NVIDIA's $100B+ M&A Empire Expands as Musk's Terafab Chip Ambitions Emerge
nashnova research
Nvidia signed over $140 billion in deals in two months and holds nearly $100 billion in equity stakes — Jensen Huang is spending real money to push his customer base from 'three buyers own half the revenue' toward a thousand-model ecosystem, while Musk's Terafab plan may hand Intel a turnaround lifeline.
$140 billion in two months — what is Nvidia racing against?
Per The Information, Nvidia signed over $140 billion in deals in roughly two months, including a $105 billion credit guarantee. Its total equity-investment portfolio now approaches $100 billion.
This means → Nvidia is converting profit-sheet cash into supply-chain leverage at a pace that goes well beyond routine financial investing.
The core anxiety: in the six months through July, three customers accounted for 44% of total sales. In plain terms = nearly half of Nvidia's revenue sits in three buyers' hands — any one of them pulling back would hit the income statement hard.
Where is the money going?
Robotics and edge AI: Nvidia is in talks with humanoid-robot maker Figure about an additional ~$1 billion investment. Figure is raising at a roughly $38 billion valuation.
AI models: Nvidia spent $6 billion to license and acqui-hire Poolside staff, advancing its own Nemotron open-source model. It is also negotiating a ~$2.5 billion investment in Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, and has invested in Anthropic, xAI, and Reflection AI.
AI platforms: Nvidia expressed acquisition interest in AI-model marketplace OpenRouter and AI search company Perplexity. Stripe ultimately acquired OpenRouter for $8 billion — Nvidia lost out because it needed more evaluation time. This reflects the trigger for Huang's faster deal pace afterward.
Data-center infrastructure: Nvidia injected $3 billion into SB Energy to support large-scale data-center development for OpenAI, and provided a $105 billion credit guarantee for OpenAI's lease. Huang said the campus could house roughly $600 billion worth of Nvidia compute.
Why is Huang personally backstopping OpenAI's lease?
Huang's own words: "Frontier AI labs have enormous demand for training and inference compute, but many companies are growing faster than their balance sheets and long-term credit profiles can support."
In plain terms = AI companies burn cash faster than they build creditworthiness — banks may not lend. Nvidia is using its own credit to vouch for them, which ultimately ensures those customers can keep buying Nvidia chips.
This means → Nvidia has shifted from "selling the shovels" to "lending the gold-rushers money to buy shovels." The alignment of interests deepens, but so does the risk exposure.
What does Musk's Terafab plan have to do with Intel?
Musk's large-scale chip-manufacturing project, Terafab, is drawing market attention. Wedbush analyst Matt Bryson noted that Intel remains a possible manufacturing partner for the project.
This means → if Musk picks Intel as the foundry partner, it would provide a material boost to Intel's contract-manufacturing business — one of the few "external mega-order" possibilities on Intel's turnaround path.
Terafab is still in its early stages. Whether it materialises is one of the key checkpoints for the next phase of the AI-compute landscape.
What are the risks in this expansion wave?
PwC's tech-sector lead Dallas Dolen estimates global data-center buildout will reach roughly $31 trillion over the next 25 years — but flags funding, political, and overcapacity risks along the way.
Nvidia currently holds $99 billion in cash and marketable securities — ample firepower.
An AI-infrastructure investor familiar with Nvidia's M&A moves put it this way: "If I were in Jensen's war room, I'd do everything I could to tilt toward a world with thousands of models serving thousands of use cases." This reflects an industry-wide consensus on betting on ecosystem breadth — yet the wider the bet, the higher the odds of stumbling into an overcapacity trap.
市场有风险，内容仅供研究参考，不构成投资建议。