The day in one read

The artificial intelligence sector entered a period of intense consolidation and scrutiny this week, marked by the imminent public listing of Anthropic and a severe reputational crisis at NVIDIA. While Anthropic prepares to price its shares at a valuation approaching $1.4 trillion, capital markets are simultaneously reassessing the risks embedded in the AI supply chain, driven by allegations of sanctions violations involving NVIDIA’s CEO. This tension between explosive growth and regulatory fragility is further complicated by shifting political winds in the United States, where new state-level regulations threaten to slow the physical expansion of data centers, and by a broader debate over whether the massive capital expenditure on compute infrastructure will yield proportional economic returns.

The Anthropic Listing and Circular Financing

Anthropic is set to list on the Nasdaq next Thursday, targeting a valuation between $1.25 trillion and $1.4 trillion by pricing shares at $70 to $78. The company aims to raise between $57 billion and $90 billion, a figure that matches the scale of SpaceX’s recent record-breaking IPO. This capital raise is supplemented by a $26 billion private placement from industry partners including NVIDIA, Microsoft, Google, and Broadcom. The prospectus reveals a company with $81 billion in run-rate revenue for October, up $10 billion from the prior month, but also discloses a $104 billion net loss over the last nine months, largely attributed to non-cash mark-to-market adjustments on convertible notes.

A central point of contention among analysts is the structure of Anthropic’s compute commitments. The company has disclosed $901 billion in compute commitments against its annualized revenue, a ratio that has drawn criticism for its circularity. One observer noted that the structure resembles an individual with a modest salary signing contracts for ten times that amount. While underwriters may push the price higher to reflect a $1.5 trillion valuation due to strong demand, the reliance on its own suppliers for capital raises questions about the independence of its growth metrics. Simultaneously, Anthropic launched a productivity suite including Claude Sheets, Claude Whiteboards, and Claude Data, a direct competitive threat to Microsoft’s enterprise software dominance that contributed to a 2.2% drop in Microsoft stock.

NVIDIA’s Sanctions Scandal and Market Reaction

NVIDIA shares have fallen approximately 10% to 11% since the revelation that CEO Jensen Huang’s partner, Alice Wang, allegedly facilitated the export of $7.5 billion in chips to Chinese companies in violation of US sanctions. The scandal, involving a firm called Megaspeed, has triggered an FBI investigation and a class-action lawsuit alleging that NVIDIA concealed national security risks. President Trump acknowledged prior knowledge of the situation, and House Republicans are considering subpoenaing Huang. The market has reacted with significant volatility, with NVIDIA’s market cap shrinking from nearly $5 trillion to $4 trillion.

Analysts are divided on the long-term implications. Some argue that while the scandal undermines Huang’s personal credibility, it may not have lasting economic consequences for the chip supply chain. However, short sellers predict a further 30% drop if Huang resigns, viewing him as a key-person risk. The incident has also prompted Microsoft to terminate contracts with Oracle and N-Scale to reduce its dependency on NVIDIA hardware. Meanwhile, the broader AI chip market is showing signs of fragmentation, with Google developing eighth-generation TPUs and projecting its chip revenue to rise from $12 billion this year to over $100 billion by 2030.

Infrastructure Capex and Political Headwinds

The physical buildout of AI infrastructure is facing new regulatory friction following US midterm election results. Approximately 41% of planned US datacenter capacity is located in states where new governors support pauses or stops on permits. In Texas, where 248 projects totaling 200 gigawatts are planned, UBS estimates 18 gigawatts of capacity are at risk due to the state falling under Democratic control. This regulatory uncertainty may accelerate the peak of AI capital expenditure to 2027, creating a "protest march" against infrastructure expansion that will slow deployment timelines.

