
Kog claims 30x faster LLM inference on existing Nvidia and AMD GPUs
The 11-person French startup hit 3,000 tokens per second on a 2B model and now needs to prove the same trick works on real LLMs by September.
Tag · 7 stories
Every story tagged Cerebras on AI Chat Daily.

The 11-person French startup hit 3,000 tokens per second on a 2B model and now needs to prove the same trick works on real LLMs by September.

The new preview mode runs at 14x standard speed, powered by a Cerebras partnership, and targets enterprise workflows where latency is the constraint.

The AI chipmaker beat Q1 estimates with $193M in revenue, but a 38–41% full-year margin forecast triggered the selloff.

Shares opened at $385 against a $185 IPO price, valuing the Nvidia challenger at $56.4B fully diluted.

The AI chipmaker is raising its target ahead of pricing as order books fill, pushing Cerebras toward a richer public-market debut.

The AI chip maker plans to raise about $3.5 billion, while a December loan converts into a multi-billion-dollar stake for OpenAI.

The Wafer-Scale Engine 3 maker is pricing 28 million shares at $115 to $125, with banks already fielding $10 billion in orders.
The tightly edited AI news email engineers, founders, and investors actually open. One email. Every weekday. Five minutes to finish.
The briefing read inside teams at