The day in one read

The Capital Crunch and the AI Valuation Gap

The dominant narrative of the day is a sharp rotation away from artificial intelligence equities toward defensive sectors, driven by growing anxiety over the sustainability of AI capital expenditure. The Nasdaq fell 3% on Thursday, with the semiconductor index dropping 7% and CoreWeave losing more than 10% to sit 55% below its yearly high. This sell-off was triggered in part by Meta’s announcement that it plans to invest $135 billion to $145 billion in data centers and AI chips this year, a figure representing more than half of its revenue. The market’s reaction was immediate: Meta’s stock lost over 7% intraday, wiping out approximately $120 billion in market value, while credit default swap costs for the company hit record highs.

The underlying tension is a widening gap between infrastructure spending and revenue generation. Gartner forecast that 80% of neoclouds will go bankrupt by 2030, arguing that $3.5 trillion in AI infrastructure spending will be impaired due to a demand gap. The firm estimated that data center owners need $8.5 trillion in revenue to cover costs, while total enterprise and consumer AI revenue is forecasted at only $3.6 trillion. Citadel CEO Ken Griffin stated that the market is "almost certainly in the middle of a bubble," citing uncertain return on investment. Conversely, Blackstone executives argued that data center debt is backed by creditworthy hyperscalers, and NVIDIA CEO Jensen Huang noted on CNBC that capital scarcity, not demand, is currently limiting growth. This divergence highlights a market that is increasingly pricing in the risk that the current infrastructure buildout may outpace the economic value it generates.

Hyperscaler Capex and Infrastructure Bottlenecks

Major technology firms are continuing to pour unprecedented capital into AI infrastructure, yet they face physical and financial constraints that are reshaping their strategies. Amazon CEO Andy Jassy stated that the company expects to spend approximately $220 billion in capital expenditure this year, noting that even this amount leaves meaningful AI demand unmet. He reported that AWS grew 36.7% year-over-year to a $169 billion annual revenue run rate, with the AI segment alone exceeding $25 billion in run rate. Jassy emphasized that 2026 capacity is sold out, 2027 is largely sold out, and a significant portion of 2028 is already committed. He argued that inference costs remain an order of magnitude too high for pervasive enterprise adoption, a gap Amazon aims to close through its Annapurna Labs chip division, which developed the Graviton CPU and Trainium AI accelerator.

Meta’s strategy is similarly aggressive but faces different scrutiny. The company reported Q3 revenue of $64.6 billion, up 26%, but its Q3 capital expenditure was $35 billion, reducing year-to-date free cash flow from $30 billion to $11 billion. Meta announced a two-for-one stock split creating Class C shares to fund AI spending and maintain Mark Zuckerberg’s control. Scott Galloway predicts Meta will be the winner of the AI race due to its distribution of 3 billion daily active users, arguing it offers the best ROI on AI investment compared to competitors. In contrast, Tesla beat Q3 expectations with $28 billion in revenue, though core automotive revenue fell 3% and net income dropped 29% to $787 million due to increased spending on AI and Optimus robots. A special committee has been formed to explore a SpaceX-Tesla merger, which analysts view as a narrative play to justify Tesla’s 330x earnings multiple, compared to the automotive industry average of 18x.

The Anthropic IPO and Market Liquidity

Anthropic is reportedly preparing to file for an IPO within days, with a private valuation of $965 billion and potential public market valuations of $2 trillion. This listing would be a significant test of market liquidity, as the company reported an $80 billion accounting loss and an $8.7 billion cash loss in the first half. The IPO is expected to require $50-100 billion in capital, potentially absorbing a significant portion of available market liquidity. Spark Capital’s 2023 investment of $75 million at a $5 billion valuation is now worth approximately $7 billion, reflecting the massive appreciation in the company’s value.

