OpenAI Gets Early Access to Nvidia’s Vera Rubin Systems as It Builds Multi-Chip AI Software

aiai-infrastructureai-compute-chipsai-inference-platforms 3 posts · 3 accounts

OpenAI already has early access to Nvidia’s Vera Rubin systems, a new hardware milestone for the company as it develops internal software that can run AI workloads across chips from multiple providers rather than relying solely on Nvidia. The effort is aimed at a future in which AI runs across a broader mix of chip suppliers.

The project matters because Nvidia’s CUDA software ecosystem has been one of the company’s biggest competitive advantages. Compute chief Sachin Katti has described a future of heterogeneous AI infrastructure and said OpenAI is open to making the software public; OpenAI is already using or planning chips from Amazon, Cerebras, AMD and its own custom silicon, while preparing Vera Rubin clusters for training later this year.

From the sources (3 posts)

@wallstengine

OpenAI is developing internal software to run AI workloads across chips from multiple providers, not just $NVDA, per The Information. Compute chief Sachin Katti described a future with heterogeneous AI infrastructure and hinted OpenAI coul

@anissagardizy8

NEW: OpenAI is open to the idea of publicly sharing software it’s been developing to make its AI run on chips from different providers, a move that could weaken one of Nvidia’s biggest advantages. Read more of our interview w/ OpenAI's Sac

@theinformation

OpenAI already has early access to Nvidia’s Vera Rubin systems, but it’s also building software designed for a future where AI runs across many chip providers. Full story:

Market impact

45 medium

NVDA AMD SOXX

Why it matters: OpenAI building (and possibly open-sourcing) software to run AI workloads across multiple chip providers directly threatens Nvidia's CUDA moat, its single biggest competitive advantage, while helping alternative silicon vendors like AMD, Cerebras and Amazon gain a foothold in OpenAI's compute mix.

At publish time: Headline facts state OpenAI is building cross-provider chip software and is "open to making it public" — a direct, capability-level attack on Nvidia's CUDA lock-in from its largest customer, not a routine product announcement.

model confidence: medium

What happened next

1–2 weeks later

fizzled moved up

Predicted CUDA-moat threat fizzled; OpenAI deepened Nvidia reliance (fall Vera Rubin training, 10GW Ohio data center, sold-out compute), bullish for NVDA, no AMD traction.

based on follow-up coverage

Follow-up coverage

Jun 7

OpenAI Plans Fall Training Run on Nvidia Vera Rubin Systems

matched on: nvidia, openai, rubin, systems, vera
Jun 7

OpenAI Plans Fall Training Run on Nvidia Vera Rubin as Compute Remains Sold Out Through Much of 2027

matched on: nvidia, openai, rubin, vera
Jun 7

OpenAI Plans Fall Training Run on Nvidia Vera Rubin as Compute Remains Sold Out Through 2027

matched on: nvidia, openai, rubin, vera
Jun 8

Nvidia Makes Late Vera Rubin Design Shift to Detachable Two-Piece Vapor Chamber, DigiTimes Reports

matched on: nvidia, rubin, vera
Jun 8

Nvidia Doubles Production Capacity as Vera Rubin Products Enter Mass Production

matched on: nvidia, rubin, vera
Jun 10

Gigabyte to Ship Nvidia Vera Rubin AI Servers in Second Half, Taking 2027 Orders

matched on: nvidia, rubin, vera
Jun 10

OpenAI Negotiates 20-Year Lease for 10-Gigawatt Ohio Data Center With Nvidia Support

matched on: nvidia, openai
Jun 10

OpenAI in Advanced Talks to Lease 10-Gigawatt Ohio AI Campus With Nvidia Backing

matched on: nvidia, openai

Nine US Trade Associations Warn AI Data-Center Demand Squeezes Memory-Chip Supply, Raises Prices for Consumer Goods Bitcoin Falls Below $61,000 as Ether Hits Lowest Since April 2025