Cognition eyes $40B valuation in new funding round

AI inference startups that have built business models by leasing compute from cloud providers and reselling it are now facing capacity constraints, forcing some to purchase their own chips. This shift threatens the asset-light economics that made these startups attractive. The move also reflects broader pressure on compute availability across major hyperscalers, pushing some startups toward smaller cloud providers, which introduces new operational and security risks.

Google DeepMind has released AlphaGenome 2, an AI model that predicts how DNA changes affect different cell types, advancing understanding of disease mechanisms at the cellular level. The model builds on last year's AlphaGenome by providing more detailed insights into gene expression patterns across diverse cell types. This capability could help researchers identify disease causes and develop targeted treatments.

OpenAI reported third-quarter revenue of $10 billion, up nearly 50% from the prior period, driven by demand for newer models and enterprise products. Operating losses widened to $10.9 billion from $10.1 billion, though the company showed margin improvement on an adjusted basis, moving from negative 150% to negative 109%. The results underscore OpenAI's continued struggle to achieve profitability despite strong top-line growth.

PSI, a startup focused on AI-driven scientific discovery, used an unreleased Anthropic model to solve a fundamental quantum computing question that researchers have pursued for nearly two decades. The achievement demonstrates the potential for large language models to accelerate scientific breakthroughs in specialized domains. The work signals both the capabilities of advanced AI systems and the emerging business model of applying them to hard scientific problems.