Broadcom, Apollo and Blackstone unveil $35 billion AI XPV Platform aimed at boosting compute by 20+ gigawatts

Broadcom (AVGO) said it has teamed with Apollo and Blackstone to launch a new large-scale financing platform for artificial intelligence infrastructure, dubbed the AI XPV Platform. The announcement comes as AI developers and chipmakers face a bottleneck not just in graphics processing units, but also in power availability and data-center buildouts required to support high-end model training and deployment.

According to the report, the platform is designed to enable more than 20 gigawatts of compute capacity for “frontier AI labs” through 2028. Gigawatts are a measure of electrical power. In this context, the figure is intended to reflect the amount of power that data-center operators and AI infrastructure projects would need to run the compute required by advanced AI systems.

The initial financing is described as a $35 billion capital solution, with the first tranche framed as a launch step for the broader effort. Broadcom Chief Executive Hock Tan is quoted in market reporting as describing the initiative as combining Broadcom’s technology roadmap with long-term capital, positioning Broadcom not only as a chip and networking supplier but also as an orchestrator of the infrastructure path that AI labs need to scale.

Apollo said it would lead the $35 billion capital solution, in partnership with Blackstone and “leading global banks,” in a separate announcement posted to its investor relations site. Apollo also said the initial investment is intended to accelerate compute capacity for Anthropic as part of the platform’s broader agenda, with Anthropic identified as an early major beneficiary.

Blackstone is named as a key partner in the platform’s launch, aligning with its Credit and Insurance business in market summaries. Blackstone is commonly associated with asset-based and structured finance, while Apollo’s involvement points to credit and capital solutions aimed at infrastructure buildout timelines rather than purely equity-style funding.

Sector context matters because AI spending has increasingly shifted upstream to the physical layer of computing. As model sizes and training runs grow, the constraint frequently becomes power, cooling, and the ability to construct and connect data-center capacity quickly enough to meet demand. Deals that bundle financing with technology roadmaps are also seen as a way to reduce the time gap between AI demand and the infrastructure required to run it.

Still, the announcement leaves several practical details unclear, at least based on the publicly available posts and summaries. The parties did not spell out how the compute capacity targets would be allocated across specific data-center sites, what the financing terms would be for each deployment, or what percentage of the $35 billion would be deployed in the earliest phase versus later years through 2028. It also was not disclosed in the accessible materials which other AI labs, besides Anthropic, would be included as named clients.

For investors and industry watchers, the next test will be whether the platform translates announced capacity targets into contracted infrastructure and measurable timelines for power and site readiness. Watch for follow-on disclosures covering the first financed projects, the customers and operators involved, and any details on how Broadcom’s technology components will integrate into the funded deployments.