. Soft data has weakened while hard data has yet to confirm it, and Powell will “meet change with no change”; a slowing economy combined with a Fed that does not take decisive action in time will ultimately weigh on corporate earnings growth.
- The near-term rally is being driven by three forces—tariff negotiations, earnings and buybacks, and extremely light systematic positioning—not proof that long-term valuations are safe. CTA and vol-control positions fell from roughly $250B last year to $35B at the trough and now stand at around $90B; a falling VIX will prompt these strategies to buy back in, while results and buybacks from Microsoft, Amazon, Google, Meta, and Apple have further amplified the rebound.
- The S&P 500 is now near the speaker’s upper range, implying roughly 21x P/E and 10% earnings growth—clearly out of line with his fundamental view. The market still expects around 6% earnings growth in 2025, versus Goldman’s roughly 3%; if growth falls to 5% or below, EPS is about $255, making even 22x expensive despite the technology weighting, while recession valuations could approach 15x. At current levels, the trade is “a bit of an awkward one,” with downside risk greater than upside.
- Even if reciprocal tariffs ultimately come down uniformly to 10%, the speaker estimates they could still cost roughly 1% of GDP and add roughly 1% to inflation. The negative real-economy impact should lag by around six months and may emerge by the end of Q3 or in Q4; even if tax cuts pass in full, the money may not enter the economy until November, pushing the macro effects to between the end of Q1 and mid-2026 and creating a policy limbo.
- The index is not worth chasing, but beaten-down AI, ASIC, and semiconductor names offer better single-stock odds; the speaker owns Marvell and Intel. In complete ASIC solutions, he cited Broadcom and Marvell; Broadcom gets more market credit because of its long track record serving Google TPU, while Marvell offers more upside torque because its AI transition is newer and it faces NDA restrictions and hedge-fund positioning. Discussing Microsoft orders and Maya 3, the speaker said it was “very certain” that the companies involved would earn about $3 per share this year and “also very certain” that earnings could grow 70%-80% next year. Potential improvement at Intel’s 18A is tied to 陈立武’s ability to connect Cadence, TSMC, and external tapeout teams. Banks, investment banks, and names such as Goldman Sachs and Morgan Stanley, by contrast, are more direct bets on how much regulation Trump will loosen.
- With the 10-year Treasury yield around 4.5%, room for another leg higher is limited; it is more likely to end the year in the 4.2%-4.5% range, making a Treasury short “not very meaningful.” At the portfolio level, the most actionable conclusion is not to put on a large short, but to follow the Fed: “don’t touch existing positions,” and wait for tariffs, hard data, and earnings to play out.
Deep dive
1. The tariff ceasefire removed tail panic, but not policy risk
The speaker attributed the selloff since February to a reversal in the market’s policy perception: investors initially assumed Trump was pro-business, then concluded that the White House seemingly “didn’t care” about asset prices and began pricing the left-tail risk of the government “messing things up.” The rally of the past 2-3 weeks has essentially removed that extreme-risk premium.
The US-China agreement was more constructive than the short-term compromise he had envisioned in April, but it did not make him outright bullish: “I don’t think this is a market you can FOMO into right now.” The substance of the negotiations after 90 days remains unknown, and policy can change fast enough to reverse the current pricing once again.
Under the speaker’s comparison, China and the US each impose 10% on the other, while much of the rest remains fluid: of the US’s original 34%, 24% depended on the 90-day negotiations, and the 20% fentanyl tariff was more a bargaining chip; the UK received a 10% tariff and an auto quota, while Taiwan accepted roughly 32%. The market may be pricing a scenario in which “10% for everyone” ends the dispute, but that is already the more pessimistic case embedded in the early-year outlook, and other countries may not continue to yield.
2. The US can still slip into recession; the Fed will wait for hard data to force its hand
Even with the US-China “ceasefire,” the speaker still puts the probability of a US recession at “about 50-50, maybe even above 50%,” clearly higher than Goldman’s latest 35%. Morgan Stanley has begun discussing new highs for the S&P, but he sees betting real money on the political direction as a different exercise from discussing policy while working in venture capital.
The tension is that soft data has already started to fall, while the data over the past week has held up and hard data will take longer to weaken. The speaker calls this state limbo: the economy may slow, but the Fed “won’t move” until hard data on inflation or recession appears, and at most will act very cautiously.
He has more confidence in his read on Powell than on the White House: there is no prospect of the Fed pre-coordinating with the Treasury Secretary to launch a joint, large-scale liquidity injection. The Fed will simply “meet change with no change.” That leaves the market exposed to earnings-revision risk between the initial soft-data deterioration and confirmation in the hard data.
3. Light positioning, buybacks, and better-than-expected earnings amplified the rally
S&P earnings grew roughly 12% in Q1, beating expectations, but the growth rate was already down from 2024. After DeepSeek emerged, the market questioned AI compute demand, cloud profitability, and the ability to keep funding capex; earnings and buybacks from Microsoft, Amazon, Google, Meta, and Apple provided support for risk appetite against that backdrop.
Positioning created another mechanical transmission channel: US CTA, vol-control, and other systematic positions were typically around $250B last year, fell to a low of $35B, and now stand at roughly $90B. Systematic money had “basically sold out,” and as the VIX falls and the market rises, it will gradually buy back in.
