CoreWeave now offers Nvidia's Vera chip for AI agent workloads, the cloud provider disclosed via its online product catalog. The catalog entry marks one of the first public commercial channels for the next-generation Nvidia silicon outside Nvidia's own direct sales channels.

CoreWeave's product page frames the Vera chip as inference and orchestration silicon for autonomous agent systems. The page identifies customers running agent-class workloads as the target buyer and lists the chip under a dedicated AI agent category. Specific instance SKUs, per-hour pricing and capacity commitments do not appear on the public listing.

What does the catalog entry confirm?

The product page records three confirmed data points:

• Vendor: Nvidia
• Product line: Vera
• Target workload: AI agent deployments

The page omits memory capacity, interconnect topology, accelerator pairings, pricing and committed capacity. CoreWeave directs prospective customers to sales for configuration and quotation.

Which specifications and commercial terms remain undisclosed?

Multiple technical and commercial fields are absent from the public listing:

• Memory capacity and HBM configuration
• Interconnect generation and topology
• Accelerator pairing rules with Vera
• Per-instance vCPU or memory allocation
• Per-hour price and minimum commitment
• Regional availability and rollout schedule

CoreWeave has not indicated whether the entry reflects general availability or a limited design-partner program.

Why does Vera target agent workloads?

AI agent deployments sustain long-running contexts, execute multi-step tool calls and orchestrate external services. Such workloads stress memory bandwidth, low-latency CPU-GPU coherency and persistent state retention. CoreWeave's placement of Vera under a dedicated AI agent category — rather than alongside general-purpose inference or training instances — signals the provider targets sustained production workloads, not short-lived training jobs.

Which customers does the listing target?

The AI agent framing on CoreWeave's catalog page points to enterprise teams running autonomous agent systems. The categorization suggests the Vera instances will compete for production workloads rather than research or training deployments. Enterprise software, customer support automation, developer tooling and code-generation stacks are typical entry points for dedicated agent infrastructure.

What questions does the listing leave open?

The catalog entry leaves several commercial and technical points unanswered:

• What per-hour pricing does CoreWeave set for Vera instances?
• When does general availability open beyond initial design partners?
• How does Vera pair with companion Nvidia accelerators on CoreWeave's reference architectures?
• Which regions receive Vera capacity at launch?
• What memory and interconnect specifications define each instance SKU?

How does the listing reshape GPU-cloud competition?

Channel availability for a purpose-built AI agent chip lowers the procurement barrier for enterprise teams. Customers who would otherwise wait for broad public availability can engage CoreWeave's sales team for early allocation. The pricing structure CoreWeave eventually announces will set an early reference point for competing GPU-cloud providers carrying the same silicon.

Industry attention will shift to CoreWeave's sales briefings and Nvidia's own Vera rollout cadence for confirmation on the open questions. Until then, the catalog listing serves as the only public signal that Vera has reached a commercial channel outside Nvidia's direct enterprise program.