NVIDIA OpenAI $100 Billion deal: CFO says it’s still a term sheet and NVIDIA’s lead “has absolutely not narrowed”

She also emphasized that NVIDIA’s performance and ecosystem lead has not narrowed.

Resume Captain

Your AI Career Toolkit:

• AI Resume Optimization
• Custom Cover Letters
• LinkedIn Profile Boost
• Interview Question Prep
• Salary Negotiation Agent

Get Started Free

What Kress said — highlights from the interview (NVIDIA OpenAI deal + ecosystem)

• OpenAI deal still a term sheet.
• The roughly $100,000,000,000 USD commitment NVIDIA announced at the end of September to invest in OpenAI remains at the investment-intent stage; no final agreement has been signed yet.
• ($100,000,000,000 USD ≈ ¥730,000,000,000 RMB; conversion used: 1 USD ≈ ¥7.3 RMB, see note below.)
• Revenue from that collaboration not included in guidance.
• Kress emphasized that any revenue tied to that agreement has not been baked into NVIDIA’s current revenue guidance.
• Platform adoption is broad.
• Kress said virtually all models and workloads — whether cloud or on-prem — are running on NVIDIA’s platform, and customers continue to deploy on NVIDIA systems.
• Hardware + software moat.
• She highlighted NVIDIA’s combined hardware and software stack (notably CUDA and its libraries) as a durable competitive advantage that improves over time via software updates.
• Anthropic deal also influences bookings.
• NVIDIA’s earlier commitment to invest up to $10,000,000,000 USD in Anthropic (≈ ¥73,000,000,000 RMB) may further increase booked capacity and chip allocations.
• Demand outlook.
• NVIDIA has said systems like Blackwell and Vera Rubin are expected to produce several hundred billion dollars of demand over 2025–2026.
• Kress repeated that demand remains very large and early in the CPU→GPU migration cycle.

Detailed Q&A points — what investors, founders, and techies need to know

1. Is there an AI bubble?

Kress acknowledged that historically much datacenter compute was on CPUs and that workloads are now shifting to GPUs.

She said the migration is still in early stages and expects trillions of dollars of spending on datacenter infrastructure this decade, with roughly half related to the CPU→accelerated-compute transition.

This, she argued, is additive capacity rather than one-for-one replacement.

2. Has NVIDIA’s lead narrowed?

“Absolutely not,” Kress said.

She emphasized NVIDIA’s system-level co-design approach — multiple chips and an integrated software stack — rather than a single fixed-function ASIC.

That combination, she said, keeps models and workloads running on NVIDIA’s platform across cloud and on-premises deployments.

3. Are NVIDIA’s customers profitable?

Kress argued that as the AI industry shifts from generative-training to inference and as models scale, paying customers and monetization increase.

That dynamic, she said, creates a “flywheel” driving further model expansion and additional compute investment.

4. Do big model developers’ unpaid bookings pose a risk?

UBS asked about the risk that some large model builders with weak revenues have booked massive amounts of compute.

Kress said much of the demand is long-term and will be resolved based on capital availability.

NVIDIA focuses on current and near-term work — whether customers have purchase orders and the capital to pay for them.

5. Vera Rubin and product roadmap

Kress said the next-generation architecture, Vera Rubin, has completed tape-out and is expected to ship in the second half of next year.

She said Vera Rubin will deliver a “multi‑fold” performance jump versus the prior generation.

6. Gross margin outlook

Despite concerns over HBM memory cost inflation, NVIDIA expects to keep gross margins in the mid‑70% range next year.

7. Inventory and purchase commitments spiked

Inventory plus purchase commitments rose by nearly $25,000,000,000 USD in the last quarter (≈ ¥182,500,000,000 RMB).

Kress said that reflects preparing supply for future growth and that much of the inventory is likely already in transit or delivered to customers.

8. Capital allocation priorities

Based on NVIDIA’s demand guidance, Kress said next year’s revenue is expected in a range around $350–400 billion USD (≈ ¥2.56–2.92 trillion RMB).

The company will prioritize capital for supply and capacity build (including launching Vera Rubin), followed by shareholder returns (buybacks and dividends) and strategic ecosystem investments.

Large-scale transformational M&A is difficult now; NVIDIA will favor smaller engineering and platform teams that contribute to its stack.

Find Top Talent on China's Leading Networks

• Post Across China's Job Sites from $299 / role, or
• Hire Our Recruiting Pros from $799 / role
- - - - - - - -
• Qualified Candidate Bundles
• Lower Hiring Costs by 80%+
• Expert Team Since 2014

Get 25% Off
Your First Job Post

Monetary figures (converted)

• OpenAI investment announced: $100,000,000,000 USD (≈ ¥730,000,000,000 RMB)
• Potential system-driven revenue mentioned earlier: $400–500 billion USD (≈ ¥2.92–3.65 trillion RMB)
• Anthropic commitment: up to $10,000,000,000 USD (≈ ¥73,000,000,000 RMB)
• Inventory and purchase-commitment increase last quarter: ≈ $25,000,000,000 USD (≈ ¥182,500,000,000 RMB)
• NVIDIA next-year revenue guidance range cited by management: $350–400 billion USD (≈ ¥2.56–2.92 trillion RMB)

Conversion note: All RMB↔USD conversions use an approximate rate of 1 USD = ¥7.3 RMB (rounded) for clarity; exact exchange rates on any given day will vary.

ExpatInvest China

Grow Your RMB in China:

• Invest Your RMB Locally
• Buy & Sell Online in CN¥
• No Lock-In Periods
• English Service & Data
• Start with Only ¥1,000

View Funds & Invest

Context & implications — what this means for investors, founders, and tech leaders

Kress’s comments underline two cross-cutting themes: large announced commitments to model developers remain subject to final contracts and payment capacity, and NVIDIA is positioning its combined hardware-plus-software platform as the long-term standard for modern AI workloads.

For investors, that suggests substantial near‑term demand but also that booked or announced intent does not automatically equal realized revenue until agreements are finalized and capital flows.

For founders and cloud operators, the message is to plan for an extended CPU→GPU migration and to evaluate software stack lock-in risks given NVIDIA’s emphasis on CUDA and integrated toolchains.

For product and engineering teams building models, the practical takeaway is to consider deployment flexibility across cloud and on-prem and to weigh performance gains against platform dependence.

Quick action checklist for stakeholders

• Investors: Monitor finalized contracts and payment milestones before assuming revenue recognition.
• Founders: Map migration timelines for GPU-accelerated inference and training costs into your GTM and pricing models.
• Tech leaders: Evaluate the operational trade-offs of NVIDIA’s stack versus alternative accelerators and software ecosystems.
• Marketers: Frame product narratives around interoperability and cost predictability during the CPU→GPU transition.

Bottom line: NVIDIA OpenAI $100 Billion deal is not finalized, and NVIDIA continues to assert its systems and software lead.

References

• NVIDIA CFO interview — original reporting (financial news) – 财联社
• 英伟达CFO：OpenAI千亿大单尚未敲定 领先优势“绝对没缩小” – 东方财富
• Investor Relations and Press Releases – NVIDIA Corporation

Keyword: NVIDIA OpenAI $100 Billion deal