Bank of America has turned more constructive on Meta Platforms, and the reasons the analysts named are silicon and software: the company's in-house AI chips and the Meta One offering.

The call, reported by Yahoo Finance, centers on two assets. The first is Meta's custom AI accelerator program — the silicon the company designs itself to run training and inference workloads for its recommendation and generative AI systems. The second is Meta One, the platform initiative that Bank of America's analysts flagged alongside the chip effort as a basis for their improved outlook.

The significance for the semiconductor supply chain is straightforward. Every large hyperscaler that shifts inference or training onto internally designed silicon redirects orders away from merchant GPU vendors and toward contract foundries, advanced packaging houses and memory suppliers. Meta's chip program is one of the largest such efforts by volume of AI compute consumed. When a bank's research desk raises its view of the company partly on the strength of that program, it is effectively endorsing the execution of a silicon roadmap — a judgment that carries weight because hyperscaler roadmaps decide where some of the biggest advanced-node wafer allocations in the industry end up.

Custom accelerators matter to Meta for a second reason: cost. Internal silicon, once validated, typically runs cheaper per unit of inference than merchant GPUs at hyperscale, and Meta operates one of the largest inference estates in the world across Facebook, Instagram and WhatsApp. If Bank of America sees the chip program as a positive, the analysts are implicitly crediting Meta with progress on the long road from dependent customer to partial self-supplier of AI compute.

Meta One, the second named driver, extends the same logic from infrastructure to product. A unified platform offering gives Meta a cleaner commercial surface for monetizing the AI capability stack that its data centers and custom silicon support. Analyst optimism here reflects the expectation that the platform converts compute investment into revenue rather than leaving it as pure capital expense.

The report does not, in the summary available, attach specific price targets, capacity figures or timeline commitments to the improved outlook. What it does establish is direction: a major sell-side research house now counts Meta's own silicon among the reasons to hold a more favorable view of the stock, rather than treating the chip program purely as a costly capital-expenditure line.

That framing change matters beyond one ticker. It signals that investors are beginning to underwrite custom hyperscaler silicon as an asset with measurable returns rather than a speculative bet — a shift that shapes how the market values the foundry, packaging and memory suppliers that ultimately manufacture these designs.

For Meta, the next test is execution at scale: shipping its AI chips in volumes large enough to visibly bend its compute cost curve, and demonstrating that Meta One converts that compute into durable revenue.