Nvidia and Wall Street Build a $500B AI Financing Platform
Chamath Palihapitiya
· Chamath Palihapitiya
· 16 августа 2026, 15:31
· ⏱ 4 мин чтения
| Читать в Substack ↗
Резюме
AI compute is becoming a financeable asset class: Nvidia and six Wall Street firms plan platforms to mobilize over $500B for Nvidia-based data centers, using GPU lease cash flows as collateral. Meanwhile, Anthropic's Decart deal targets the cost side of AI inference, and the app layer keeps raising large rounds — the market takeaway is that AI infrastructure demand is expected to remain durable enough to support securitization, with cost-optimization and non-Nvidia compute options gaining strategic value.
•On August 10, Nvidia signed MOUs with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR; the six platforms aim to mobilize more than $500B from outside investors, though capital commitments and revenue contracts have not yet been signed.
•KKR's Waldemar Szlezak said A100s still earn at high utilization six or seven years in, which supports the logic of securitizing GPU lease cash flows across investor classes.
•Larry Fink compared the planned financing to the birth of the mortgage-backed securities market in the 1970s and cited roughly $9 trillion sitting in U.S. money market funds as potential capital.
•Anthropic is in talks to buy Decart for about $6B, a 50% premium to Decart's nearly $4B May valuation; Decart optimizes models across Nvidia GPUs, Google TPUs, Amazon Trainium, and AMD chips.
•If closed, Decart would be Anthropic's fourth acquisition of 2026 and its largest, and Anthropic's widely anticipated IPO is aiming for a $2T valuation.
•River AI raised $1.1B led by General Catalyst and AMP PBC with Nvidia and AMD Ventures; Lovable raised $400M at a $13.3B valuation; Cognition is discussing a $40B round; Blacksmith's Series B valued it at $550M, up from $60M in under a year.
Nvidia signed MOUs with six financial firms to build platforms mobilizing over $500B for Nvidia-based data centers, and A100s reportedly still earn at high utilization six or seven years in — this sup
Nvidia signed MOUs with six financial firms to build platforms mobilizing over $500B for Nvidia-based data centers, and A100s reportedly still earn at high utilization six or seven years in — this supports the durability/financeability of Nvidia GPUs and could extend the AI capex cycle.
Risk: The MOUs are non-binding; if capital commitments or lease securitization fail to materialize, the financing tailwind is delayed.
Decart's Optimization Stack tunes models across Nvidia GPUs, Google TPUs, Amazon Trainium, and AMD chips; a hardware-agnostic performance layer that lowers AI inference costs could make AMD's data cen
Decart's Optimization Stack tunes models across Nvidia GPUs, Google TPUs, Amazon Trainium, and AMD chips; a hardware-agnostic performance layer that lowers AI inference costs could make AMD's data center GPUs more competitive for AI workloads.
Risk: Article provides no comparative performance data; Nvidia's ecosystem and faster annual cadence remain dominant.
This newsletter, published August 16, 2026,
features Chamath Palihapitiya
discussing NVDA, AMD.
2 trade ideas extracted by AI with direction and confidence scoring.