Market participants are looking for signals of clarity and directionality, but they keep getting conditional moves. Macro is stable enough to lift major indices and selective outperformers, but not solid enough to reward laziness. If yields stop cooperating, the same portfolios that own NVDA, MU, SOX, crypto beta, and long-duration software will discover they own one trade: falling discount rates plus AI capex confidence.
Key Takeaways
- One crowded macro trade: falling discount rates plus AI capex.
- The market still supports risk, but it is no longer subsidizing weak structures.
- Long-duration Treasuries are still in a multi‑year drawdown.
- U.S. consumption is being propped up by wealthy households, not broad household health.
- Crypto is underperforming equities and needs a stronger marginal buyer.
- NVDA remains the kingmaker, but the market will start policing customer concentration and spend durability.
- The S&P’s record run is driven by AI earnings while consumers dip into savings.
- Do less, but make each unit of risk work harder.
Macro: Growth Holds, Cuts Wait
Headline PCE rose 0.4% m/m and 3.8% y/y, up from 3.5% in March and the fastest annual pace since May 2023; core PCE rose 0.2% m/m and 3.3% y/y, up from 3.2% in March. The headline number matched consensus, so the market did not get a shock, but the mix is still bad for risk: prices are rising faster than incomes, and the consumer is spending by drawing down savings.

The equity rally is happening while the old portfolio hedge is still broken. The U.S. Treasury total return index has been in drawdown for 69 consecutive months, with TLT down roughly 40% from its April 2020 peak. The point is not that bonds cannot bounce. The point is that long-duration Treasuries are not supplying the liquidity release that risk assets usually want at this stage of a rally. That makes the semi move more fragile than the price action looks.

The Fed can debate the rate of change; voters and consumers remember the price level. The market is making highs while the consumer feels worse, and that split matters for breadth. University of Michigan final May sentiment fell to 44.8 from 49.8 in April and 52.2 a year earlier; 57% of consumers spontaneously cited high prices as damaging their personal finances, up from 50% in April. The pressure is coming through gas, essentials, and inflation expectations, not just politics.

Inflation risk has moved back from “last mile” annoyance to positioning constraint. March PCE was still +3.5% y/y, with the next BEA print due May 28, while April CPI accelerated to +3.8% y/y and core CPI rose +2.8% y/y; energy rose 17.9% y/y and gasoline rose 28.4% y/y. The Fed’s next meeting is June 16–17, and the bar for cuts is higher.

The U.S. consumer is holding up because the top decile is still spending, not because the broad household sector is healthy. According to Moody’s Analytics, the highest-earning 10% of households now account for nearly half of U.S. consumer spending, up from just over one-third about 30 years ago. RBC makes the same point from financial-account flows, arguing aggregate consumer data overstates resilience because the top 10–20% drive the bulk of outlays and hold the liquid, income-producing assets that benefited most from the post-2020 asset boom.
Crypto: Collateral, Cash Flow, or Cut

BTC — BTC is trading ETF redemptions and macro oxygen. The same ETF rail that made BTC feel structurally bid in April and early May is now a forced seller during a hot PCE / higher-oil week. Equity risk and crypto risk have decoupled short term: SPX and NDX are holding records on AI and semis, while BTC is getting sold through regulated products.

ETH — Ethereum has an execution-risk problem amidst the ongoing Foundation brain drain. Carl Beek and Julian Ma became the latest senior EF exits, after seven and four years respectively; Beek worked on Beacon Chain design, while Ma worked on censorship resistance and cross-layer bridge architecture. That takes the recent tally to at least eight high-profile departures, with five clustered in May. Barnabé Monnot and Tim Beiko are moving on, Alex Stokes is taking sabbatical, and Will Corcoran, Kev Wedderburn, and Fredrik now coordinate the cluster.

The short-term defense is that the roadmap still has teeth. The EF’s May 11 update points to a 200M post-Glamsterdam gas-limit floor, ePBS running on a multi-client devnet with external-builder testing, EIP-8037 repricing finalized, and Hegotá groundwork moving through FOCIL prototypes and native account-abstraction scoping.

HYPE — HIP-4’s prediction/outcome markets are the first live test of whether HYPE can grow from a perp-exchange token into broader market-structure collateral. These are fully collateralized binary contracts that trade between 0.001 and 0.999, settle to 0 or 1, and avoid funding or liquidation risk; launch-day volume reached 6.05M contracts.

The May 25 update extends HIP-4 outcomes to canonical markets based on offchain events, with validators voting on deployment and settlement quality. That creates a real product wedge and moves validator judgment closer to the P&L of traders.

