Sebastien Naji

Rene answered directly that V9 and CSS migration insulates Arm, then handed to Jason who conceded incremental slowdown versus start-of-year expectations.

Rene said every partner asked endorsed the strategy. Over 50 partners provided quotes, references, or videos. Emphasized the silicon strategy was driven by customer demand.

Rene emphasized power efficiency as the key differentiator driving hyperscaler adoption. All incremental compute announced has been Arm-based, driving Neoverse royalties more than doubling YoY.

Rene said Arm China tracks global trends with consistent growth. H20 export control changes have minimal impact. China was 21% of Q1 revenue, up from 15% last quarter and 14% a year ago.

Rene explained the deal gives Malaysian startups CSS-level access for chip design. Would not forecast new deals but sees potential knock-on effect from other governments.

Management said bigger switches carry higher ASP, but AI-fabric features beyond lanes drive content past $1,000 per accelerator.

Management said custom solutions have higher attach (one per accelerator) offsetting lower ASP, and content per XPU keeps growing each generation.

Management said IP constraints drive China toward PCIe Gen 5 scale-up with more GPUs per cluster, a larger per-cluster opportunity for Astera.

Management said general compute is a tracked transition with new Aries, Taurus and Leo design opportunities from PCIe Gen 6 and CXL CPUs.

Bill gave rich detail on the Pilot telemetry platform (per-link sensing, ESD/dust detection) and confirmed the growing dataset feeds better next-gen solutions.

CEO called Neoclouds one of the biggest trends and agreed the group collectively could be on the order of 20% of revenue over coming years.

Bill said Credo secured wafer and packaging supply across 12/7/5/3nm, and flagged industry-wide memory as a concern that Weaver's DDR-over-HBM approach can help address.

Bill said Credo doesn't track ASIC versus merchant GPU mix and is agnostic to GPU type; he believes customers order connectivity alongside GPUs, limiting inventory risk.

Todd outlined three priorities: campus expansion with VeloCloud, geographic acceleration (strong Asia numbers), and new logo acquisition through channel investment.

Jayshree noted balanced spend between AI and traditional cloud in 2025 vs the sharp AI pivot of 2023, with environment much more balanced.

Jayshree sees NVIDIA's 80-90% GPU share potentially falling to 50-50 in 2-3 years, positioning Arista as GPU-agnostic scale-out networking provider.

Robbins expects unified platform to benefit Catalyst adoption by adding cloud management. With 35 million devices cloud-managed, customers get choice of on-prem or cloud-managed operation.

Robbins said current hyperscaler clusters mostly use native connectivity within clusters. Scale-up Ethernet opportunity could emerge over time, Cisco has unannounced road maps to address it.

Robbins said no meaningful shift to white box, noted Cisco can integrate silicon into white box when customers prefer it. Herren added two-thirds of $600M+ orders were systems vs optics.

Francois clarified F5 isn't competing directly in CDN against players like Akamai. Instead, CDN is bundled with security capabilities. A number of the 1,000+ DCS customers took the full bundle including CDN.

Francois noted new customer wins occur across all product families, with distributed cloud services growing from 500 to 800+ customers. However, majority of growth comes from existing customer expansion and cross-selling.

Francois acknowledged competition will emerge but emphasized F5's two decades of Layer 4-7 traffic management expertise is unique and directly applicable to AI workloads. Noted remaining work is in go-to-market and commercial models.