Capital & Industrial Strategy

Anthropic wrapped up a secondary share sale with limited employee participation, according to Bloomberg. Some investors were unable to acquire their planned allocation because employees sold fewer shares than anticipated, suggesting confidence in future valuation despite current regulatory challenges. Separately, a federal appeals court declined to pause the Pentagon's designation of Anthropic as a supply-chain risk, even as a California lower court has blocked broader government restrictions on the company's technology, leaving the legal landscape fragmented across jurisdictions per Wired and Politico.

The company is simultaneously pushing into enterprise with new managed agent services designed to lower technical barriers for business deployment, as reported by Wired. Anthropic also disclosed that its Mythos model is too powerful for public release after exposing thousands of unpatched software vulnerabilities, according to Semafor, though the company is providing early access to select tech firms for cybersecurity preparation. Reuters notes Anthropic may have narrowed the revenue gap with OpenAI, which has implications for both companies' eventual public market valuations.

Meta released Muse Spark, the first major model from its restructured AI organization led by chief AI officer Alexandr Wang, who was brought in as part of a multibillion-dollar overhaul following the disappointing reception of the company's previous model release over a year ago, according to Bloomberg, CNBC, and the Financial Times. The model is described as purpose-built for social media applications, directly addressing investor skepticism about Meta's massive AI capital expenditure. Wired reports that benchmarks suggest formidable performance, positioning Meta more competitively against OpenAI and Google.

The release comes as Meta is simultaneously financing a $3 billion data center campus in Ohio through syndicated loans arranged by Natixis, MUFG, and Societe Generale, per Bloomberg. This infrastructure investment signals Meta's commitment to scaling compute capacity even as the company faces investor pressure to demonstrate returns on AI spending. The Wall Street Journal notes this is a major test of the company's AI ambitions following the expensive restructuring.

OpenAI CFO Sarah Friar confirmed the company will allocate IPO shares to retail investors, while CRO Denise Dresser disclosed that enterprise revenue now represents 40% of total and is projected to equal consumer revenue by the end of 2026, according to CNBC. This marks a significant strategic shift toward B2B monetization ahead of the company's anticipated public markets debut. The retail allocation decision is notable given typical tech IPO structures favor institutional investors, suggesting OpenAI is attempting to cultivate a broad shareholder base similar to consumer-facing technology companies.

Semafor argues that recent concerns about OpenAI's growth trajectory are overstated, though the article does not provide specific financial metrics. The shift in revenue mix toward enterprise customers likely reflects OpenAI's efforts to build more predictable, contract-based recurring revenue ahead of increased public market scrutiny.

Banks are syndicating $3 billion in loans for Meta's Prometheus data center in Ohio, demonstrating continued institutional appetite for AI infrastructure debt financing even as the Iran conflict creates broader geopolitical uncertainty, per Bloomberg. Goldman Sachs Asset Management is advising clients to increase exposure to semiconductor companies and other infrastructure providers, arguing the AI capital expenditure cycle will continue to accelerate despite tensions, according to Bloomberg.

Separately, CNBC reports that advanced packaging at TSMC has emerged as the next critical bottleneck for AI chips, with Nvidia having reserved the majority of TSMC's most advanced packaging capacity. This post-fabrication step requires chips—including those manufactured in the U.S.—to make a round trip to Taiwan, highlighting persistent supply chain dependencies despite domestic semiconductor production efforts. Meanwhile, G42 in the UAE is proceeding with data center construction for OpenAI despite regional infrastructure attacks, according to Bloomberg, and Alibaba launched a 10,000-chip data center in China with domestically produced processors per CNBC, signaling parallel infrastructure buildouts across geopolitical blocs.