Are we at the crossroads of the AI compute infra, it being the major dictator of how AI will progress and find its way in the days to come. Sam Altman’s OpenAI is on track to bring “well over 1 million GPUs online” by the end of the year. Whatever may be the nature, quantum and extensive utility of the model, the fact of the matter is, there is a model in town which needs 8 Nvidia H20 chips to operate. Additionally what we are talking about is the product being open source.
China has defined and redefined technologies in a variety of areas and in a variety of ways even within it. Just think of high speed trains and you can visualise how different geographies are a world apart. And when it comes to scale there, none can compare with China. The AI tech and the business model created by the US for the world was running fine till China upset the apple cart. The AI challenger is born and as expected it’s China. The DeepSeek R1 model released in Jan 2025 marks a watershed moment in the history of AI.
Nvidia’s stocks declined by 17-18 percent erasing roughly $600 billion in the market capitalization in a single day – its largest ever single day market value loss. Now one more from the Chinese tech stables. On Monday, a new bombshell on the tech world came in the form of Startup.ai, formerly known as Zhipu, claimed that its new GLM-4.5 said it would cost less than DeepSeek to use. Z.ai said its new GLM – 4.5 is built on Agentic AI, meaning that the model would automatically break down a task into sub- tasks in order to complete it more accurately.
Announced on July 28th, 2025, GLM – 4.5 has intensified global competition,pushing boundaries of cost, efficiency and performance in the rapidly evolving AI landscape. GLM – 4.5 stands out for its aggressive pricing structure, charging just $0.11 per million input tokens and 0.28 per million output tokens. Z.ai’s model is open source under the Apache 2.0 licence – free for developers – and engineered on just eight Nvidia H20 chips, roughly half the hardware DeepSeek needs, making it especially appealing for scale-ups and startups. No-one knows how AI will fan out in the days to come. Ironically, based on business outcomes, projections and patterns are being worked out.
OPEN INNOVATION DOESN’T JUST LOWER THE COST OF INTELLIGENCE, IT DRAMATICALLY RAISES THE STAKES FOR EVERYONE.
Sanjay Sahay
Have a nice evening.