Skip to content

Pulse - June 6

6 Jun 202619 min
ContentsOverview

Bubbles aren’t necessarily a function of over-valuation. They are often the result of concentration, or being too late to an already way too crowded party. This edition is all about navigating a market built on one story. Expect analogies to past crises and trillion-dollar stories with thin margins of error. Indeed, the market is making highs with fewer stocks, richer multiples, and weaker households underneath.

Key Takeaways

  • BTC down (hard) while equities hold up; the market is paying for AI earnings, not generic risk.
  • Strategy’s 32 BTC sale was tiny in size and large in signal.
  • SPX keeps making highs, but breadth is poor and concentration is doing most of the work.
  • AI is the single point of failure: equity indices are levitating on a narrow cohort of AI-linked mega-caps.
  • Diversification is an illusion when the same buyer hides behind all major tickers.
  • Overall, 1929‑style concentration risk layered on 1907‑style shadow credit and 2008‑style stressed households.

Macro: The 1929, 1907, 2008 Market

If you’ve been reading our recent issues, you’ve probably noticed how often we come back to the gap between rising stock prices and falling consumer sentiment. 2026 combines a 1929-style valuation and concentration problem, a 1907-style shadow-credit channel, and a 2008-style consumer-credit fragility. AI capex, passive index flows, private credit, mega-cap tech, and crypto beta now share more of the same marginal buyer.

We’re on the verge of three possible $1 trillion IPOs: SpaceX, Anthropic, and OpenAI. Let’s go back to the 1920s analogy. Radio, electricity, autos, and aviation changed the economy. AI will inevitably have a similar, or even more accentuated effect. Bubble risk isn’t a claim that the tech is fake; it’s that markets are pricing in a utopia of perfect adoption and unlimited, cheap capital. Therefore, we are in a market that’s not expensive in a casual way; it is expensive in a mechanically fragile way.

The S&P 500 tech sector now represents more than 39% of index market cap, above the March 2000 internet-bubble level, while the AI-linked share of the index rises to more than half once Alphabet, Amazon, Meta, and other AI capex buyers are included. That is the 1929 echo: not just high prices, but high prices resting on a small group of winners.

The 2008 rhyme is in the personal savings rate, which fell to 2.6% in April, a four-year low and, excluding June 2022, represents the lowest since 2008. The New York Fed’s Q1 household-debt report put total household debt at $18.8T and aggregate delinquency at 4.8% of outstanding debt

Altogether, the equity market has a 1929 problem: valuation and concentration. The funding market has a 1907 problem: credit has migrated into less transparent nonbank channels. The consumer has a 2008 problem: cash-flow stress is rising under the surface while headline spending still looks fine. None of those forces guarantees a crash, but together they argue for lower exposure, faster exits, and less faith in “diversified” risk when the same AI/liquidity buyer owns the whole market.

Crypto: Under the AI Shadow

BTC — By Monday’s opening BTC was already trading ~72k, a level not visited since April. It dipped below $65k by Thursday. Topping it off was MicroStrategy’s sale of 32 BTC ($2.5M at an average price of $77,135). The problem for BTC, however, is marginal demand, not the 32 coins. Spot BTC ETFs have already been bleeding, just as stocks keep grinding higher. This divergence is getting harder to ignore. BTC really needs a strong catalyst to offset ETF redemptions, Strategy’s capital-stack needs, and Nasdaq’s AI-led pull away. Taken together, we are at a point where 50% of BTC in circulation is now held at an unrealized loss.

ETH — It opened the week near $2k and ended around $1.7k. The roadmap is extensive, but loaded with maintenance work that doesn’t bring much optimism. ETF demand is also messy: May had stronger flow narratives, yet June opened with outflows.

SOL — As May closed SOL ended its 8th straight red monthly candle. SOL opened in the low-80s and finished ~$70. App revenue stayed strong, ETF/institutional access is improving, and the chain continues to be a versatile bet for payments, trading, and RWAs. However, a network can look healthy and its token still dips when BTC breaks. If SOL cannot reclaim the high-$70s quickly, the market will treat $80 as resistance.

HYPE — Everything is lining up in their favor right now: new highs, deep perp liquidity, regulatory validation for the product category, and a listed equity wrapper that gives U.S. capital a way to chase the exposure. It’s gotten to a point where upside hinges more on the entry of the trade than on the thesis. This week Grayscale launched HYPG, a Nasdaq-listed Hyperliquid Staking ETF with a 0.29% fee, joining competing 21Shares and Bitwise products and pushing HYPE further into brokerage-account capital. The risk is crowding, now trading ~20% below all-time-highs.

