Skip to content

Pulse - May 9

9 May 202622 min
Read by asset class: Macro
ContentsOverview

Stocks are back at all-time highs while consumer savings look thinner and the 30-year Treasury is pressing against the 5% ceiling again. AI capex is carrying the growth story, but duration is starting to compete with equities. Crypto is not broad enough for blind beta in these conditions. The market is pricing a boom, but the boom is narrow. That leaves us with a simple playbook: respect the equity trend, do not ignore the 5% Treasury ceiling; and in crypto only pay for liquidity, value accrual, and catalysts that can actually move price.

Key Takeaways

  • Stocks are high, savings are low, and the consumer cushion is thinning.
  • The 30-year near 5% is the pressure point; equities can ignore it until they can’t.
  • AI capex is no longer a tech-sector story; it is the U.S. growth story.
  • The flaw in the AI trade is concentration: too much growth now sits on too few balance sheets.
  • Semis stole crypto’s speculative role this week: scarce compute, memory, and storage got the bid.
  • BTC still sets crypto risk appetite; alts need confirmation.
  • RWA headlines keep accruing to companies first; token holders still need a direct cash-flow path.
  • DePIN/storage names are catching AI beta, but most still need proof of durable demand.

Macro: ATH Stocks, Empty Wallets

SPX and NDQ are pressing records as oil’s pressure is coming out and AI/earnings leadership is carrying the market. Asia is also catching record-high bids, while gold recovered toward a 2-week high. BTC is holding the risk-on handoff around $80-82k, helped by ETF inflows and better macro tone, but the trigger is still further consolidation above the low-$80k area.

Stocks at all-time highs just as the U.S. Personal Savings Rate falls to 3.6%, the lowest reading since 2022. Employment is contracting while rising prices become more sticky.

The 30-year Treasury is tagging 5% (bps away from an 18-year high), revisiting a level that has been a hard ceiling for the past two years. At 5%, long-dated Treasuries start to look competitive with stocks, while mortgages, corporate loans, and government financing all get more expensive at once.

Markets have seen this movie twice already—late 2023 and early 2025—when yields flirted with 5% but couldn’t stay there, and the S&P 500 rolled over each time. High-net-worth portfolios are also crowded into equities, while hedge funds have been cutting U.S. tech exposure at the fastest two-week pace in roughly a decade.

AI: Capex Is the Cycle

AI, conceived as the master key technology, cannot be framed in isolation. AI capex is not a sector-specific issue but a macro problem. Morgan Stanley now expects Amazon, Alphabet, Meta, Microsoft, and Oracle to spend roughly $805B on capex in 2026 and $1.1T in 2027, with the 2026 spend alone near the capex of the entire non-tech S&P 500 combined.

Turns out that AI capex has already been a roughly 2% boost to GDP growth this year. Morgan Stanley’s latest research puts that figure closer to 2.5-3%, >3% in 2027 and already drove roughly 75% of Q1 GDP growth. That’s bullish for compute providers, semiconductors, data centers, and the “AI as infrastructure” trade, but it also concentrates the whole U.S. growth story into a select few balance sheets.

Why it is bullish: every dollar of hyperscaler capex shows up as revenue for someone else. The big platforms—Amazon, Google, Microsoft, Meta, Oracle—sign off on the build, and a whole ecosystem gets paid: Nvidia/AMD/Broadcom, memory makers, server manufacturers, fiber and cooling vendors, electrical gear suppliers, grid contractors, utilities, data‑center REITs and private operators, and private credit funds.

That’s why Morgan Stanley describes AI as a physical infrastructure super‑cycle at roughly $2.9T in global datacenter capex between 2025 and 2028, with about $1.4T funded directly from hyperscaler cash flows and the rest raised via debt and other capital sources. This explains why this is no longer just a “software productivity story,” but an industrial investment boom centered on power, land, chips, financing, and construction.