The current bid is being paid by three buyers: fee-funded open-market buybacks, new ETF/ETP access, and traders underwriting Hyperliquid as the onchain venue for 24/7 macro/event risk. The bias is to stay long while volumes feed the Assistance Fund, canonical markets resolve without controversy, and HYPE holds the May breakout zone; cut exposure if an offchain settlement turns political, the buyback run-rate drops with volume, or the token loses momentum while ETF headlines stay positive.
NEAR — Near’s privacy + AI bet continues to pay off, with one caveat: the market is now awake. The thesis is intents + agent execution + June’s upgrade introducing dynamic resharding and post-quantum-safe signing. NEAR Intents crossed $19B of cumulative volume, and sentiment also flipped hard after Arthur Hayes put NEAR into the “AI/agent” conversation.

ZEC — ZCash remains the better privacy trade, but similar to NEAR, it is no longer under-owned. It resembles a violent two-way positioning trade, not a quiet privacy compounder. The swings increasingly feel more oversized as perps volume is orders of magnitude greater than spot (roughly $1.5B of OI and $4.5B of futures volume against about $450M of spot.)

UNI — UNI’s week is about fee expansion failing to create urgency. Uniswap Foundation Proposal 96 went live on May 24 and would extend protocol-fee collection and burn infrastructure to BNB Chain and Polygon. Value accrual is increasing gradually, slowly. A trade needs more confirmation that fee activations will become visible buy pressure. Volume is low on spot, which makes it a no-go for momentum traders.

MORPHO — The Midnight whitepaper was released, which MORPHO from “Aave relative winner” into a broader credit-market bet. We laid out the thesis in our latest memo. The new protocol is EVM-based, non-custodial, fixed-rate, fixed-maturity lending: markets are isolated, immutable, permissionlessly created, and loans settle through credit and debt units that behave like zero-coupon obligations. The important design choice is offered liquidity. Makers can quote across multiple markets without locking capital, then source funds only when an offer fills; the same balance can sit in Morpho Blue or elsewhere until execution.

OKB — OKX is trying to move exchange infrastructure into the protocol layer, imitating the Hyperliquid playbook. Exchange OS on X Layer lets developers and institutions deploy spot, perpetual, and outcome markets using OKX-style matching, margining, liquidation, settlement, and risk infrastructure. Builders can configure assets, oracles, revenue models, and compliance frameworks; venue creation requires OKB staking. The first test comes in June with 2026 World Cup Outcomes, but OKX calls it a simulated market with no real-world value.

WLD — The most direct OpenAI-adjacent crypto trade pulled a double-digit up move reaching close to ~$1B in daily volume. This move, however, came after printing capitulation. WLD had just hit an all-time low at $0.2303 on May 18. By May 24 it had only recovered to about $0.2947, still stuck below the $0.30–$0.31 battle zone. The violent move came after it cleared that zone, not on the first OpenAI IPO headline, which came out last week on May 20.

ARKM — The market is digesting OpenAI’s IPO rumors, looking for beta beyond WLD. Arkham was backed by Sam Altman. The thesis is onchain intelligence plus AI-adjacent data distribution. Arkham has recently shipped prediction markets, a decentralized trading product, an AI-agent API, and more wallet-label coverage. Like WLD, ARKM was also >30% up over 7d at the beginning of the week, with volume up >200% day-over-day, though it still sits roughly 96% below ATH and carries locked supply overhang. This week the ZEC debate put Arkham’s labeling product back in the feed after coverage resurfaced its claim that it had linked about $420B of Zcash activity to identifiable entities.

UB — UB is trading as AI-agent memory beta, with enough product language to attract flow and enough token-structure risk to punish late entries. Only 2.5B of 10B tokens are circulating, and the token is up >80% on the weekly timeframe. The current driver is the agent stack: Unibase sells persistent memory, onchain agent identity, x402 payments, and agent-to-agent commerce. Price already did the hard part earlier in May, breaking out of a seven-month $0.02–$0.06 range, clearing $0.16–$0.17, and tagging roughly $0.21. That means the trade is now execution-sensitive. A hold above $0.17 keeps the discovery structure alive; failure back below that band turns the setup into an overbought AI chase with poor float optics.

ONDO — Ondo founder Nathan Allman died unexpectedly, with no cause disclosed. Longtime president Ian De Bode will take over as CEO. De Bode had already overseen strategy, product, and daily operations for more than two years.

RAIL — RAIL is the privacy rotation in its purest DeFi form. The token has been up >100% over the past 7 days, even after pulling back to roughly $3.63 and daily volume already cooling to about $2.5M (having reached $7.5M, which is roughly 10x typical levels). ZEC reopened the trade, but RAIL gives traders a smaller-cap way to express private DeFi, wallet-level privacy, and Ethereum-adjacent privacy tooling. The market also resurfaced the Ethereum Foundation / Kohaku SDK integration narrative, Railgun Connect, and hardware-wallet integration work, which makes RAIL look less like a mixer proxy and more like wallet infrastructure.

