Episode 6 of MI&S IT Talk breaks down key takeaways from Google Cloud Next, including Google’s progress with Gemini Enterprise, its agentic AI strategy, and the evolving role of TPUs in enterprise AI infrastructure. Matt Kimball and Jason Andersen also discuss the expanding AI silicon market, why inference is creating room for more chip players, and what enterprise IT leaders should watch as AI platforms become more integrated, distributed, and competitive.
Key Points & Takeaways
Google Cloud is becoming more enterprise-ready.
Jason sees a major shift in how Google Cloud is packaging its AI and agentic tools for enterprise buyers. Gemini Enterprise and the broader agentic suite feel more like an integrated IT product than a loose collection of technologies.
Gemini Enterprise is now in the front pack.
Compared year-over-year, Google’s agentic AI platform has made a significant leap. Jason suggests Google is now more competitive with offerings from Microsoft and Salesforce, though integration breadth remains a key question.
Google still needs to show more for professional developers.
Jason was impressed by the no-code and low-code agentic tooling but wanted to see more emphasis on pro developer tools, especially for teams building enterprise-grade agents, applications, and AI-enabled systems.
TPU architecture is strong, but enterprise positioning is less clear.
Matt praises Google’s TPU architecture, especially the split between TPU 8T for training and TPU 8I for inference. However, he questions how well the TPU story maps to real enterprise environments, particularly hybrid or heterogeneous infrastructure.
The AI silicon market is expanding beyond NVIDIA.
Both hosts agree NVIDIA will remain the dominant market leader, but AMD, Intel, AWS, Microsoft, Google, Apple, and emerging chip startups all have opportunities as inference expands across the data center, edge, and devices.
Inference will drive broader silicon diversity.
As AI moves closer to enterprise workflows, manufacturing floors, devices, and local environments, the hosts expect a wider mix of chips and accelerators to support different workload needs.
Orchestration may become a major enterprise opportunity.
As companies adopt multiple AI hardware and software stacks, Matt points out that someone will need to help orchestrate across all of it. That could become a significant layer in the enterprise AI stack.
Google’s “all-in” ecosystem approach may limit some buyers.
Jason notes that Google’s agentic and AI infrastructure approach works well for organizations building the Google way, but it may be less attractive for companies needing broad multi-platform agentic governance or third-party integration.
AI vendors need clearer competitive benchmarks.
Matt argues that claims like “80% faster than the previous generation” are not enough for enterprise buyers. He wants clearer comparisons against market leaders, especially around cost per token, tokens per watt, and inference economics.
AWS and OpenAI are becoming more interesting partners.
Jason discusses the OpenAI-AWS relationship as partly a marketing move today, but potentially more meaningful if the companies collaborate on new models or deeper integrations through Bedrock and managed AI services.
Watch the full episode here:
Disclaimer: This show is for information and entertainment purposes only. While we will discuss publicly traded companies on this show. The contents of this show should not be taken as investment advice.