For markets, it’s also worth tracking what the public opinion is saying, especially as AI leaders change their discourse from the dream that “AGI replaces labor” toward a reframing of AI as a job-creating, human-augmenting layer. It turns out “we are building the thing that makes you obsolete” was a terrible sales pitch.

The pendulum swings fast. As token costs drop, more complex agentic workflows start becoming economically viable. The insight is that these agentic workflows in turn consume far more tokens through longer context windows, repeated reasoning loops, validation, tool use, and always-on background monitoring.

The feedback loop kicks off once enterprises feel FOMO about leaving chips on the table if they aren’t firing on all cylinders 24/7. An increase in token usage inevitably results in better infrastructure, which strengthens unit economics, creating more space for hyperscalers and model providers to continue improving and reinvesting in model quality, distribution, and capacity.

Signals Radar

BTC — Started the week rising above $80k for the first time since Jan 31st and held up strong up there while MicroStrategy acted as both a structural bid and headline risk. Strategy reported a Q1 net loss of $12.54B holding 818,334 BTC as of May 3 after having raised $11.68B YTD. That marked its BTC at a $64.14B market value with an average cost basis around $75,537. The latest disclosed buy was 3,273 BTC for ~$255M at ~$77,906 per coin. The wrinkle: Saylor said on the May 5 earnings call that Strategy would probably sell some BTC to fund a dividend as a market “inoculation.”

ETH — Flow repair without big momentum enthusiasm while the Foundation now focuses on Interop and ETF flows look constructive.

SOL — Western Union launched USDPT on Solana, and Alpenglow’s ~150ms finality keeps SOL leading the payments/ICM race among majors. At Consensus Miami Anatoly said Alpenglow could arrive as early as next quarter, but no formal Solana Foundation timeline is live yet.

HYPE — Beyond fundamentals, points speculation adds extra convexity to this bet. This changes user behavior and can skew revenue multiples while HIP-3 and HIP-4 keep expanding Hyperliquid’s ecosystem, which now supports builder-deployed markets, tokenized equities, commodities, pre-IPO contracts, and prediction markets.

ZEC — Started the week with double-digit pumping and Multicoin set a fuse under it saying they have been buying since February. Crypto is now bifurcated into two lanes: the institutional TradFi route, for which there aren’t many investable liquid tokens, and the cypherpunk ethos that kicked off the entire movement, for which ZEC now acts as the kingmaker of the privacy narrative. Beyond market sentiment, fundamentals look substantially better than in prior cycles, as NU6.1 sends 12% of block rewards to a coinholder grants program and 8% to Zcash Community Grants through the 2028 halving.

TON — The token pumped >30% into a multi-month high as Durov posted on X that he intends to turn Tonchain into an explicit Telegram trade. Fees are down 6x to near-zero, and Telegram is now set to replace the TON Foundation as the network’s driving force becoming TON’s largest validator. This may increase centralization risk, but could also result in better execution, stronger distribution, lower UX friction and tighter alignment with one of the largest social networks across the globe.

AAVE — After the April 18 rsETH bridge incident, AAVE buybacks have been paused since April 19 to preserve DAO balance-sheet flexibility; that removes the mechanical bid the market had been leaning on. Now the May 1 restraining notice and May 4 emergency motion over roughly $71M of frozen ETH turns the recovery into a court/governance test, with Aave arguing those funds belong to victims, not outside creditors.

MORPHO — Constructive setup ahead of Midnight, although MORPHO is still not a cash flow asset. Blue already proved share capture: Morpho’s borrow book rose from ~9% of Aave’s in Nov. 2024 to 33% today, with ~$7B TVL and ~$3.9B borrowed, while its fixed-market design lets Coinbase, Fireblocks, Anchorage, Taurus, Apollo, Bitwise, Steakhouse, Gauntlet and other curators build isolated credit products without waiting on protocol governance. Midnight moves Morpho from crypto leverage into institutional credit, where duration, collateral, and funding costs can be underwritten. Distribution is already there, as enterprise integrations were $1B of $4B active loans, exchanges/fintechs supplied $1B of $6B+ lent assets, Fireblocks opened vault access to 2,400+ enterprise customers, and Apollo affiliates may buy up to 90M MORPHO over 48 months.

