Agents move into infrastructure

Google Cloud’s agent push is moving past office software and into the systems that design chips and wire up networks. (cadence.com) On April 15, Cadence said it is working with Google to optimize the Cadence ChipStack AI Super Agent with Gemini on Google Cloud. Cadence said the system is aimed at chip design and verification, and that it is available now through Google Cloud Marketplace. (cadence.com) Cadence said its agent combines large language model reasoning with electronic design automation tools, the software engineers use to lay out and test semiconductors before manufacturing. The company said the setup can deliver up to 10 times productivity gains across digital design, verification planning, regression management, and debug. (cadence.com) Equinix made a parallel move on April 15, launching Equinix Fabric Intelligence as an available preview. Equinix described it as an artificial intelligence-native operational layer for managing network infrastructure across clouds, data centers, and edge sites. (investor.equinix.com) Equinix said Fabric Intelligence automates how connections are set up, adjusted, and maintained for artificial intelligence workloads, replacing more manual network operations. The company said its Fabric Super Agent can let staff manage those environments through prompts in Slack, Microsoft Teams, or the Equinix customer portal. (investor.equinix.com) (sdxcentral.com) The common idea is that an “agent” is software that does multi-step work instead of only answering one prompt. In these announcements, the work is not drafting email or slides; it is running chip-design tasks and configuring network paths that usually require specialist tools and human operators. (cadence.com) (investor.equinix.com) Google Cloud has been framing that shift as part of its infrastructure agenda ahead of Google Cloud Next 2026 in Las Vegas on April 22 to April 24. Google’s event materials highlight sessions on “agentic cross-cloud infrastructure,” “agents” on Google Kubernetes Engine, and using Gemini with Config Connector to build “10x cloud teams.” (googlecloudevents.com) (cloud.google.com) That timing matters because companies are now trying to fit artificial intelligence workloads into older technical plumbing built for more predictable traffic. Equinix cited Omdia research saying 93% of organizations see network automation as essential for future change, and 88% said artificial intelligence itself will be required for effective network automation. (investor.equinix.com) Cadence is making a similar argument in semiconductors, where design teams use sprawling tool chains and large compute clusters before a chip ever reaches a factory. The company said Google Cloud’s elastic compute is being used for Gemini reasoning, Cadence tools, and the ChipStack agent in what it called a click-to-deploy workflow. (cadence.com) The near-term test is whether these agents save engineers time without adding new failure points in expensive design and network environments. For now, the clearest signal is that Google Cloud’s agent ecosystem is being positioned as infrastructure software, not just productivity software. (cadence.com) (investor.equinix.com) (cloud.google.com)