ZEC — From narrative leader to supply scare. The token extended its selloff past 50% after the disclosure of a security vulnerability that propelled long liquidation cascades. The flaw was present in an Orchard zero-knowledge circuit, where an incomplete operation could potentially lead to double-spend. Orchard was temporarily disabled by soft fork at block 3,363,426 on June 2, then NU6.2 re-enabled the corrected circuit at block 3,364,600 on June 3. The issue was remediated and the ZCash Foundation and ZODL report no known exploitation.

The read-through for crypto is uglier than the Zcash-specific damage: one of the oldest protocols in the industry, with real cryptographers and a functioning disclosure process, still needed an emergency consensus change after an AI-assisted audit; most DeFi stacks add unaudited dependencies, admin keys, bridges, oracle assumptions, liquidation reflexivity, and recycled collateral on top of weaker response processes.

TON / GRAM — After announcing that Telegram will take over TON (instead of the TON Foundation), Durov is now changing the name of the coin, from TON to GRAM, while TON remains the blockchain name. GRAM was the original Telegram token the SEC blocked in 2020, forcing Telegram and TON Issuer to return more than $1.2B to investors and pay an $18.5M penalty. Bringing the name back sends a strong message: Telegram wants the market to price TON as its native money layer again, not as a foundation-led chain loosely attached to a messaging app.

XLM — Stellar opened the week up >70% after a short-squeeze. DTCC announced that its tokenization service would connect to Stellar; traders then chased XLM from roughly $0.15 to $0.224 while short liquidations outpaced long liquidations at $12.41M versus $6.82M since May 28. The stronger tell is that open interest rose 10.9% to about $361M as spot volume expanded. That move was fresh longs plus shorts getting run over. It worked while volume kept confirming.

NEAR — Last week we warned about the opportunity getting crowded, which materialized this week. Intents and onchain metrics still reflect very strong fundamentals, but price action is an independent variable. Price started retracing this week as traders took chips off the table because of BTC’s anemic move. Figures like Arthur Hayes, which were making rounds last week, were quick on their feet to dump the position as things got turbulent.

PENDLE — Pendle stands out as a clear winner of the CLARITY Act approval. The reason is that the protocol turns yield into a market instead of handing users passive APY for holding. The White House has targeted July 4 for digital-asset market-structure legislation and Senate Banking advanced CLARITY 15–9. Pendle benefits because, unlike issuers of yield-bearing assets, its yield comes from actual economic activity instead of passive holding. If the law pushes capital away from passive yield wrappers, Pendle is one of the few live venues where yield is priced, hedged, and traded.

ENA — After a relatively “boring” period absent of news, Coinbase is now providing a distribution tailwind. Coinbase Ventures will be backing the project through an open-market ENA purchasing program while Ethena and Coinbase co-develop an onchain savings product for Coinbase’s 100M+ users. Coinbase is already Ethena’s primary custodian, wallet provider, and perpetuals venue, and USDe yield exposure is expected to distribute through Base and the wider Coinbase ecosystem. As a result, Ethena no longer depends only on DeFi loops, Pendle trades, and CEX collateral programs; it gets a potential funnel into idle Coinbase stablecoin balances.

ONDO — ONDO is the same CLARITY trade at the infrastructure layer. If passive stablecoin APY gets squeezed, compliant tokenized T-bill and collateral rails should become more valuable, and Ondo already has scale. The catch is token value accrual. At roughly 84% below ATH, the market is pricing regulatory access risk and token-economics doubt.

EIGEN — The backdrop for unlocks is pretty hostile right now. 36.82M EIGEN unlocked on June 1, worth about $7.7M, split between investors at 19.75M tokens and early contributors at 17.07M tokens. That's nearly 4.97% of the float, part of a schedule that gives investors and contributors a 4% monthly unlock after the one-year cliff.

RAIL — The privacy narrative cooled fast after the squeeze. Tokens like RAIL, which aren’t listed on many CEXs, serve as a reminder of how fast price can snap back when liquidity thins. Good theme, and the project still has a strong roadmap, but most liquidity still lives onchain and reflexivity kicks both ways.