ZRO — Bryan’s acknowledgement is the right kind of uncomfortable message that rebuilds long-term trust. The protocol, aiming to stay unopinionated and permissionless, pushed many security assumptions to the edges, presuming that 1/1 DVN configurations wouldn't be common among protocols integrating LayerZero’s infra. With Zero on the horizon, expected for Q3, the bull case remains asymmetric if trust stabilizes. Zero gives the token a higher-utility path, Stargate keeps funding buybacks, and the next fee-switch window reopens governance optionality.

ENA — ENA presents the right asymmetry for those trusting its team’s execution, but it’s a victim of bad timing. Unlocks are small enough to absorb if buyers are present, but they can get loud during periods of illiquidity. The bull case hinges on broadening USDe collateral and revenue sources into institutional lending, RWAs, equity/commodity basis, and Hyperliquid-linked opportunities. Short-term, however, we are looking at key metrics that have retraced significantly from prior highs in recent months.

SKY — Once seen as vulnerable to Ethena and widely criticized for a messy rebrand, Sky’s USDS and sUSDS keep growing: record Q1 revenue of $123.79M, $46.04M protocol surplus, $11.70B USDS supply, and $6.49B sUSDS supply, while live SSR sits at 3.75% with ~6.23B sUSDS outstanding.


PENDLE — Despite being the market leader and kingmaker in the yield trading space, Boros wasn’t enough to re-rate the token. The change from vePENDLE to sPENDLE tokenomics, however, is starting to pay off. Right now 80% of all protocol revenue is used for market buybacks of PENDLE, and distributed to sPENDLE. That’s over $1.8M in buybacks this year so far. Thanks to Hyperliquid’s HIP-3, Pendle now has a path from crypto funding to equity funding, index funding, and eventually a broader rate surface.

DRIFT — Trust-repair optionality starts now. After the April 1 exploit, confirmed as DPRK-affiliated and sized at ~$295M in user losses, Drift has laid out a recovery plan that issues affected users a separate recovery token, seeds the pool with remaining assets, then funds redemptions through exchange revenue, up to $127.5M from Tether, and up to $20M from partners until the pool reaches ~$295.4M. It’s a setup that can get reflexive quickly, with downside being linked to delays, governance fights, or user-recovery disappointment.

JTO — Introducing JTX as an all-in-one onchain trading platform of which 80% of revenue is allocated to JTO. Jito does not want to wait for third-party apps to build the trading experience on top of its infra. Jito built the Solana trading engine first: MEV-aware infra, JitoSOL, TipRouter, StakeNet, and BAM-style block assembly meant to improve execution and reduce toxic ordering games. Now it wants to own the user interface too. JTX is the consumer-facing trading app: spot first in July, later perps via Phoenix and prediction markets, with the pitch being self-custody, CEX-grade execution, pro order types, and support for RWAs.

MEGA — Post-TGE MEGA is quality tech wrapped in hard float math. The initial float is small and FDV is large while public-sale and Echo/Fluffle vesting dates create future sell pressure. The first post-listing move already showed how quickly a high-attention launch can lose altitude.

CFG — Coinbase named Centrifuge a Preferred Tokenization Infrastructure platform and made a strategic investment. This aligns Base’s distribution with Centrifuge’s institutional stack across Apollo, Janus Henderson, and S&P Dow Jones Indices. The edge is distribution + usability: Centrifuge’s deRWA framework already puts deSPXA on Base, giving eligible non-U.S. users tokenized S&P 500 exposure that can move through DeFi rails instead of sitting trapped in a permissioned vault.