ARB — The Foundation is asking the DAO for another year of runway: $16M in RWAs/stablecoins, 1,740 ETH, and 230M ARB, worth roughly $43.5M at current marks. The ARB piece alone is 2.3% of max supply, or about 3.7% of current circulating supply, for one year of operations and ecosystem spend. The proposal cites 4.7M+ daily transactions, $8.6B in stablecoin supply, roughly $800M in RWAs, and $23.49M of 2025 gross profit flowing to the DAO. Still, the market will not pay a premium for a token whose clearest recurring use is funding the operating layer again. At the ask level, the DAO is paying about 1.9x last year’s gross profit for next year’s Foundation budget.

Equities: Margins, Chips, Distribution
SPX — The S&P500 continued rising to its highest level on record, marking a 9-week win-streak since 2023. The index is still levitating on AI earnings and semi leadership even though macro refuses to cooperate. BEA’s April data also showed consumer spending through weaker income: personal income was basically flat, disposable income fell 0.1%, spending rose 0.5%, and the saving rate fell to 2.6%. It is an earnings and AI-driven market sitting on a thinner consumer base and higher inflation tail. Stay long while SPX holds above 7,500 and semis keep leading. If yields push higher after PCE and SPX fails to hold the record breakout, the right move is to cut cyclical and lower-quality beta first.

NDX — NDX is the most clear expression of the current market, and that is also the problem. The move is being pulled by AI infrastructure, memory, data-center software, and the broader semi squeeze, not by a calm macro backdrop. It is holding near records despite oil, inflation, and Iran headlines keeping pressure on rates. The pain trade is still up, but the hidden position is the same across portfolios: long AI capex, long falling real-rate tolerance, short anything that challenges the multiples.

NVDA — Nvidia is still digesting a monster print after reporting revenue of $81.6B, up 85% year-over-year, with Data Center revenue of $75.2B, up 92%, plus an additional $80B buyback authorization and a dividend increase. The after-print fade came from expectations, not weak demand. The company delivered the quarter the market asked for, then traded like the bar had already moved higher.

GOOG — GOOG’s current bid comes from full-stack AI distribution, not model hype alone. Google Cloud Next introduces Gemini Enterprise, agent tooling, and eighth-generation TPUs. The Blackstone TPU-capacity deal turns Google’s internal AI infrastructure into a cloud product and pressures neocloud margins. Alphabet remains one of the most attractive AI-distribution longs as it owns chip, cloud, model, app, browser, and search entry points.

AAPL — Apple’s driver is pre-WWDC AI repricing ahead of delivered product revenue. The stock traded around records >20% up QoQ as investors price a stronger AI story into the June 8 WWDC event. Bank of America raised its target from $330 to $380, arguing Apple can win through iPhone distribution, identity, permissions, trust, and app-level control.

META — Meta is caught between efficiency and AI capex anxiety. The market still likes the ad machine, but it is forcing Meta to prove the AI spend. Meta will cut about 8,000 jobs, eliminate 6,000 open roles, and reassign roughly 7,000 employees toward AI, while capex guidance has moved into the $125B–$145B zone as data-center and memory costs rise. The bull case is self-funding AI: ads and engagement pay for compute, AI tools improve ad yield, and wearables become a second interface. The bear case is capex outrunning evidence, with Google and Apple now attacking the same agent/glasses surface.

TSLA — Tesla is technical repair plus SpaceX/AI optionality. It’s been forming a three-weeks-tight pattern inside a cup-with-handle with a $450 buy point after a roughly 30% recovery from the April low. The SpaceX IPO provides a breath of fresh air, but the constraint is still on the demand side: Chinese retail sales were reported down 15% year-to-date through April. A clean break above $450 can pull in technical buyers; a failed breakout says the SpaceX/AI halo is masking weak auto data again.

MSFT — OpenAI is preparing to file confidentially for a U.S. IPO in the coming weeks, with a possible listing as early as September; OpenAI itself closed $122B of committed capital at an $852B post-money valuation. Microsoft’s 27% stake gives it exposure but MSFT is too large for this to trade as a proxy. At MSFT’s current ~$3T market cap, a 27% look-through stake at OpenAI’s $852B mark is roughly 7.4% of MSFT; even a jump to a $1T OpenAI valuation adds only about 1.3% of MSFT market cap. Limited convexity.

SPCX — SpaceX is reportedly coming at a ~$1.75T valuation and seeking up to $80B, while OpenAI is trying to force itself into the same portfolio conversation. If Anthropic follows the window, public investors will need to reserve a risk budget for AI access on the board. The bull case is Starlink + reusable rockets + passive index demand; the bear case is the filing: Q1 revenue of $4.69B, total operating loss of $1.94B, Starlink operating profit of $1.19B, and AI losses of $2.47B on only $818M of AI revenue; AI also consumed 76% of SpaceX’s $10.1B Q1 capex after the xAI purchase. Overall, SPCX is a strange hybrid: one proven infrastructure monopoly, one capital-intensive rocket platform, and one AI loss furnace being used to justify a much larger TAM. The buyer base is likely retail enthusiasm plus passive/index absorption.

