Advanced Micro Devices, Inc.

NASDAQ Global Select
Back to all articles
Somewhat Bullish +35

Pat Gelsinger Just Backed a $100 Million Challenge to Nvidia’s AI Networking Moat. AMD Could Benefit - Insider Monkey

AI-generated analysis

🚀 Pat Gelsinger backed Delos Data's $100 million raise to build heterogeneous networking chips that work across different AI compute hardware.

🔗 Delos aims to break Nvidia's tightly integrated stack by enabling data centers to mix various accelerator types without friction.

💡 AMD benefits indirectly as hardware-agnostic networking lowers the barrier for mixing its Instinct accelerators with other vendors.

⚠️ Delos faces a steep challenge proving performance improvements against Nvidia's mature software and established scale.

📈 Hedge fund exposure to AMD increased in Q2, with Marshall Wace holding approximately 3.9 million shares.

📉 AMD had about 41.7 million shares sold short as of August 31, representing roughly 2.6% of its float.

🏗️ The strategy targets the specific bottleneck of idle compute waiting for data to arrive in AI clusters.

🔄 Success for Delos would allow AMD to compete more effectively against Nvidia's full-rack system sales model.

Bullish Signals
  • Indirect benefit from Delos Data's $100 million funding round aimed at creating hardware-agnostic networking that lowers friction for mixing AMD accelerators with other vendors.
  • Strategic advantage gained if Delos succeeds in making mixed AI clusters easier to build, potentially weakening Nvidia's dominant ecosystem lock-in.
  • Alignment with Pat Gelsinger's view on reducing idle compute waste, positioning AMD as a key player in the next phase of AI infrastructure efficiency.
Risk Factors
  • Indirect risk that Delos fails to prove claimed performance improvements in commercial deployments against Nvidia's established scale and mature software stack.
  • Continued vulnerability to Nvidia's integrated system sales model which protects its moat when customers buy accelerators, networking, and software together.
Full Analysis
Pat Gelsinger, general partner at Playground Global, backed Delos Data's $100 million funding round to develop heterogeneous networking chips for AI data centers. The startup aims to create hardware-agnostic networks that function across different compute architectures, directly challenging Nvidia's tightly integrated ecosystem of accelerators, interconnects, and software. This development presents a strategic opportunity for Advanced Micro Devices (AMD) by potentially lowering the friction associated with mixing AMD hardware with other vendors in large-scale clusters. If Delos succeeds in making mixed AI clusters easier to build, it could weaken Nvidia's dominant position where customers typically purchase entire systems together, thereby opening wider doors for AMD and other accelerator challengers. Despite the potential upside, significant risks remain as Delos must prove its claimed performance improvements in commercial deployments against an incumbent that spends billions on research. While hedge fund exposure to both Nvidia and AMD has increased recently, the article notes that Delos does not need to replace Nvidia entirely to matter, only to make interoperability more viable for data center operators. The narrative highlights a shift in focus from raw accelerator speed to network efficiency as the next bottleneck in AI infrastructure. By addressing idle compute waiting for data, this approach aligns with Gelsinger's view on energy and capital waste, offering AMD an indirect but meaningful avenue to gain market share in the evolving AI networking landscape.