Frontier AI race widens beyond OpenAI and Anthropic

Frontier AI competition has shifted from a perceived OpenAI-Anthropic duopoly into a fluid oligopoly. Recent releases from SpaceX’s Grok 4.5 and Meta’s Muse Spark 1.1 put both companies back in contention, while Alphabet’s expected Gemini 3.5 Pro launch will test whether its vertically integrated strategy and TPU advantage remain competitive. OpenAI is positioned around Codex momentum and consumer distribution, while Anthropic’s revenue run-rate is estimated to be over $60B but faces trust and ecosystem risks.

The reopened frontier has not made model building broadly accessible. Alphabet, Meta, and SpaceX can fund repeated attempts through existing businesses, while Chinese labs have raised large sums and pursued open-source distribution to gain adoption. That strategy has brought visibility but weaker monetization, since inference providers capture part of the economics and Western customers may avoid Chinese API endpoints.

Long-term outcomes hinge on whether any lab reaches recursive self-improvement, using AI to accelerate AI research and pull capital and talent toward a single leader. Without that breakthrough, leadership is likely to rotate and labs will need conventional moats in distribution, workflows, cost, and customer ownership. For investors, broad exposure to AI capex suppliers looks less reliable as capex approaches the $1T scale, making bottleneck durability, pricing power, and revenue growth more important.