RUNE — Thorchain was once front and center; now it lives in the back pages. On May 15 the network experienced a signing exploit. According to the official post-mortem, a newly churned node operator entered two days before the attack and used a GG20 threshold-signature vulnerability to drain about $10.7M from one of five vaults. Trading and signing were halted and RUNE fell about 12%, not recovering much since. ADR028 proposes absorbing the loss through protocol-owned liquidity, with no new RUNE mint. That protects dilution. Trust is still damaged as negative sentiment prevails.

GNO — Gnosis Pay was hit through the Zodiac Delay Module; one week earlier, a vulnerable SquidRouterModule drained roughly $3.2M from 86 Safe wallets across Ethereum and Base in about two hours. The common thread is permissions wrapped in convenience. Smart accounts are becoming payment rails, but every helper module adds another place for a signer, delay, route, or policy check to fail. All of this hits hard because they’re the default multisig and wallet infrastructure provider, and the oldest name in the game. And they are still having security issues, just as DeFi security is back at the top of everyone’s worry list.

Prediction Markets — Polymarket did free marketing for Hyperliquid’s HIP-4: Strategy’s June 1 filing says it sold 32 BTC between May 26 and May 31 at an average $77,135, before the market’s May 31 cutoff; Polymarket’s own rule said the market resolves “Yes” if MicroStrategy sells any Bitcoin by the date in the title, yet the market moved into final review after two disputed “No” proposals because confirmation arrived on June 1, outside the market window.

Polymarket uses UMA’s optimistic oracle for dispute resolution, and it turns out that the most disputed Polymarket markets had more than half of UMA voting power coming from the ten largest wallets, with many active UMA voters tied to live Polymarket accounts. Hyperliquid’s HIP-4, in contrast, ensures outcome markets sit inside HyperCore, are fully collateralized, use the same order-book stack, and canonical offchain markets are deployed and settled by validators rather than a separate token-vote jury. That does not make HIP-4 incorruptible; validators can still approve bad markets or carry their own incentives. But they are more likely to be aligned with the protocol’s traders.

A UMA whale can win one ugly settlement and externalize the damage to Polymarket’s brand; a Hyperliquid validator that games outcomes attacks the exchange, the fee machine, the bridge trust, and HYPE itself.

Equities: An Index Balanced on Ten Names

SPX — The S&P 500 keeps printing records. It logged its 19th record close of 2026 by late May, close to one-fifth of trading days, and Goldman raised its year-end target to 8,000. However, the equal-weighted S&P 500 has lagged badly; the equal-weight/cap-weight ratio is near 1.1, close to the lowest since 2003 and below the 2008 low. Since February 2023, the S&P 500 is up about 81%, while the equal-weighted index is only about 34%, and the Nasdaq 100 about 151%. Top-ten concentration is already extreme. The ten largest S&P names represented almost 40% of the index in April and Nvidia alone 8.04%.

NDX — Nasdaq is the reason the crypto drawdown feels worse. QQQ opened Monday at $737.04 and moved sideways yet slightly up while BTC had a double-digit drawdown in the same window. That divergence shows that the market is still buying AI earnings and capex beneficiaries, instead of broad risk. The risk-on label is too lazy here as well. Nvidia, Microsoft, memory, software, AI PCs, and data-center supply chains have an earnings buyer; BTC has ETF outflows and a weakened treasury-buyer narrative.

IGV — To contextualize the Nasdaq-BTC divergence it’s helpful to use software ETFs as an example. IGV is up roughly 12% since May 14 while BTC is down about 10%, with the 20-day BTC/IGV correlation down to 0.58, a level last seen before the late-2023 and mid-2024 BTC rallies. IGV offers targeted exposure to North American software and interactive-media names. Software has reclaimed its 200-day as AI-disruption fear cools, while BTC is still below its 200-day. Either BTC catches up fast, or IGV’s recovery starts to wobble and kills the risk-on signal.

NVDA — Nvidia is the story underneath the story. The stock is roughly flat since Monday open, but it still represents the bid that’s sucking liquidity out of everything else. For them, the market is no longer only paying for hyperscaler GPUs. It is starting to pay for the whole AI compute path: cloud training, inference, memory, local agents, and developer workflow.

