NVIDIA launched a formal revenue-sharing and credit-support model on July 1, letting capital-constrained AI clouds access Grace Blackwell GPUs without upfront purchase — in exchange for a recurring cut of cloud revenue. First partners Sharon AI (40,000 GPUs) and Firmus Technologies (170,000 GPUs, 360MW Batam data center) give the program 210,000 GPUs at launch, extending the playbook from NVIDIA's $6.3B CoreWeave capacity guarantee.
NVIDIA launched a formal revenue-sharing and credit-support model on July 1, letting capital-constrained AI clouds access Grace Blackwell GPUs without upfront purchase — in exchange for a recurring cut of cloud revenue. First partners Sharon AI (40,000 GPUs) and Firmus Technologies (170,000 GPUs, 360MW Batam data center) give the program 210,000 GPUs at launch, extending the playbook from NVIDIA's $6.3B CoreWeave capacity guarantee.
NVIDIA published a blog post on July 1, 2026 — co-authored by CFO Colette Kress and Raj Mirpuri — announcing a formal revenue-sharing and credit-support model for AI cloud operators. Under the structure, NVIDIA backstops GPU infrastructure buildouts for smaller cloud providers in exchange for a recurring share of the cloud revenue those GPUs generate. NVIDIA earns standard upfront hardware revenue plus an ongoing percentage of cloud service revenue on supported capacity.
The first two named partners signal hyperscaler-class scale. Sharon AI (Nasdaq: SHAZ) will deploy up to 40,000 NVIDIA Grace Blackwell GB300 GPUs. Firmus Technologies is building a DSX AI factory campus in Batam, Indonesia, expected to scale to 360 megawatts and house up to 170,000 NVIDIA GPUs — giving the two deals a combined commitment of roughly 210,000 GPUs.
The model solves a structural problem in AI infrastructure finance: GPU residual value. Traditional banks will not collateralize GPU clusters because only NVIDIA knows what next-generation architecture will do to current hardware depreciation. By guaranteeing to purchase idle capacity at a predetermined price, NVIDIA effectively sets a floor on the asset's value, making the cluster bankable. Lenders are now lending against NVIDIA's balance sheet as much as the hardware.
The program extends a model NVIDIA established with CoreWeave in September 2025, when the chipmaker committed to purchase all of CoreWeave's unsold computing capacity through April 2032 under an agreement valued at $6.3 billion. Sources indicate NVIDIA is negotiating a similar financial backstop for OpenAI's planned data center projects.
Firmus expects to collect between $25 billion and $30 billion in committed customer offtake during the first six years. The company raised $505 million in April 2026 at a $5.5 billion valuation, with NVIDIA participating as an investor, and arranged a $10 billion debt facility. Sharon AI listed on Nasdaq in February 2026 via a $125 million IPO and raised an additional $1.6 billion in private placement in June 2026.
NVIDIA's broader infrastructure commitment: over $40 billion across cloud providers and data center operators in 2026 alone, with stated plans to invest up to $500 billion over four years in U.S. AI infrastructure alongside partners like TSMC and Foxconn. The company is also aiming to raise at least $20 billion in debt to fund general corporate purposes.
Market reaction was mixed. SHAZ fell 14.2% to $67.91 on July 3 on volume of 2.6 million shares, reflecting investor ambivalence about Sharon AI's margin profile as a revenue-share partner.
Securitize (NYSE: SECZ, ~$5B AUM) launched Securitize Stocks today at 13:00 UTC: 12 tokenized US mega-caps — including Apple, Microsoft, Nvidia, Tesla, Circle, Strategy and Palantir — each backed 1:1 by a real share, settling in USDC on Solana with Jump Trading running the proprietary AMM and clearing firm RQD* handling custody and settlement. It is the first product live under the SEC's five-year Innovation Exemption for onchain stock trading, granted September 17.
Anthropic's much-anticipated IPO prospectus — reported in detail by Reuters and the Financial Times — pursues a $2 trillion valuation while disclosing a $42 billion net loss for 2025 and plans to spend $518 billion on cloud, computing, and infrastructure in the coming year. Roughly 80 of the filing's 261 main-body pages cover risk factors, including that advanced AI could pose 'catastrophic or existential risks to humanity,' versus 48 pages describing the business. Nvidia separately disclosed that Anthropic has more than $180 billion in contracted compute commitments — 2.5GW of Nvidia infrastructure shipping through 2028.
Nvidia's board authorized an additional $150 billion in share repurchases on Monday — which the company says is the largest single authorization increase ever announced by a public company — lifting total remaining buyback capacity to $235 billion, expected to be spent through fiscal 2028. NVDA rose 2.8% to a $5.42 trillion market cap as Jensen Huang called AI 'a once-in-a-generation platform shift.'