MON — Monad’s RWA angle improved as its main competitor, MegaETH, became more active with the Aave-Ethena deal boosting its TVL. Centrifuge brought tokenized Treasury bills, Janus Henderson AAA CLO exposure, and Apollo credit strategy exposure to Monad, including freely transferable deRWA wrappers for 24/7 DeFi use. Tokenized U.S. stocks are also live through Monday Trade/Anchored with USDC settlement and 24/5 trading. The first major unlock is not until Nov. 24, 2026.

WLFI — World Liberty filed suit against Justin Sun, while Sun’s side has already fought the asset freeze. The extra 5.9B WLFI private sale, with most early investor holdings still restricted, adds supply complexity and keeps the marginal-buyer question open.

ONDO — Continues to line up institutional tokenization headlines. DTCC’s tokenization working group includes Ondo alongside BlackRock, Goldman, JPMorgan, Circle, Robinhood, and others; DTCC is targeting limited production trades in July 2026 and full launch in October 2026. This won’t result in short-term cash flows for holders, but it expands Ondo’s distribution credibility. The reality, though, is that news can keep accruing to companies while token holders wait.

FIL — As semiconductors started rallying in TradFi, FIL acted as liquid storage beta to the DRAM/memory rally in equities. The 2026 roadmap may have some catalysts rallying behind in the form of onchain deals, stronger network economics, flagship client adoption, and Filecoin Onchain Cloud as the payment layer for storage, retrieval, and compute workflows.

AR — Similar to Filecoin, Arweave rallied as the permanent-data/provenance leg of the storage bid, helped by the same “AI needs data rails” rotation lifting FIL/STORJ/ICP.

AKT — Aiming to become a DePIN leader, the priority is to achieve value-accrual through BME (Burn-Mint Equilibrium), finding a sweet spot where compute demand can translate into AKT buy-and-burn pressure via ACT credits, while providers still receive payments in stables. The 2026 roadmap also points toward shared security to reduce staking inefficiency and operational overhead.

POL — Polymarket’s migration to another chain is top-of-mind, which Polygon is countering with headlines: Visa added Polygon to its stablecoin settlement program, and Meta is using Polygon rails for USDC creator payouts.

GNO — The topic of RFV Raiders (coordinated groups looking to capitalize from distressed DAO treasuries) came back, this time picking Gnosis DAO as victim. Gnosis raised $12.5M in 2017 and now has >$200M in assets without any fundraise in between. In the past, these attempts targeted protocols whose treasury value was greater than their token’s market cap.

The GNO-150 proposal is more complex, since the discount to NAV is quite narrow. The Gnosis treasury sits at $223M in ETH, stables, and ecosystem tokens; 1.3M GNO tokens eligible to redeem against it. That’s a ~20% gain with each redeemable GNO claiming ~$170 in treasury backing. This came on the back of token holders noticing Gnosis Ltd takes ~$30M/year of DAO money and produced $400k of revenue while its token trades below NAV.

CARDS — Low-cap reflexivity that can return violently when attention returns. Beyond a narrative trade, the protocol might as well be framed as a cash flow machine. Collector Crypt did $146M gross revenue and ~$9M gross profit in Q1, implying ~$37M annualized profit against a market cap quoted under $20M.

GME — The equity proxy for cross-asset risk-appetite. It proposed a $56B cash-and-stock bid for eBay at $125/share, with a 5% stake built and $20B debt support lined up; the stock popped, then erased gains and traded lower. For crypto, it offered some relief to see that animal spirits are still alive, although they are less forgiving.

COIN — Brian Armstrong posted publicly that Coinbase is cutting ~14% of staff, or about 700 roles. He framed it as operational leverage, but ultimately alluding to a number of factors like AI-driven productivity and volatile crypto volumes. The irony is that two days later Coinbase reported a weak Q1: revenue fell to ~$1.43B from $2.03B a year ago, transaction revenue dropped ~40% to $756M, subscription/services fell 13.5% to $583.5M, and the company posted a $394M net loss versus a $65.6M profit last year.

