Goodfire raises $150M at $1.25B to decode AI black boxes
What's the deal? San Francisco-based AI research lab Goodfire has raised $150M in Series B funding at a $1.25B valuation, less than a year after its Series A round. B CapitalDealroom has a profile for this one. Try Dealroom →led the round, with participation from new investors DFJ Growth, Salesforce Ventures and Eric Schmidt,Dealroom has a profile for this one. Try Dealroom → plus existing backers Juniper VenturesDealroom has a profile for this one. Try Dealroom →, Menlo Ventures, Lightspeed Venture Partners, South Park Commonsand Wing Venture Capital.Dealroom has a profile for this one. Try Dealroom → The company uses interpretability — the science of understanding how neural networks work internally — to design AI models that can be debugged and steered with precision.
Why now? Foundation models are being deployed at unprecedented speed across production code, internet applications and even military systems, yet their internal decision-making remains opaque. Goodfire argues that the current black-box approach to building AI is fundamentally broken, with teams unable to understand why their models behave as they do. The company recently demonstrated practical applications of its approach by reducing hallucinations in a large language model by half and identifying a novel class of Alzheimer's biomarkers — the first major finding in natural sciences obtained from reverse-engineering a foundation model.
What could go wrong? The company is tackling what some would call "impossible ambition", attempting to solve what it considers the most important technical problem of this generation. Interpretability research remains an emerging field, and Goodfire's approach requires fundamental breakthroughs in understanding neural network internals. The company's model design environment must prove it can scale beyond research demonstrations to production systems.
The signal: Goodfire represents a new wave of "neolabs" — AI companies pursuing research breakthroughs in training models that scaling-focused labs like OpenAI and Google DeepMindhave neglected. The company's rapid valuation growth and ability to attract talent from frontier labs signals growing recognition that understanding AI systems may be as important as making them more powerful. With partnerships spanning Mayo ClinicDealroom has a profile for this one. Try Dealroom →, Arc InstituteDealroom has a profile for this one. Try Dealroom → and Microsoft, interpretability is moving from academic curiosity to commercial necessity.
Sources:
PR Newswire
Goodfire
Business Wire
Investing.com
B Capital
The AI Journal
Bloomberg
B.S.
