Making Sense of This Week’s AI Market Crashes | #307

Alphabet’s 13% drop on Thursday erased about $500 billion in market value even as revenue rose 25% to $127 billion, Google Cloud grew 100%, and the Gemini app reached one billion users. Investors focused on slower search-ad growth and the cost of AI investment: search ads grew 12%, down from the prior quarter’s pace, Alphabet reported a $9 billion quarterly free-cash-flow burn, and it maintained annual capital-expenditure guidance of $195 billion to $205 billion. The selloff underscored the market’s concern that impressive AI progress may not quickly offset pressure on high-margin businesses or the cash demands of building capacity.

Google’s Gemini for Carbon was presented as a leading cyber model, with reported scores of 59 on the Artificial Analysis Intelligence Index, 88% vulnerability detection on SEC Bench Pro, and working patches for 71% of flaws. But access is restricted to trusted cyber defenders through Fairwind, and Alex cautioned that comparisons with Opus 5.5 are limited by availability, cost, and unknown margins or subsidies. Google’s Genie 4, an interactive world model with simulated physics and native audio, could support robot training and industrial planning, but it was still being rolled out to testers. The company’s AI lead therefore remains a possibility, not a settled commercial advantage.

The hosts linked pressure on Alphabet to a broader shift in technology economics. Emad forecast that search advertising could peak as AI agents change how people find and buy products, while another speaker said cloud could become Alphabet’s largest business within a year or two. That transition carries a margin risk: search ads are more profitable than Google Cloud, and cloud expansion depends on constrained chip supply. The discussion also cited limited high-bandwidth memory and rising RAM prices as constraints on data centers and consumer devices. Dave warned that the buildout may worsen market conditions before it pays off, particularly if companies cannot earn enough to sustain their investment.

AMD and Intel shares also fell; Intel was reported down 28% after beating earnings estimates but giving a weaker-than-expected outlook. Dave said major technology companies are betting that control of chips or algorithms will shape the future, with Meta borrowing tens of billions to fund AI investment. He warned that AI firms’ current margins may not last: Anthropic and Nvidia were cited at 80–85%, and TSMC at 50%. If AI companies move further toward vertical integration, competition could push margins down and threaten assumptions behind chip-sector growth. Chinese developers, meanwhile, were described as making efficient, lower-cost models through optimization, distillation, and engineering under tighter compute constraints.

The crypto discussion framed another kind of market risk. Chamath reportedly warned that Bitcoin’s capped supply and limited buyer pool leave it vulnerable to hacking and, eventually, a collapse to zero, potentially accelerated by AI and quantum computing. Salim called the reversal a “PSYOP,” noting Chamath had previously predicted a $1 million Bitcoin, and argued that cryptography can adapt. Alex said the central challenge may be reaching agreement on an upgrade, rather than the inability to strengthen encryption. Eamon expected more sophisticated attacks but said Bitcoin could survive as a historic asset and upgrade, with about half its hash rate able to do so.

A reported $6.3 billion Binance theft illustrated the risks of human and technical security: attackers allegedly compromised an engineer’s machine, and North Korea was accused. The speakers said the attack took about 20 minutes and roughly $3 billion had been frozen; much of the stolen value was reportedly in Tether and Circle stablecoins, which their issuers could freeze, while transfers remained traceable. Dave described markets as down about 3–3.5% that day and said volatility could continue as announcements, Chinese models, and benchmark results change investor expectations. He still characterized the economy as growing overall and advised diversification rather than avoiding AI exposure.