MSTR — Strategy now trades with its liabilities visible. The 32 BTC sale is immaterial against Strategy’s 800k-plus BTC hoard, but it funded cash obligations and represented the first sell order since 2022. This is a trade that’s increasing in complexity, no longer just tied to BTC’s price action, but polluted by other dynamics such as common issuance, preferred dividends, convert management, cash balances, and BTC-per-share math. Saylor can call small sales a “nothing burger,” but the stock owns the path dependency.

MicroStrategy is now facing its biggest unrealized loss in history, at -$10.8 billion. After 6 years of BTC purchases they are now down ~17% on their position. For comparison, the S&P500 is up ~116% over the same period. MSTR is down almost 80% from the highs.

ARM — Oracle Cloud Infrastructure joined Arm’s AGI CPU ecosystem, joining ByteDance. The bet is that more of AI’ s inference and agentic work shifts back toward CPUs. They have already doubled expected AGI CPU demand to $2B across FY2027–FY2028 and sees the product reaching $15B annual revenue in about five years. Customer concentration is the risk; shorts can still continue getting squeezed.

MU — Micron has crossed the Rubicon from a cyclical memory stock to an AI-related asset, trading above $1k and worth $1T+ in market cap after first joining the trillion-dollar club on May 26. That was when UBS raised its target to $1,625 and argued that AI demand, long-term supply agreements, and memory shortages have changed the earnings profile of the business.

HIVE — HIVE represents the kind of top-line growth miners need to stay relevant. The strategy consists in renting out dense GPU compute in data centers to AI and other heavy workloads (HPC, or High Performance Computing), alongside Bitcoin mining operations. FY2026 revenue rose 158% to $297.8M, with $278.3M from Bitcoin mining and $19.5M from HPC. It comes down to having access to cheap power, Paraguay scale, and GPU compute. That gives HIVE two ways to monetize data centers.

PURR — Looking for convex exposure to Hyperliquid, PURR (Hyperliquid Strategies) is where market structure gets interesting. Call demand for its options has been rising. There is heavy June call activity and shallow upside strikes. When traders buy calls in a rising stock, dealers often buy shares to stay hedged; if higher strikes get listed, traders can roll up and dealers have to buy more stock as price and volatility rise. That is the next 60-trading-day squeeze case. The hard stop is supply: PURR’s S-1 registered over 35M shares for resale, and the same option mechanics can work in reverse if HYPE stalls.

BOT — ROBOSTRATEGY — RoboStrategy is trying to run the MSTR playbook with robotics and physical AI as the underlying asset. For reference, BTC value per diluted MSTR share moved from roughly $2.34 when Strategy began buying Bitcoin in August 2020 to $164.71, or $129.53 net of credit and preferred shares, a roughly 55x move in 5.7 years.

BOT listed on Nasdaq on May 11 as a closed-end fund giving public investors exposure to private, pre-IPO, and public robotics names including Figure AI, Apptronik, Dyna Robotics, Standard Bots, and Dexmate; four days later it announced an optional $2B equity facility with Roth Principal Investments. The edge, if it exists, is the flywheel: issue stock at a premium, buy scarce private robotics assets, and let public-market demand pay for access that venture funds usually gate. The risk is the same machine in reverse. Closed-end funds can trade below NAV, private marks can lag reality, and the equity facility becomes dilution if management sells into a weak market.

Anthropic — The confidential filing caught prediction market traders by surprise. The company confidentially submitted a draft S-1 to the SEC four days after closing a $65B Series H at a $965B post-money valuation led by Altimeter, Dragoneer, Greenoaks, and Sequoia. The filing does not give investors share count, price range, or audited public numbers yet; the SEC’s confidential-review process lets issuers work through comments before the registration statement becomes public. The bull case is that Claude, Claude Code, enterprise adoption, and compute access can support a trillion-dollar public comp before OpenAI trades. The skeptical read is that, once the S-1 lands, investors will finally see how much gross revenue survives compute cost, reseller accounting, safety spend, and AWS/Google/SpaceX dependence.

Conclusion

Bubble or not, AI isn’t a growth theme anymore; it’s one crowded macro trade that now links equities and credit to a single set of overlapping marginal buyers. Index highs rest on a narrow AI mega-cap spine, funded by shadow credit and passive flows. Meanwhile, households and savings look like 2008 under the surface. It’s increasingly harder to find beta through diversification; one exposure (AI) bundles 1929‑level concentration, 1907‑style shadow funding, and 2008‑grade household fragility.

Alea Research · 6 Jun 2026 · 19 min

Back to top