For COIN, the bull case is that they hit an all-time-high 8.6% crypto trading volume market share, derivatives TTM volume grew 169% y/y, retail derivatives passed $200M annualized revenue, prediction markets hit $100M annualized in less than two months, and Base is now the dominant venue for onchain stablecoin and agentic payment flows. But the stock is still hostage to spot volumes, risk appetite, and operating leverage.

INTC — Bullish sentiment was already primed by Intel’s Q1 beat: $13.6B revenue, +7% y/y, non-GAAP EPS of $0.29, and Q2 revenue guide of $13.8–$14.8B. This week the asset took off adding $55B to its market cap on a 12% move on Tuesday. Apple has held early talks with Intel and Samsung to manufacture main device processors in the U.S., giving Apple a possible second source beyond TSMC. AI CPU demand + U.S. foundry nationalism + Apple optionality turns INTC from broken legacy into a strategic supply-chain call option.

AMD — Adding almost $100B in market cap after reporting earnings, rising above $400/share for the first time in history. Q1 revenue was $10.253B, +38% y/y; data center revenue was $5.8B, +57%; non-GAAP EPS was $1.37; and the company guided Q2 revenue to $11.2B. The market is now looking for NVDA beta, starting to pay for a CPU + GPU inference ticker with tailwinds like Meta’s plan to deploy up to 6GW of AMD Instinct GPUs, plus Samsung collaboration on HBM4 supply for future Instinct GPUs.

Semis — The equity casino stole crypto’s job as the market started bidding for scarce compute, memory, and storage. In memory, South Korea’s Hynix printed record Q1 numbers, with revenue +198% y/y to KRW 52.6T and operating profit KRW 37.6T, while saying client HBM requests for the next three years already exceed production capacity. In Korea, KOSPI ripped through 7,000 on May 6, with Samsung +14.4% and SK Hynix +10.6%. A mental model for this basket can be as follows: SNDK/MU/SK for pricing-power and supply-shortage plays; AMD/ARM/MRVL/QCOM for AI compute, networking, and silicon demand plays; and INTC/GFS for capacity beta and further upside if execution keeps improving.

OpenAI — Word on the street is that OpenAI is eating Anthropic’s lunch. Numbers don’t lie, and they recently started reflecting the negative sentiment vs Anthropic and the skyrocketing adoption of the Codex app. OpenAI owns momentum now; Claude’s side got hit by access/rate-limit friction and negative usage-limit sentiment while Codex moved from 3M to 4M weekly developers right before an enterprise push along the “Codex for everything (not just developers)” message.

Anthropic — Anthropic’s onchain pre-IPO proxy on Jupiter now implies a $1.2T valuation, +20% in 7 days and +900% since Oct 2025. But it might as well be reflexive private-market heat, rather than a more useful public equity proxy. Anthropic’s last official round was $380B post-money in Feb. Run-rate revenue increased from $14B in Feb to >$30B by April, >1,000 customers now spend $1M+ annualized, and Claude Code already cleared the $2.5B run-rate revenue mark. The flaw is sentiment: the developer crowd is turning against the core coding product just as valuation goes vertical. X has been pushing “developers ditch Claude Code for Codex” narratives after Anthropic’s April quality regression/post-mortem, while user complaints cluster around usage limits, Pro-plan uncertainty, and users saying Codex is now better value or “Claude Code is no longer the king.”

Wednesday’s catalyst caught many by surprise. Anthropic signed for all of SpaceX Colossus 1—300MW+, 220k+ NVIDIA GPUs within the month—while doubling Claude Code’s 5-hour limits, removing Pro/Max peak-hour reductions, and raising Opus API limits.

Conclusion

This is a market where the macro looks fine until long yields say otherwise. AI capex is the growth engine, but the engine is concentrated. The opportunity is still there, but it is not broad. AI infrastructure has the strongest macro tailwind, while crypto remains a market of isolated winners rather than a rising tide.

Alea Research · 9 May 2026 · 22 min

Back to top