The short-term rally is therefore not mysterious: tariff talk improved sentiment, earnings and buybacks provided support, and light positioning created demand from investors chasing the move. But the speaker stressed that these forces pushed the index rapidly to his upper range without changing the fact that economic growth ultimately determines earnings.
4. Current valuation requires near-perfect earnings; risk/reward has tilted downward
The market expects roughly 6% S&P earnings growth in 2025, versus Goldman’s roughly 3%; the speaker thinks the number should be revised down to 5% or below. At around 5% growth, EPS would be approximately $255, while the current price pays about 21x P/E and implies nearly 10% earnings growth.
A higher technology weighting does mean historical valuation multiples cannot be applied mechanically. But earnings growth at Microsoft, Amazon, Google, and similar companies is also slowing. As ASICs take a larger share of capex, NVIDIA’s growth in 2025 and even 2026 should continue to decelerate, making extremely high valuations difficult to sustain indefinitely.
His valuation yardstick is that a soft landing or recession typically falls in the 15x-20x range. Technology can lift the upper end, but 22x is already very high and amounts to a mild bubble in the US market; in a recession, valuations could move toward 15x. “At current prices, I think it’s actually a bit of an awkward trade.”
This is not a recommendation to put on a large short, because policy could quickly become “another story.” But even if the US and China reach a complete agreement, chasing the rally means taking on more risk upfront. The speaker’s overall view is that “the S&P’s risk is leaning to the downside.”
5. A 10% tariff is still a real shock; tax cuts cannot fill the gap in time
The speaker estimates that a 10% reciprocal tariff alone could cost roughly 1% of GDP and add roughly 1% to inflation. Without reciprocal tariffs, the estimated economic impact would be about 0.5%, with an inflation impact of 0.4%-0.5%. Goldman’s tariff estimate has not yet been revised for the China policy changes; the speaker expects the overall impact to be revised up by 0.3%-0.4 percentage points and GDP by around 0.4 percentage points, leaving the outlook “not particularly optimistic.”
Policy can deliver an immediate sentiment shock to equities, but its impact on the real economy takes roughly six months to work through: the effect of the April tariff increases may not appear until the end of Q3 or in Q4. A slowdown in growth from Q4 2025 through Q1 2026 remains the speaker’s most likely path.
Trump’s supporters believe roughly $200B-$300B of tariff revenue could be returned by year-end. But the House Ways and Means Committee’s proposal would cost around $5T, already exceeding congressional authorization, and also involves the debt ceiling around June 30. Even if the tax cuts pass in full, the money may not reach the economy until November, with the effects delayed until the end of Q1 or mid-2026.
6. AI and ASIC offer better odds beneath the index
Even with earnings being revised down across the board, the speaker still sees AI and technology as relatively defensive, especially semiconductors that suffered the steepest declines. Compared with banks, investment banks, and names such as Goldman Sachs and Morgan Stanley, which depend more directly on how much regulation Trump ultimately loosens, he sees greater upside in selected AI, ASIC, and semiconductor stocks.
The ASIC path runs in parallel with NVIDIA’s GPU business, and the speaker identifies Broadcom and Marvell as the core companies capable of providing complete ASIC solutions. Broadcom has a long history of serving Google TPU and a stronger track record, so the market is more willing to trust its earnings guidance; Marvell has only recently shifted toward AI, faces more skepticism over its order pipeline, and therefore offers greater upside torque.
Marvell’s problem is not that projects will necessarily disappear, but that cloud providers and IC turnkey contractors have strict NDAs, preventing the company from promptly disclosing which project it has won or how it is progressing. Hedge-fund positioning is also heavy, making it easy during a downturn to short or trade the stock around fragmentary news and turning it into “a very good stock to manipulate.”
Discussing the Microsoft order and Maya 3, the speaker said it was “very certain” that the companies involved would earn about $3 per share this year and “also very certain” that earnings would grow 70%-80% next year; he also mentioned Training 3 and Training 4. The source does not make clear which company “it” refers to, so the figure is not assigned to Marvell here. Overall, he sees greater upside than downside in some AI and semiconductor names that suffered heavy earlier losses.
7. Intel 18A and long bonds need to be priced on cycles, not headlines
The market is focused on whether Intel can win Microsoft’s roughly $1.8B networking-card order, but the speaker notes that semiconductor projects typically run 12 to 18 months. The relevant testing, routing, and bring-up work has been underway for some time; this is not a deal suddenly negotiated after 陈立武 took over. If 18A yields are solid, many of the outcomes are already relatively certain.
What Intel lacks most is an external team to tape out the new production line, refine the process, and communicate with the market. Customers have previously worried both about committing R&D capital and about offending TSMC. The speaker believes 陈立武’s background as Cadence CEO, along with his role as an important financial investor in TSMC while at Walden and his close relationship with TSMC, gives him the ability to bring in Cadence, NVIDIA, and other ecosystem teams to solve the bottleneck. Marvell and Intel are his personal holdings.
On 10-year yields, he expects a modest steepening of the curve, with the long end roughly holding at 4.2%-4.5% by year-end. At around 4.5%, there is little obvious room for another leg higher. Treasury issuance is lower than last year and domestic demand is relatively stable, while slower growth from Q3 and Q4 2025 through Q1 2026 should weigh on the term premium. A Treasury short is therefore “not very meaningful”; the better overall strategy remains to trade less and wait for hard data.