Key Takeaways
- • Ares Management Corp. is avoiding financing deals that back graphics processing units (GPUs) used to run sophisticated AI models, according to Bloomberg.
- • The alternative asset manager's concern is that GPU-backed assets could become obsolete shortly after release, leaving the firm holding depreciated collateral.
- • Ares is instead directing capital toward data centers that house AI compute hardware, shifting its exposure from the chips to the infrastructure.
Summary
Ares Management Corp. is steering clear of financing transactions that would back the high-powered processing units powering advanced AI models, according to Bloomberg. Instead, the firm has chosen to invest in the data centers that house these chips, a strategy that places it in the physical infrastructure layer of the AI buildout rather than the hardware itself.
The rationale centers on asset obsolescence risk. A person close to the firm, who asked not to be identified because the information is private, said Ares is concerned that if it backed graphics processing units, it could be left holding assets that become obsolete not long after they are released. Given the rapid product cycles in AI accelerators, lenders secured against such hardware face the risk that collateral value erodes faster than loan amortization schedules.
The positioning reflects a broader distinction within private credit and alternative investing between financing compute hardware, which carries short useful lives, and financing data center real estate and infrastructure, which typically supports longer-duration contracts. The source did not provide specific deal values, fund sizes, or timelines for Ares' data center investments.
The rationale centers on asset obsolescence risk. A person close to the firm, who asked not to be identified because the information is private, said Ares is concerned that if it backed graphics processing units, it could be left holding assets that become obsolete not long after they are released. Given the rapid product cycles in AI accelerators, lenders secured against such hardware face the risk that collateral value erodes faster than loan amortization schedules.
The positioning reflects a broader distinction within private credit and alternative investing between financing compute hardware, which carries short useful lives, and financing data center real estate and infrastructure, which typically supports longer-duration contracts. The source did not provide specific deal values, fund sizes, or timelines for Ares' data center investments.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.