In brief
Nvidia signed memorandums of understanding on August 10 with Apollo Global Management, BlackRock, Blackstone, Brookfield Asset Management, Goldman Sachs and KKR to mobilise more than $500 billion for AI infrastructure.
The partnership will use computing power as collateral, treating compute as a new investment class comparable to real estate or bonds, with debt financing as the primary instrument.
S&P downgraded Oracle to BBB- with stable outlook, warning of cash deficits by 2027. Alphabet posted negative quarterly cash flow for the first time since its 2004 IPO and launched an $85 billion share issuance program.
Nvidia is developing Nemotron 4 open-source models with at least 1 trillion parameters, with possible release as early as late fall. On August 11 it unveiled Nemotron 3.5 Lightning and NeMo Switchyard.
Hyperscalers including Google, Anthropic and Meta have spent over $1 trillion in three years on AI infrastructure, while also designing their own chips to reduce dependence on Nvidia.
- 10 AugNvidia signs MOUs with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to mobilise over $500bn for AI infrastructure
- 11 AugThe Information reports Nvidia developing Nemotron 4 open-source models with at least 1 trillion parameters
- 11 AugNvidia unveils Nemotron 3.5 Lightning and NeMo Switchyard open-source routing library
Nvidia's $500 billion financing alliance
On August 10, Nvidia signed memorandums of understanding with six of Wall Street's largest financial institutions to mobilise more than $500 billion for artificial intelligence infrastructure development. The consortium includes Apollo Global Management, BlackRock, Blackstone, Brookfield Asset Management, Goldman Sachs, and KKR. The partnership aims to create dedicated financing platforms that offer Nvidia's customers access to capital at competitive rates, accelerating the construction of data centres, energy infrastructure, and what the company calls "AI factories" worldwide.
Jensen Huang, Nvidia's CEO, framed the initiative as a turning point for how computing capacity is funded.
“We are bringing together the world's leading providers of long-term capital to independently secure AI infrastructure.”
He also wrote on the platform X that this is the beginning of an open capital market for AI infrastructure. The partners have not yet disclosed the timeline or the structure of the financing products, and the agreements are initially only letters of intent.
Compute as an asset class
The partnership treats computing power as a new investment class, comparable to real estate or bonds. Nvidia's products are described as an asset in which one can invest. The financing will focus on debt instruments, using computing power as collateral, a model that channels the auto-loan market. Executives indicated that multiple negotiations are already underway within this framework.
Jim Zelter, president of Apollo, described compute as a scarce and strategically important resource, one he expects to generate economic growth and productivity for years. David Solomon, CEO of Goldman Sachs, said the world is in the midst of a "historic AI investment cycle." Huang spoke of a self-reinforcing positive cycle of the industrial AI revolution.
The hyperscaler strain
Companies using Nvidia processors, including Google, Anthropic, and Meta, have spent over $1 trillion in three years building AI infrastructure. That spending has driven Nvidia's market value up fivefold. BlackRock has already closed a separate deal with Meta to finance and take a majority stake in a data centre in Texas. Anthropic has struck agreements with Macquarie Asset Management and GIC, Singapore's sovereign wealth fund, to support compute demand behind its Claude AI assistant, without disclosing the investment amount.
The capital appetite is worrying rating agencies. S&P downgraded Oracle to BBB- with a stable outlook, citing the growing weight of AI infrastructure on a balance sheet historically focused on enterprise software and databases. S&P warned that Oracle risks significant cash deficits by 2027. Alphabet and Microsoft remain further from downgrade thresholds thanks to diversified activities and stronger operating cash generation. However, Alphabet reported a quarter of negative cash flow for the first time since its 2004 IPO and launched an $85 billion share issuance program to fund growth. S&P also noted that free cash flow at Amazon and Meta will be under pressure in the next two years, and that Chinese competition on base models, semiconductors, and infrastructure is growing.
- Nvidia partnership target
- 500 $B
- Big Tech 3-yr AI spending (floor)
- 1000 $B
- Alphabet share issuance
- 85 $B
Data from the article text
Nvidia's open-source push
Separately, The Information reported on August 11 that Nvidia is developing Nemotron 4, a new family of open-source AI models intended to rival top open-source models globally. The largest Nemotron 4 model is expected to have at least 1 trillion parameters, according to multiple employees working on the project. Nvidia has not set a release date and has yet to complete final training, though employees said the model could be ready as early as late fall.
On the same day, Nvidia unveiled Nemotron 3.5 Lightning, aimed at code review, tool use, security alert monitoring, billing inquiries, and other tasks. The company also released NeMo Switchyard, an open-source model-routing library designed to automatically direct AI tasks to the most suitable models. Nvidia last month formed a coalition with other companies to develop and share tools for AI safety and cybersecurity, and signed an open letter with Microsoft backing open-weight models to prevent innovation from drifting overseas.