Nvidia’s earnings reveal systems advantage stretching beyond the GPU

👁 117 reads

Nvidia’s earnings report on Wednesday prompted investors to look past the GPU and toward the full rack of orchestration hardware the company now sells alongside it - a shift that reframes how durable the chipmaker’s competitive position might be as AI compute scales into the gigawatt range.

The company is currently rolling out its Vera Rubin architecture, which pairs the Rubin GPU with a Vera CPU, a Groq 3 LPX inference accelerator, and dedicated storage and networking racks. Each unit targets a different bottleneck outside the GPU itself. The Vera CPU, in particular, focuses on moving data to the GPU at the right time - a problem that grows more acute as data centers scale memory capacity alongside raw compute.

“We saw upwards of 3x improvement in these operations, where the Vera CPU is allowing for acceleration. So now we can use our flash to its fullest potential, because we can get all that performance out of it without bottlenecking.”

Nvidia’s share price grew roughly 10x between the start of 2023 and mid-2025, then entered a more modest trajectory over the past year as hyperscalers including Amazon and Google began building their own chips. The new narrative emerging after Wednesday’s results is that GPU competition, while real, addresses only one layer of what Nvidia sells.

A contrasting approach comes from OpenAI’s Jalapeño chip, which the company says was designed to “minimize data movement and communication delays” by keeping entire workloads within one connected system - sidestepping the orchestration problem rather than solving it with dedicated hardware.

Practitioner takeaway: For developers and businesses choosing or pricing infrastructure, the implication is that raw GPU specs are an incomplete guide to cost and throughput. Vendors that bundle orchestration hardware - handling memory traffic, storage and networking together - may deliver meaningfully better tokens-per-watt than GPU benchmarks alone suggest, so total-system efficiency is worth asking about when evaluating cloud tiers or on-premise deployments.

// read moreClaude Opus 5 System Prompt Reveals Export-Control SuspensionNewsNvidia backs $500B data center fund, promises GPU resale value supportMagazineSpaceX closes $60B Cursor acquisition, promising GPU fleet accessMagazine

Kinetyra