MARKET UPDATE: Anthropic S-1, Market Bottoms, What’s Next?

Anthropic’s S-1 disclosed $132 billion in cash against more than $500 billion in forward reported compute commitments. One speaker said the company is losing money and estimated it would need to raise about $80 billion to continue, making the commitments dependent on exponential growth or outside support; they called the outcome a coin flip. But because expectations for the filing had become so negative, the speaker thought much of the risk was already priced in. After the release, Intel rose from about $85 to $94, Sandisk gained 6%, and the Nasdaq advanced roughly 2%, a move they attributed to investors having de-risked in advance and short positions being caught out.

The speakers saw the market as broadly sound and the AI trade as still in its first or second inning, while cautioning that a hot bull market can bring sharp volatility. One expects a possible second leg in AI, perhaps after the midterms, and said they were adding to Intel. They expect undercapitalized neoclouds to consolidate or be absorbed by hyperscalers, and warned that slower Anthropic growth could pressure those companies. Their preferred approach was to avoid neoclouds they consider uncreditworthy, buy Micron on dips, and add to QQQ or SPY on market pullbacks. They cited SK Hynix as sold out through 2030 and expected DRAM shortages to persist.

An Anthropic growth plateau was framed as a possible buying opportunity, not necessarily the end of the AI trade. One speaker estimated that Anthropic fulfilling even half its commitments, together with Meta’s announced commitments, would put about $200 billion of equity capital into the system. They argued losses for some neocloud investors could help fund Meta’s AI expansion and subsidized models or compute, potentially benefiting Meta stock. The midterms were identified as a major market uncertainty, with fears of a Democratic sweep and politicians targeting AI contributing to de-risking. The speaker expected a rally under either election outcome, potentially larger after a Republican win, and viewed an initial selloff after a Democratic win as a buying opportunity, though algorithms could amplify selling.

The political outlook remained uncertain: one speaker put the odds of either party winning the midterms at 50% and estimated a 21% chance of a radically left-of-mainstream Democratic nominee, or about a 10% overall chance of that outcome. They regarded Newsom, Ossoff, and Shapiro as business as usual for markets, but saw AOC or DSA-aligned candidates as a major risk; another speaker cautioned that midterm results would not make a later Democratic victory inevitable. Speakers also disagreed over affordability: one emphasized tariff-driven price increases and public dissatisfaction, while another considered tariffs a smaller factor and said conditions were improving. Housing supply in major cities was identified as especially important for people earning roughly $80,000–$200,000.

One speaker saw political and regulatory risk as greater for crypto than for AI, arguing that a Democratic administration could more readily tighten rules for Coinbase, wallets, and other crypto businesses than halt AI progress. They suggested shorting crypto or buying puts, especially on Ethereum, and considered layer-1 networks such as Solana more exposed than Bitcoin. The same speaker predicted Bitcoin could reach $1 million within 10 years, citing continued liquidity creation and currency debasement. They argued that Bitcoin and major AI trades are global bellwethers, whereas smaller crypto and AI-adjacent projects are more vulnerable to U.S. political shifts. They doubted AI agents need tokens or blockchain rails, pointing to Stripe accounts and existing payment networks, and expected Bitcoin to be the only necessary token as others trend toward zero.

The speaker’s long-term portfolio remained biotech, ARK Genomic Revolution ETF, Intel, QQQ, and the S&P 500; the ARK fund was up about 5–6% that day and moving toward new highs. They were watching uranium but did not own it, citing potential data-center demand and broader European acceptance of nuclear as clean energy. Speakers differed on timing: one said nuclear’s seven-to-ten-year buildout could not meet near-term needs, while another favored modular reactors and noted China was pursuing them. Natural gas was described as quick and inexpensive to build near pipelines, but Jona doubted political support for highly polluting plants; they allowed that urgent shortages might prompt officials to cut red tape for modular gas turbines. Jona predicted data-center-driven brownouts could lead an AOC-led government to ban AI, while governments led by anyone right of the DSA would build power plants regardless of party.