The optical bill arrives, and the software buyers keep shopping.

Bottom lineThe largest deals bought hard infrastructure and sensing. The smaller deals bought the operating layer that makes AI useful inside a specific workflow.

• Over the next three to twelve months, watch optical and edge infrastructure for follow-on transactions. The Marvell/Lumentum agreement casts optical interconnect as a strategic category alongside compute, and ADI’s Alif completion gives larger semiconductor buyers a clear edge-AI precedent. A counterfactual is straightforward: a sustained slowdown in AI data-centre orders would weaken the case for paying infrastructure-scale premiums.^{1, 2, 7}

• Defence and dual-use sensing should stay active if procurement demand remains durable. Shield’s terms show that radar, manufacturing capacity and counter-drone capability can command strategic value when they are attached to deployed systems. Any weakening of order visibility, export constraints or financing availability would test that thesis.^{4, 6}

• In application software, the likely buyer preference is for a workflow with records, an action path and a domain team. The QuestionPro, Command Alkon and Windward deals are small beside the semiconductor transactions, yet they show how software consolidators can add AI through customer engagement, operational analytics and proprietary sensing rather than buying broad model vendors.^{18, 21, 24}

• For sellers, the resulting screen is stern but usable: prove access to a hard-to-replace dataset or operating moment, show that the product already works inside a buyer’s adjacent workflow, and separate reported customer evidence from aspirational AI claims. Deals will still happen without those features. They will simply be harder to price as strategic assets.^{13, 16, 20}