AI Infrastructure Demand Broadens Beyond Chips as Lattice, Jacobs Post Record Results

The second-quarter reporting slate delivered a coherent message: AI capital expenditure is no longer a story confined to accelerator vendors. Lattice Semiconductor posted record revenue of $201.1m, up 62% year over year, with management attributing the acceleration directly to AI data center demand for low-power programmable logic at the rack and server-management layer. Jacobs Solutions, an engineering and infrastructure services firm, disclosed that AI data center work now represents 11% of revenue and helped push backlog to a record $29bn — a durable, multi-year revenue visibility metric that converts hyperscaler announcements into contracted work. Viasat, meanwhile, reported a 22% jump in defense awards with new satellite capacity approaching service entry, underscoring how national-security connectivity spending is running alongside the commercial compute cycle. The counterargument is being voiced with increasing force: commentary circulating this session focused on the concentration risk embedded in the trade, where the demand signal ultimately traces back to the financing capacity of OpenAI, Anthropic and a small set of hyperscalers, with Chinese model competition compressing the pricing power that underwrites the capex. For now, the order books say otherwise. Investors should watch backlog conversion rates and book-to-bill at the services layer as the cleanest tell on whether the cycle is extending or merely pulling demand forward.