AI that matters, from the architect's desk. Curated and engineered by Saaket Varma, PhD — no hype, just signal.
HOW TO READ THIS Read top to bottom: Baseten seeks funding, the round and valuation target, the inference-over-training fork, then the live-serving payoff.
AI inference startup Baseten is reportedly raising $1.5B at a $13B valuation, only months after its last mega-round. The signal isn't the number — it's where the capital is pointing: serving models at scale is now a bigger business than training them. As frontier labs commoditize the models themselves, margin moves to whoever runs them cheapest and fastest. If you're building on AI, treat inference cost and latency as a first-class architecture decision now, not a post-launch optimization.
HOW TO READ THIS Read top to bottom: Jio builds AI into its network, one layer fans out to every phone, reaching 500M+ people.
Reliance is wiring AI into telecom for 500M+ users — a preview of AI as default infrastructure, not a feature.
HOW TO READ THIS Read top to bottom: Snap housed the unit, then it exited, pushed out by mounting compute costs, becoming standalone Dotmo.
Even an audience-rich company can't carry frontier video compute alone — a candid read on how expensive AI video has gotten.
HOW TO READ THIS Read top to bottom: Elastic buys the startup, the deal size, how its agent scans and fixes code, then its place inside Elastic.
Elastic's up-to-$85M acquisition is another marker that debugging is quickly becoming an agent's job.
High-throughput, memory-efficient inference and serving engine for LLMs.
Fast serving framework for LLMs and vision-language models.
Production toolkit for deploying and serving large language models at scale.