Marvell Technology (MRVL) The Six Five Summit: AI Unleashed 2026 summary

AI infrastructure evolution and market transformation

• AI is moving from data centers to networks, endpoints, and application layers, with connectivity now seen as the primary bottleneck for scaling AI workloads.

• The industry is in the midst of the largest infrastructure build-out in history, with the AI revolution still in its early stages despite rapid growth since 2022.

• Compute and memory challenges are being addressed, but the focus is shifting to unleashing connectivity and interconnect technologies to enable true AI scale.

• The transition from electrical to optical connectivity, including CPO and NPO, is accelerating, with significant investments and acquisitions to build a comprehensive photonics and optics portfolio.

• Mass deployments of GPUs and AI accelerators are still ahead, with future innovations in connectivity expected to unlock new use cases and performance gains.

Strategic technology investments and product leadership

• Early bets on photonics, such as the acquisition of Inphi and Celestial AI, have positioned the company as a leader in silicon photonics, broadband analog, and DSPs for data center and interconnect solutions.

• The company offers a broad portfolio, integrating optics with XPUs and switches, supporting Ethernet, UALink, and NVLink, enabling large-scale GPU communication.

• Memory expansion technologies, especially CXL-based solutions, are seeing rapid adoption, driven by AI's demand for disaggregated and scalable memory architectures.

• Custom silicon, including XPU attach products, is gaining traction as hyperscalers seek economic and technical advantages tailored to their workloads.

• Strategic partnerships, such as a $2 billion investment from NVIDIA, enable interoperability and shared IP, reinforcing leadership in custom and merchant silicon ecosystems.

Market outlook and future opportunities

• The total addressable market (TAM) for AI infrastructure is expanding rapidly, with forecasts for data center CapEx rising from $10.7 trillion to over $12 trillion by 2030.

• The market is not zero-sum; multiple players can succeed as demand for AI infrastructure and customization grows across clouds and architectures.

• The company has consistently executed on its strategy, winning dozens of sockets for XPU attach and custom silicon, with significant revenue potential per accelerator.

• The era of abundance in AI infrastructure is just beginning, with many years of growth and innovation ahead as new connectivity and memory technologies are deployed.

• The technological advancement required for AI at scale is still in early innings, with much of the opportunity and innovation yet to come.