COIN — Coinbase is trading like a policy-and-product call option. The problem this week is that the SEC delayed its tokenized-stock “innovation exemption,” which directly hits the near-term story around Coinbase becoming the venue where stocks, crypto, USDC balances, prediction markets, perps, and options collapse into one account. Coinbase’s own stock product is live for U.S. users with 24/5 stock access, zero commissions, USDC funding, and up to 3.5% USDC rewards for Coinbase One users, but that same reward mechanic remains exposed to the stablecoin-yield fight in Washington.

CRCL — As an alternative to Coinbase, Circle is the stablecoin pure-play, but the stock is really a margin fight wrapped in a USDC growth story. The market is no longer paying for Circle’s good numbers. Circle’s Q1 showed the good part: $77B of USDC in circulation, up 28% y/y; $21.5T of onchain USDC transaction volume, up 263%; $694M of total revenue and reserve income, up 20%; and $151M of adjusted EBITDA, up 24%. The less comfortable line is operating leverage. Reserve income rose 17% because average USDC circulation grew, but the reserve return rate fell 66 bps; distribution, transaction, and other costs rose to $407M, while operating expenses rose 76% on post-IPO stock comp and investment spending. That is why CRCL does not trade like a simple USDC supply chart. More circulation helps, but rates, Coinbase economics, rewards rules, and distribution costs decide how much reaches equity.

SOX — The AI trade is now an index concentration problem at ~18% of S&P 500 market cap, while SOX has outpaced the Magnificent 7 by a wide margin since early 2025. Semis are no longer a sector sleeve. They are market beta, AI capex, U.S. industrial policy, Taiwan risk, power demand, memory scarcity, and duration all in one trade. That can keep working, but it means “diversified AI exposure” often owns the same factor five ways.

MU — Micron is the marginal AI trade this week. The stock kicked off the week up >20% after UBS lifted its target from $535 to $1,625, arguing that long-term agreements and AI demand can make Micron’s earnings less cyclical than old memory cycles. Micron’s fiscal Q2 revenue was $23.86B, nearly triple last year’s $8.05B, non-GAAP EPS was $12.20, non-GAAP gross margin reached 74.9%, and the company guided fiscal Q3 revenue to $33.5B ±$750M with about 81% gross margin.

Semis — Semis remain the market's upside engine and its most crowded hidden macro trade. BlackRock’s Q2 outlook says hyperscaler capex estimates through 2030 rose about 25% since October, even as Middle East energy pressure reduced hopes for easier Fed policy. NVDA at ~$215.33 and SMH at ~$576.32 still work while capex revisions rise and earnings revisions confirm. The risk is correlation, not valuation alone: SPX, NDQ, Mag7, AI power, semis, and crypto beta now share the same liquidity/capex impulse.

Miners — Miners are finally getting paid for operating margin rather than just metal optionality. The World Gold Council’s Q1 report showed a record $4,873/oz average LBMA PM gold price, supply up only 2% y/y to 1,231t, and demand mix shifting toward investment demand while jewelry volumes fell. GDX trades near $85 and GDXJ near $112. Own miners when gold holds high while oil and labor costs stay contained; reduce if gold breaks lower, AISC inflation (the rise over time in the all‑in sustaining cost per ounce that miners report, driven by higher inputs like labor, diesel, explosives, maintenance, royalties, and sustaining capex) catches up, or generalist money rotates back into AI without confirming miner flows.

ENHA — The direct Enhanced Games ticker, closer to a post-SPAC credibility test than a long setup. Enhanced said the May 24 inaugural Games secured more than $32M of sponsorship deal value, with partners including ROKU, RUM, and RZLV, while also warning that sponsorship deal value is not GAAP revenue and may differ from recognized revenue. The market cared more about what it delivered, and the stock fell more than 40% after the event, wiping out almost $800M of market value. The Games created headlines, but the athletic result undercut the pitch. Only one unofficial world record was recorded. That was Kristian Gkolomeev’s 20.81-second 50m freestyle swimming, helped by a banned skinsuit on top of the enhancement protocols.

Conclusion
The right posture is selective risk, not blanket risk-on. Own what has passive flow, recurring demand, or a credible product loop. Fade what depends on fresh catalysts. The market can keep grinding higher, but the pain trade is no longer random beta. The pain trade is owning too little of the few assets still getting paid and too much of everything that only rallies when positioning gets sloppy. Do less, but make the exposure count.


