AI Infrastructure: Outlook for 2027
Memory, Compute & AI
AI infrastructure will claim more semiconductor capacity in 2027 than any other market. This session looks at where the revenue and the shortages will land, from HBM and NAND supply to the custom silicon taking over inference.
Memory's share of 2027 semiconductor revenue
Silicon photonics share of AI optical transceiver revenue
2027 datacenter power IC market
What you'll learn
Where AI demand meets supply in 2027
Each topic draws on the 2027 Outlook reports. Bring your questions for the audience Q&A.
The impact of memory scarcity on chip and system design
For chip and system designers, memory will become an architectural constraint. Designs must now account for what memory types can actually be sourced.
The move from training to inference and the rise of custom ASICs
In 2027, focus shifts from training, where GPUs and NVIDIA have excelled, toward inference, where custom silicon such as Google TPU, AWS Inferentia, Meta MTIA and Microsoft Azure Maia is expected to be the fastest growing segment.
Silicon photonics, 1.6T optical ports, and datacenter power ICs
Silicon photonics will account for 42% of AI optical transceiver revenue, and 1.6T optical ports will become the dominant port speed. The market for datacenter power ICs will surpass $4.5B.
Reports included
