Intel and Lattice Semiconductor square off in an AI-hardware debate

A recent market comparison set up a straightforward question for investors watching the AI hardware supply chain, which is better positioned: Intel’s efforts to expand in AI PCs, data-center platforms, and specialized chips, or Lattice Semiconductor’s focus on low-power field-programmable gate arrays (FPGAs) used as control, connect, and compute “companion” logic in infrastructure.

Intel says its client roadmap is already tied to the AI PC shift. In an earnings-call summary released for 4Q 2025, the company said its Core Ultra Series 3 line (formerly Panther Lake) is built on its Intel 18A manufacturing process and that it delivered the first Series 3 product SKUs ahead of its own timeline. Intel also said it formally launched Series 3 at CES with OEM partners, adding that the platform is powering more than 200 notebook designs.

On the data-center side, Intel’s same materials described a broader “DCAI” approach that ties CPUs, accelerator plans, and ASICs together, alongside a continuing push to stand up a “trusted US Foundry.” The company said it centralized its data-center and AI businesses under the same leadership and pointed to strong demand for traditional servers while it ramps capacity. It also outlined server roadmap changes aimed at focusing resources, while continuing work with Nvidia on an integrated XEON design using Nvidia’s NVLink technology.

Lattice Semiconductor argues from a different angle. In its latest annual report, the company describes itself as the “low power programmable leader,” selling FPGA devices and related silicon-enabling offerings across the edge to the cloud. Lattice said its FPGAs enable control, connect, and compute technologies, and it points to demand tailwinds that include hyperscale data-center expansion, 5G infrastructure buildout, electrification and sensor proliferation, and AI deployments at the network edge.

Those strategic claims are paired with financial detail. Lattice reported total revenue of $523.3 million for the year ended January 3, 2026, up 2.7% from $509.4 million for fiscal 2024. Within that, the company said revenue in its Communications and Computing end market rose 28% in fiscal 2025 versus fiscal 2024, attributing the increase to stronger demand in data-center applications, including general-purpose and AI-specific servers, and wireline networking components.

Where the debate lands is on what “AI hardware exposure” means. Intel is trying to benefit from AI workloads through mainstream compute platforms, hybrid-AI enablement, and the scaling of next-generation process technology (18A) while also developing accelerator and ASIC options for AI use cases. Lattice, by contrast, is positioning its low-power FPGAs as enabling infrastructure inside servers and networking equipment, emphasizing dataplane, control, security, power and thermal management, and edge AI offload use cases rather than competing directly on the highest-end AI compute accelerators.

Still, important details are not fully answered by the comparison or the public materials it leans on. Intel’s earnings-call summary does not quantify, in dollar terms, how much incremental AI PC or data-center demand is expected to flow to specific product lines beyond broad roadmap priorities. Lattice’s annual report similarly emphasizes market drivers and end-market growth, but it does not break out a precise revenue forecast tied to any single AI server “attach” rate or to a specific new product ramp in the way a short market-news comparison would typically imply.

What to watch next for both companies is less about headline plans and more about execution indicates. For Intel, investors will be watching Series 3 ramp progress into 2026, Intel 18A yield and capacity improvements, and how the company’s combined CPU-and-accelerator narrative evolves in data centers and ASICs. For Lattice, the key indicates are continued growth in Communications and Computing, evidence that design wins translate into shipments in far-edge and server-related applications, and how the company’s low-power FPGA portfolio performs as customers adjust inventory and expand capacity for AI and networking workloads.