Google on Monday unveiled two new variants of its in-house artificial-intelligence chips. The push comes just as market leader Nvidia is grappling with an investigation into reports that its CEO Jensen Huang was targeted by a foreign intelligence effort.

Google, which is owned by Alphabet, said it has expanded the eighth-generation lineup of its Tensor Processing Units with two new systems, the TPU 8tx and TPU 8i Max.

The search company unveiled the initial TPU 8 chips -- designed in collaboration with Broadcom -- earlier this year, for training AI models and inference, or running those models. Now it is making more specialized versions.

Alphabet said the TPU 8tx was targeted at businesses and governments running AI with sensitive data and was optimized for "secure and confidential inference." Due to its hardware design, developers can fine-tune models and train AI agents without needing to move data between separate computing clusters.

Meanwhile, the TPU 8i Max is designed to overcome the so-called "memory wall" in running large AI models, where there is a gap between how quickly an AI chip can do calculations and a memory chip can feed in data, leading to inefficient processing.

"TPU 8i Max scales compute and memory in lockstep...so processors remain equipped with essential data and so that large workloads do not stall," Google said.

The increased range comes just as Google markets its hardware more fiercely as an alternative to Nvidia's. The company recently announced TPUs would be made available through software-and-cloud computing company Oracle's cloud infrastructure platform from 2027. This is the first time TPUs will be offered on a so-called hyperscale cloud platform outside of Google's own.

"Our aim is to make TPUs more easily accessible to every developer and Cloud customer," Google said. "By combining modular custom silicon with an open and easy-to-use software stack, we can deliver new hardware and system configurations at the pace of the AI industry."

That's hard to see as anything other than a direct challenge to Nvidia, which has marketed its graphics-processing units as the best choice for AI workloads. Nvidia's big advantage has generally been the familiarity AI developers have with its CUDA software, as well as their general-purpose flexibility.

Alphabet stock was up 3.6% in early trading while Nvidia shares were down around 7% as the latter company was hit by concerns over CEO Huang's future amid an investigation into his romantic relationship with an executive suspected of working for Chinese intelligence.

The announcement comes as Nvidia faces scrutiny over Huang -- though Google and Nvidia are generally friendly rivals. But the AI hardware race is moving at a rapid pace and any distraction could prove costly.