Google unveiled two specialized Tensor Processing Units designed to power the emerging wave of AI agents. The chips split inference and training workloads across dedicated hardware.
Google introduced its latest generation of TPUs, positioning them specifically for what the company calls the "agentic era"—a shift toward autonomous AI systems that can perform complex tasks with minimal human intervention.
The dual-chip approach marks a departure from previous TPU designs. One chip handles inference, executing trained models at scale, while the second specializes in training, optimizing model development and fine-tuning. This separation allows each processor to be tailored for its specific workload, potentially improving efficiency and performance.
The timing aligns with Google's broader push into AI-driven enterprise tools. The company simultaneously updated Workspace with new automated functions powered by Workspace Intelligence, its custom AI system. These tools aim to automate routine office tasks, positioning AI as an operational assistant rather than a supplementary feature.
The TPU announcement reflects intensifying competition in AI infrastructure. As companies race to deploy increasingly complex AI agents—systems capable of reasoning, planning, and taking actions autonomously—hardware becomes a critical differentiator. Purpose-built chips can reduce latency and energy consumption compared to general-purpose processors.
Google has not disclosed detailed specifications, pricing, or availability timelines for the new TPUs. The company typically makes chips available through Google Cloud, allowing external developers and enterprises to build AI applications on the hardware.
The focus on agent-capable hardware suggests Google expects significant demand for systems that can operate more independently than current AI assistants. Whether in workplace automation, customer service, or data analysis, agents require sustained computational resources and rapid response times—exactly what specialized silicon addresses.
Google's dual approach—building both the chips and AI applications that run on them—gives the company vertical integration advantages in the competitive AI market.
Z.ai released GLM-5.3's weights on Hugging Face under a new license that requires large companies to undergo security review before hosting the model. The change marks a departure from the standard MIT license.
Anthropic has introduced the Model Hardware Standard (MHS), a unified interface enabling AI agents to operate robotic arms, lab instruments, and other physical devices. Early testing shows integration time has dropped from weeks to hours.
Open-weight AI companies—those releasing freely available models—are attracting major acquisition interest from tech giants. The trend reflects growing capital investment in the business model of distributing AI models at no cost.
Google Deepmind has upgraded its Co-Scientist AI system to autonomously plan experiments, operate lab equipment, and publish scientific papers. The Gemini-based multi-agent platform demonstrated experimentally validated results across materials science, chemistry, and medical AI development.