The market reaction to Anthropic’s potential listing is mixed. Some investors view the company as a cornerstone of the AI economy, while others worry that the sheer size of the offering could exacerbate the current sell-off. Anthropic’s IPO valuation has softened from $2 trillion to $1.5 trillion in recent rumors, with OpenAI’s fundraising expected to close at $1.25 trillion. The company is also rumored to be in talks for a $45 billion contract with N-Scale, while NVIDIA is rumored to invest hundreds of millions into Firmus. These moves suggest that the largest AI players are consolidating their supply chains and securing long-term capacity, even as the broader market questions the sustainability of the current spending cycle.

European Sovereignty and Regulatory Pressures

European policymakers and tech leaders are increasingly focused on digital sovereignty, driven by concerns over dependency on US-controlled technology and the risks of AI-driven job displacement. During the Microsoft Frontier Tour in Köln, Microsoft CEO Satya Nadella emphasized the company's readiness to partner with German industry and politics, announcing that Microsoft’s platform now offers over 11,000 models for selection, including options from both OpenAI and Anthropic. Microsoft is also launching specific "Sovereignty" products and tools targeted at the European market. However, critics argue that true sovereignty is unattainable through hyperscaler dependency, noting that Germany lacks the chips, models, energy, and capital for independent AI development.

Benjamin Balde, head of Control AI’s German activities, argues that artificial superintelligence must be legally banned because it exceeds human control capabilities. He notes that current AI already outperforms most humans in coding and solves complex mathematical problems, with Anthropic reporting that AI performs up to 26% of its own research. Balde asserts that the path to superintelligence involves automating AI research, where current models assist in writing 80-90% of the code for next-generation models. He warns that existing containment measures are insufficient and advocates for a global ban on general-domain autonomous self-improvement. Control AI is funded primarily by Jan Thalen, the Skype founder, and operates in the US, Canada, UK, and Germany. Balde urges citizens to contact their representatives to push for regulation before irreversible societal changes occur.

Security Vulnerabilities and AI-Driven Exploitation

Critical security vulnerabilities in PHP and Node.js have triggered widespread infrastructure instability, with Cloudflare reporting a surge from 1,000 to millions of malicious requests within 10 hours of disclosure. The PHP flaw, identified by the security firm Calif, was known to maintainers for two months and fixed in collaboration with OpenAI. The vulnerability allows remote code execution without touching user code, though exploitation requires specific configurations like display_errors enabled and low memory limits. AI tools have accelerated exploit development, reducing the time needed to find and test payloads. This incident highlights the growing risk of AI-accelerated exploitation of legacy vulnerabilities, a concern that is driving European efforts to develop sovereign AI infrastructure.

The Ethereum GlamStem update was also delayed due to security concerns found by Chain Security. In the AI sector, CoreWeave’s stock dropped 25-30% amid fears over its debt structure, though its credits are collateralized by individual data centers. AWS reported being fully booked through 2028, with power as the primary constraint rather than GPU availability. The discussion highlighted the "AI bubble" concerns, noting that the current security landscape is becoming more volatile due to the coexistence of legacy software vulnerabilities and AI-generated code. The Starlink incident in Ukraine, where service was restricted and then restored, was cited as a warning about dependency on private tech providers for critical infrastructure.

Also Notable

Bitcoin trades in the low $80,000s, holding strength despite macro headwinds, with analysts noting a decoupling from the AI trade. Solana launched the Alpenglow upgrade, reducing block finality from 12 seconds to 200 milliseconds and cutting validator costs by 70%. In China, DeepSeek raised approximately $15 billion, two days after the U.S. ordered federal civilian agencies to remove its AI technology over security concerns. Spotify reported 788 million monthly active users and nearly 5 billion euros in revenue for the third quarter, with Co-CEO Gustav Söderström emphasizing the company’s strategy to become an "intelligent media service encapsulated as an agent." Moderna shares plunged 17% after a study showed its cancer vaccine reduced recurrence or death risk by 42% in combination with Merck’s Keytruda, a result that disappointed investors who had priced in higher efficacy. Insomnia Cookies founder Seth Berkowitz outlined the company’s expansion strategy, noting the chain will finish the year with roughly 393 bakeries globally, with a long-term target of 1,800 locations.