Despite these headwinds, demand for compute remains robust. CoreWeave reported quarterly revenue of $3.5 billion to $3.66 billion, nearly tripling year-over-year, and raised its full-year forecast to $13 billion. However, the company requires $3 in investment for every $1 in revenue, and its capacity is sold out through 2027. Mike Volpe, a former Cisco executive, argues that the current AI infrastructure buildout is fundamentally more valuable than the 1990s internet boom because it builds "intelligence" rather than just communication. He notes that NVIDIA Hopper H100 GPUs, originally expected to have a five-year life, now command rental rates of $250 to $300 per hour due to persistent scarcity in semiconductor manufacturing. Volpe predicts an economic correction among the 200 identified NeoClouds, but maintains that the underlying compute assets will retain long-term utility.

Tokenization and Institutional Crypto Adoption

Institutional adoption of tokenized assets is accelerating, with the New York Stock Exchange selecting Avalanche to settle tokenized stock trading. Securitize, a tokenization platform with a market cap of $1.3 billion, reported Q3 earnings that surprised the market, driving its stock up over 20%. The company is expanding into credit markets by acquiring lending platform River One and serves major clients like BlackRock. Other major exchanges are launching tokenized stock products; OKEx is launching tokenized stocks on its X Layer in partnership with ICE on November 23rd, including 67 tokens with 59 ETFs participating.

Bitcoin is trading near its 200-day moving average, with analysts debating whether the current pullback marks a bottom. Kyle Reedhead identified a potential support level at $71,000-$72,000, noting that while traditional equity markets experienced a capitulation event, Bitcoin has shown resilience with a slow, low-volume decline. The stablecoin market capitalization has declined from $304 billion to $288 billion, a drop analysts attribute to institutional players like Goldman Sachs and Citigroup preparing to launch their own dollar-backed stablecoins in 2027. This shift, combined with higher yields in traditional finance driven by AI-related debt, is diverting capital away from existing on-chain stablecoins.

AI Agents in Consumer and Enterprise Workflows

AI agents are expanding into diverse consumer touchpoints, from wearables to business discovery. Meta announced that its Muse AI agent is expanding to Apple Watch and Android’s Wear OS, allowing users to interact with the assistant via voice calls and text messages. Spotify upgraded its AI DJ feature, adding more voice options and a toggle to reduce commentary, catering to background music use cases. OpenAI rolled out a business hub to enhance ChatGPT as a consumer discovery engine, allowing companies to create verified business profiles and launch specialized agents.

In the enterprise and personal productivity space, the focus is shifting toward orchestration and persistent context. Ross Mike demonstrated a workflow using GrokBot as a central orchestrator for engineering, design, and finance, arguing it serves as a superior interface to the internet compared to direct model access. He noted that the Cloud Code $200 monthly subscription provides $8,000 to $12,000 worth of API token credits, making it financially irresponsible not to utilize this subsidy. Similarly, Scott Farrar developed a household management system using Claude Code to read and write local markdown files, creating a transparent, persistent memory layer for AI interactions. This approach allows for grounding AI responses in verified household data, such as appliance serial numbers and insurance policy details, eliminating the opacity of standard chat interfaces.

Also Notable

US core CPI rose 0.3% in October, exceeding the 0.2% forecast and triggering a 10-basis-point spike in 10-year Treasury yields to 4.99%. Fed Governor Christopher Waller suggested a 0.5% rate hike in December is possible. Hensoldt, a German defense sensor maker, targets €10 billion in revenue by 2035, with analysts divided on its valuation; Jefferies maintains a Buy rating with a €98 target, while MBB Research recommends Sell with a €70 target. SoftBank reported a $7 billion quarterly loss, its first in two years, driven by prepayments for US datacenter power plants and declining Intel holdings. China blocked SoftBank’s $5.5 billion acquisition of ABB’s robotics division, citing data security concerns. In the labor market, US white-collar jobs remain resilient, with management and scientific consulting adding 17,300 jobs, while document preparation roles fell by 17%, highlighting a phenomenon compared to the Solow Paradox where AI productivity gains are not yet translating into significant wage increases.