For business owners
AI moves into the plumbing: chips, security, and your business data
November 3, 2026 in summary: a $115 billion chip deal, security extended to production AI workloads, and business data opened to coding agents. The common thread is that AI is moving from a separate tool into the infrastructure underneath everyday work.
- Daily brief
- AI news
- Automation
November 3 did not produce a single breakthrough product. It produced something quieter and more consequential: three moves that push AI further into the infrastructure underneath everyday work. A chip deal reshapes who supplies the hardware, a security partnership extends protection to AI workloads running in production, and a data platform opens business records to the coding agents that now build software. Taken together, they describe the same shift. AI is no longer mainly a tool you open. It is becoming part of the plumbing, and the businesses that benefit will be the ones that treat it that way, with the same attention to cost, security, and data control they already give to the rest of their systems.
The hardware underneath AI is consolidating
The largest signal of the day came from the supply chain. Marvell is buying Lumentum in a $115 billion all-stock deal, with Lumentum shareholders receiving 4.5 Marvell shares each and closing expected in the first half of 2028, according to the TechCrunch report. A deal of that size is not really about one product line. It is about control over the optical and chip components that carry AI workloads, and about how much of that stack ends up in fewer hands. For a business owner, the practical point is not the price tag. It is that the costs and availability of AI capability are increasingly set by consolidation decisions far upstream, which is one more reason to avoid building a core workflow around a single vendor's pricing or roadmap without an exit plan.
Security now has to cover AI workloads, not just offices
The second move fills a gap that consolidation alone does not address. CrowdStrike and CoreWeave have expanded their partnership so CoreWeave customers get Falcon platform protection and around the clock managed security for AI workloads running on its cloud, as described in the CrowdStrike announcement. That matters because AI workloads are not like a website sitting behind a firewall. They ingest data, call tools, and act on results, which means a breach or a manipulated input can travel further than a stolen password would. Security that only covers laptops and email no longer covers where the work actually happens. If your business runs AI on a cloud platform, the question to ask your provider is simple: is that workload protected with the same monitoring and response as the rest of your systems, or is it still treated as an experiment?
Your business data is becoming agent-ready
The third move brings the shift closest to home. Microsoft has brought Dataverse to Copilot Cowork, to coding agents such as GitHub Copilot and Claude Code, and to n8n, so business data and its security rules travel with the builder wherever they work, without custom plumbing, according to the Microsoft Power Platform blog. In plain terms, the records a small business already keeps, customers, orders, inventory, can now be reached by the agents that write and automate software, under the same permissions that already govern them. That removes a lot of tedious integration work, but it also raises the stakes on data hygiene. An agent can only be as trustworthy as the records and permissions it is given, so cleaning up who can see what is no longer a back office chore. It is the foundation the automation stands on.
What this means for a small business
Read together, the three stories point to the same practical checklist. First, treat AI as infrastructure, not as a gadget: ask what it depends on, what it costs if a supplier changes direction, and how you would switch. Second, extend your security thinking to AI workloads themselves, including the data they touch and the tools they can call. Third, get your business data and its permissions in order before you hand it to an agent, because that is what determines whether automation helps or creates a mess at greater speed. None of that requires a large budget or a new platform. It requires the same discipline that already makes a business run well, applied to a new layer of tools.