Big Tech keeps spending billions on AI data centers and chips

In short: There is no clear sign of an AI infrastructure “glut”, and the biggest tech companies are still pouring money into chips, cloud servers, and data centers.

What's going on

Some commentary says companies are scrambling because there is too much AI. But the reporting and industry numbers point the other way. Demand for the basics that power AI is still rising, especially computing power (the rented “engine room” that runs AI) and the buildings that hold it, called data centers.

Big cloud providers are leading the spending. Microsoft has talked about roughly $80 billion in planned spending, Alphabet up to $75 billion, Meta around $60 to $65 billion, and Amazon about $100 billion. These companies control much of the cloud infrastructure that AI makers rely on to train and run their systems, similar to how most websites rely on a few big hosting companies.

That demand is also pulling in more specialized suppliers. Chip makers like Nvidia, AMD, Intel, Cerebras, Groq, and Tenstorrent sell the hardware many AI systems need. Newer “AI-first” cloud companies, including CoreWeave, Lambda Labs, Together AI, and Voltage Park, offer on demand access to large pools of graphics chips, which are commonly used for AI.

Surveys cited in the reporting suggest the pressure is still growing. One set of figures says 96% of organizations plan to expand their AI computing power, and 64% already use generative AI (tools that create text, images, or code).

What to watch

The risk is not oversupply so much as overspending. If companies do not earn back what they invest, budgets could tighten and layoffs could continue. Still, the biggest cloud providers can often profit either way, because they charge ongoing fees for the computing time that other companies rent.

Source: NYTimes