Cisco is reportedly on the verge of investing in CoreWeave, one of the hottest neo-cloud providers specializing in AI infrastructure, in a transaction valuing the GPU cloud provider at $23 billion. CoreWeave has rapidly scaled its AI capabilities by utilizing NVIDIA GPUs for data centers, becoming a key player in the AI infrastructure market.

Who is CoreWeave?

CoreWeave is a specialized cloud infrastructure provider focused on HPC and AI workloads. Founded in 2017 and headquartered in Roseland, New Jersey, CoreWeave quickly positioned itself as a significant player in AI cloud computing, catering to industries requiring large-scale, computationally intensive workloads.

CoreWeave has attracted major investments from notable technology and financial institutions, including NVIDIA, Magnetar Capital, Coatue Management, Jane Street, and Fidelity.

The company has seen significant financial backing due to the increasing demand for AI infrastructure:

• In May 2023, CoreWeave raised $1.1 billion in a Series C funding round.
• The company also secured an additional $7.5 billion in debt financing to support its GPU infrastructure and further expansion.

Aligned with Cisco’s AI Strategy

Cisco has increasingly focused on positioning itself in the AI infrastructure market. An investment in CoreWeave aligns with its broader strategy:

• AI-Focused Networking Solutions:
  • Expansion of custom ASICs and AI-specific network designs to push Ethernet deeper into AI clusters.
  • Integration of NVIDIA’s SuperNICs into Cisco platforms, offering lower latency than traditional Ethernet.
• AI Datacenter Expansion:
  • The investment in CoreWeave signals Cisco’s interest in ensuring its hardware becomes integral to AI infrastructure growth.
  • Cisco aims to capitalize on the increased data movement in AI-centric workloads, potentially driving demand for its networking products in AI-focused data centers.

Strategic Considerations

• NVIDIA’s Role in the Ecosystem:
  • NVIDIA’s InfiniBand is the de facto standard in AI compute fabrics, commanding a significant share of large-scale AI training deployments (~90% of the market).
  • Cisco’s investment in CoreWeave signals an attempt to compete with or complement NVIDIA’s networking dominance by offering hybrid solutions where InfiniBand is used for compute clusters and Cisco Ethernet is used for storage, orchestration, and management.

Market and Competitive Landscape

• Competitors in the AI Datacenter Space:
  • Arista Networks and Juniper Networks: Both competitors are active in the cloud and data center networking markets. Arista, for example, focuses on high-performance, low-latency networking for cloud and AI workloads, making it a direct competitor to Cisco in this space.
  • Traditional Hyperscalers: CSPs like AWS, Microsoft Azure, and Google Cloud offer GPU-based AI services and have their own networking infrastructure investments—Cisco’s investment in CoreWeave positions it to compete in supporting smaller, specialized AI infrastructure providers.
• Potential Competitive Impact:
  • Cisco’s hardware integration into CoreWeave’s infrastructure could give it a strategic advantage over competitors in the AI networking space, especially for non-compute network layers.
  • Given that NVIDIA dominates AI compute networking with InfiniBand, Cisco may focus on Ethernet for hybrid use cases or customers looking for more flexibility in network infrastructure. This would make Cisco competitive with companies like Arista, which have traditionally dominated the high-performance Ethernet market.

Analysis

From Cisco’s perspective, the rationale for the investment is clear: AI workloads are set to dominate data center infrastructure, driving an exponential need for high-speed, reliable networking solutions. Cisco’s portfolio, which has been evolving to cater to AI applications, fits squarely into this opportunity. By potentially embedding its technology into CoreWeave’s AI-heavy environments, Cisco benefits from the accelerated growth of data movement in AI processing—where speed, low latency, and scalability are paramount.

At a more granular level, Cisco’s challenge will be integrating its Ethernet-based solutions into an ecosystem currently dominated by NVIDIA’s InfiniBand technology, which remains the preferred choice for AI compute fabrics due to its low-latency, high-throughput characteristics.

While CoreWeave is heavily invested in InfiniBand, there is ample room for Cisco’s networking gear in other parts of the infrastructure, including management, orchestration, and storage layers, which don’t require the same latency precision.

Cisco’s investment may also signal a longer-term bet on broadening the adoption of Ethernet in AI compute environments. As evidenced by its recent collaborations with NVIDIA, Cisco is already moving to push Ethernet deeper into AI clusters with custom switch ASICs and integration of NVIDIA’s BlueField-3 DPUs and SuperNICs into its platforms. This deal could be a springboard for Cisco to expand its foothold in the AI data center space by influencing future infrastructure designs.

For CoreWeave, the partnership with Cisco adds significant industry credibility and access to networking expertise as the company prepares for a possible IPO. Already backed by major investors like NVIDIA, Fidelity, and Coatue Management, CoreWeave has enjoyed a meteoric rise on the back of surging demand for GPU-based computing, particularly for AI applications. Cisco’s involvement will likely accelerate this momentum, giving CoreWeave the networking backbone necessary to scale further in an industry expected to grow exponentially over the coming years.

Oveall, Cisco’s potential investment is more than just a capital infusion—it’s a calculated move to secure a stake in the future of AI infrastructure. By aligning itself with CoreWeave, Cisco ensures that its products and technology become indispensable to the rapid build-out of AI data centers. This space will define the next decade of enterprise computing.

The real challenge for Cisco will be to find ways to insert itself deeper into AI compute clusters while coexisting with Nvidia’s InfiniBand dominance. If successful, this could pave the way for Cisco to realize its broader AI ambitions, which, according to the company, include hitting $1 billion in AI product orders by 2025.