The AI Chip War Is Moving Inside the Cloud
Google, AWS, Microsoft, Meta, and OpenAI are building their own AI chips. The goal is not to replace NVIDIA everywhere — it is to control the cost of every AI task.
NVIDIA generated $75.2 billion in Data Center revenue in a single quarter, up 92% from a year earlier. Its non-GAAP gross margin was about 75%. Those numbers show just how powerful NVIDIA’s position has become.
But NVIDIA’s biggest customers are also building alternatives.
Google is now on its eighth generation of TPUs. Amazon is scaling Trainium. Meta plans four new generations of its MTIA chips within two years. Microsoft already launched Maia 200, and a new report says Maia 300 could arrive as soon as September. OpenAI has also unveiled its first custom accelerator.
This does not mean NVIDIA is about to lose the AI market.
The more important question is different:
As AI usage explodes, who controls the cost of running it?
That is what the custom-chip race is really about.
Smarter insights on money, markets, business, and the trends shaping what comes next.
Get:
3 premium briefings every week
Deep market & business analysis
Early signals on emerging opportunities
Full access to premium PulseLine posts
$10/month or $100/year
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.