Sebastien Naji

Sebastien Naji from William Blair asked for an update on the Arm China business and whether recent U.S. export controls could boost its data center contribution.

Answer

CEO Rene Haas stated that recent export controls do not significantly impact Arm's business and that the China market's momentum tracks the rest of the world. CFO Jason Child added that China represented about 21% of revenue in Q1, up from ~15% in the prior quarter and ~14% a year ago, indicating strong growth.

Sebastien Naji inquired about Astera Labs' opportunities in China, particularly the willingness of Chinese hyperscalers to adopt open technologies like PCIe or UALink.

Answer

Jitendra Mohan, CEO and Co-founder, noted that due to IP and technology constraints, Chinese hyperscalers show significant demand for PCIe-based scale-up solutions. He explained that limited IP availability for higher data rates (e.g., 200 Gb) means they often use more GPUs (16-24 vs. 8 in the U.S.) to solve similar problems, creating a larger opportunity for Astera Labs' switching and retiming solutions. Regarding NVIDIA's shift to cableless designs, Mr. Mohan reiterated that Astera Labs' opportunity arises when hyperscaler customers customize high-performance GPU platforms, such as Blackwell or Rubin, for their specific infrastructure needs.

Sebastien Naji inquired about NVIDIA's shift to a more cableless design with their Rubin servers or rack, and if this design change impacts Astera Labs' opportunity with Aries or Taurus.

Answer

Jitendra Mohan, CEO, stated that Astera Labs' opportunity with NVIDIA-based designs arises when hyperscaler customers customize these platforms for their own infrastructure, as seen with the Blackwell platform and expected for Rubin. He emphasized that the choice of backplane (cable vs. PC board) is less relevant than the customization opportunities for Astera Labs.

Sebastien Naji inquired about non-AI market trends, specifically PCIe Gen5 adoption in general-purpose servers, and asked for Astera's perspective on how emerging scale-up Ethernet solutions might compete with its PCIe-based offerings.

Answer

President & COO Sanjay Gajendra confirmed that the general-purpose compute transition is a growth driver for the Ares, Taurus, and Leo CXL product lines. CEO Jitendra Mohan positioned UALink and PCIe as superior for scale-up due to lower latency and higher efficiency, arguing the key differentiator versus scale-up Ethernet is UALink's open, multi-vendor ecosystem.

Sebastien Naji asked about Arista Networks' investments in enterprise go-to-market strategies, specifically where the most progress is expected in 2026, such as geographic expansion, channel investment, or cross-selling to existing enterprise customers.

Answer

Jayshree Ullal, Arista Networks' CEO and Chairperson, expressed excitement for the large $30 billion TAM in the enterprise segment. Todd Nightingale, Arista Networks' President and COO, highlighted three key dimensions: expansion into the campus (completed with VeloCloud), geographic expansion (noting good numbers in Asia), and reaching new logos by investing in the channel.

Sebastien Naji asked about Arista's investments in its enterprise go-to-market strategy, seeking insight into where the company expects to make the most significant progress in 2026, including geographic expansion, channel investment, or cross-selling to existing customers.

Answer

Jayshree Ullal, CEO and Chairperson, highlighted the excitement for the $30 billion enterprise TAM. Todd Nightingale, President and COO, detailed three key areas of focus: expanding into the campus market with VeloCloud, accelerating geographic expansion (noting good numbers in Asia), and reaching new logos by investing in the channel to drive more opportunities.

Sebastien Naji asked for a ranking of the opportunities across Arista's three AI networking product families: the 7060 series, the 7800 AI Spine, and the new 7700 Ethalink switch.

Answer

CEO Jayshree Ullal provided a qualitative ranking. The 7060 series is the most popular by unit volume. The flagship 7800 series is smaller in volume but the most strategic in revenue dollars. The new 7700 series, co-developed with Meta, is positioned as a 'best of both worlds' solution for mid-scale clusters up to 10,000 GPUs and has no direct competitor. Chief Platform Officer John McCool added that customers often mix these products to optimize deployments.

Sebastien Naji of William Blair & Company asked about the long-term opportunity for Ethernet in scale-up AI connectivity and whether the Meraki-Catalyst platform integration would shift hardware purchasing behavior.

Answer

CEO Chuck Robbins acknowledged a potential future shift to Ethernet for scale-up, confirming Cisco has roadmaps to address it. Regarding the platform integration, he believes it will make Catalyst more attractive to customers wanting simple cloud management, thereby expanding its appeal rather than causing a shift away from it.

Sebastien Naji of William Blair inquired about the adoption of white box solutions by hyperscalers and whether AI was driving a shift toward that model, given Cisco's flexible offerings.

Answer

CEO Chuck Robbins stated he has not seen a meaningful shift away from full systems, which remain the predominant purchase from hyperscalers. CFO Scott Herren supported this by highlighting that two-thirds of the quarter's AI orders were for complete systems, not disaggregated components, reinforcing the trend observed in the previous quarter.

Sebastien Naji asked about the mix of growth from new versus existing customers and inquired about the competitive positioning of the NVIDIA BlueField solution, including the risk of NVIDIA developing its own alternative.

Answer

President and CEO François Locoh-Donou stated that while new customer acquisition is strong, particularly for Distributed Cloud, the majority of growth comes from expanding within the existing customer base. Regarding the NVIDIA partnership, he emphasized that F5's unique L4-7 traffic management expertise is a key differentiator that is difficult to replicate, providing a compelling and early solution to the industry-wide problem of GPU utilization. He acknowledged the main work ahead is on the go-to-market model.