Apollo in AI: GPU Financing & Data Center Infrastructure

What Apollo is

Apollo Global Management is an alternative asset manager: it raises capital from institutions and, increasingly, individual investors, and deploys it across private credit, private equity and real assets rather than public stocks and bonds. Leon Black, Josh Harris, Marc Rowan and Tony Ressler, former bankers at the defunct Drexel Burnham Lambert, founded the firm in 1990. Marc Rowan is now chairman and chief executive. Apollo went public in 2011, converted to a standard C-corporation in 2019, and joined the S&P 500 in December 2024. As of June 30, 2026, it reported roughly $1.05 trillion in assets under management, most of it in credit strategies, plus a large retirement-services business run through its Athene subsidiary. [2][1]

Credit, not buyout private equity, is now Apollo's largest business by assets, and that shift matters for why the firm shows up on the timeline: private credit funds are set up to originate large, long-duration loans against hard assets, the same shape of financing that data centers and the chips inside them increasingly need. [1]

Turning GPUs into collateral

On August 10, 2026, NVIDIA announced financing platforms with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR intended to mobilize more than $500 billion in third-party capital for AI data center buildout. Each firm runs its own platform rather than a single shared fund; NVIDIA's role is to supply the compute and the CUDA ecosystem the financing underwrites, while the six asset managers supply long-term capital, infrastructure expertise and capital-markets distribution. NVIDIA's announcement describes the goal as making its compute function as an investable asset class, with financing aimed at frontier AI labs, enterprises and AI cloud providers. [3]

Apollo's president, Jim Zelter, described the shift in terms specific to the firm's credit business: "Modern compute has emerged as a scarce, mission-critical asset class with compelling investment characteristics." That is a private-credit pitch. The loans a platform like Apollo's would write are structured against data center and compute assets whose value depends on continued demand for AI workloads, which is a different risk than the equity risk a venture investor takes on a model company. [3]

Why this belongs in an AI timeline

Atlas interpretation: Apollo builds no models and sells no AI products. It matters to the timeline because the buildout everything else here depends on, the data centers and chips behind every frontier model, is now large enough that traditional venture and big cloud provider balance sheets are not the only capital funding it. Six of the largest alternative asset managers agreeing to structure GPU-backed financing at half-a-trillion-dollar scale is a statement that Wall Street now treats AI compute the way it treats other infrastructure: as a collateral class, not just a growth bet. Whether that financing performs the way its structure assumes is a question the sources cannot yet answer; it can only note that the bet has been placed. [3]