Google takes on Nvidia with specialised AI chips, Bloomberg reports

What Bloomberg reported

On 9 November 2026, Bloomberg reported that Google is intensifying its competition with Nvidia by releasing specialised versions of its artificial intelligence chips ^S1. The report frames the step as an escalation in a contest between the two companies over the hardware that runs AI models ^S1.

The published excerpt is a headline only. It does not name the chip families, describe what the specialisations are, or state prices, volumes, customers or launch dates ^S1. Those details would be needed to judge how serious a challenge the move represents.

Why chips are the money story

Large AI systems are trained and run on silicon built to perform enormous numbers of simple calculations at once. In plain English, the chip is the engine room and the model is the software that uses it.

Buying that hardware is one of the biggest costs in building AI. A "specialised" chip generally means one tuned for a narrower job, such as running a trained model rather than training a new one, in exchange for lower running costs or better performance on that specific task. Generality and efficiency trade off against each other: the more narrowly a chip is designed, the less easily it moves to other work.

For the company selling those chips, revenue depends on convincing buyers that the narrow job is the one they will do most often. For the company buying them, it depends on whether the savings outweigh the risk of depending on a single supplier.

Why the rivalry framing matters

The framing of the report is itself the signal: two large technology companies are being placed in direct competition over AI silicon ^S1. When a company with both a cloud business and an AI research operation brings out chips tuned to particular workloads, the question for anyone holding technology stocks is whether those buyers shift budget away from the incumbent supplier ^S1.

The report supplies no figures that would settle that question ^S1. It does not say who will sell the specialised chips, whether they will be offered through Google's cloud platform, or whether outside customers can buy them at all.

What the report leaves out

A single headline cannot tell a reader the size of the investment, the manufacturing partner, the benchmark performance, the price per chip, or any customer commitments ^S1. Without those, claims about winners and losers are guesswork.

What to watch

• A fuller version of the Bloomberg report, with specifications and pricing.
• Whether the chips are sold as hardware or rented through cloud services.
• Any public response from Nvidia.
• Comment from large AI buyers on their supplier plans.