Musk’s $20 Billion AI Power Play: Nvidia Funds Its Own Chips in a Wild ‘Circular’ Mega-Deal

Elon Musk’s artificial intelligence venture, xAI, is once again raising eyebrows — and capital — as Bloomberg reports the company is expanding its latest funding round to $20 billion, a mix of equity and debt that showcases both financial innovation and market froth. The deal’s structure is as unconventional as its size: rather than a traditional venture raise, xAI’s financing is built around a special purpose vehicle (SPV) designed to purchase Nvidia GPUs and rent them back to Musk’s AI company for use in its Colossus 2 data center project in Memphis.

The round, according to people familiar with the matter, includes roughly $7.5 billion in equity and as much as $12.5 billion in debt. Nvidia, already an investor in xAI, is contributing up to $2 billion in new equity—making it both supplier and financier. The debt portion is being arranged by Wall Street heavyweights including Apollo Global Management and Diameter Capital Partners, with Valor Capital reportedly leading the equity tranche. The SPV will own the GPUs outright, securing the debt against the hardware rather than xAI’s balance sheet. Over five years, xAI will rent the chips, allowing investors to earn back their capital through the lease payments.

It’s a setup that looks suspiciously like the “AI circularity problem” analysts have been warning about. Nvidia’s extraordinary rise to dominance has created an ecosystem where the chipmaker is effectively funding its customers’ ability to buy its own products—an echo of the vendor-financing loops that inflated prior technology bubbles. The logic is simple: the faster Nvidia’s partners can build compute-heavy data centers, the faster demand for Nvidia chips accelerates. Yet the feedback loop risks distorting price signals and inflating valuations untethered from real cash flow.

This is not an isolated case. Nvidia’s leadership, including CFO Colette Kress, has said publicly that the company intends to use its immense cash flow to “accelerate AI adoption” by supporting partners’ infrastructure spending. In practice, that means deals like xAI’s—where Nvidia supplies both the chips and part of the capital stack to acquire them. The result is an increasingly self-reinforcing dynamic: capital flows from Nvidia’s balance sheet into its customers’ capex, only to return in the form of product sales and market share gains.

The SPV model allows Musk to scale compute capacity without bloating xAI’s corporate leverage. Because the debt is asset-backed—secured by the GPUs themselves—creditors have recourse to high-value, liquid collateral. If xAI stumbles, the lenders still own top-tier hardware they can resell. For Musk, this structure offers immediate access to the compute power needed to train large models without giving up additional control or diluting the company further. For Wall Street, it’s a lucrative form of “hardware leasing” in an era where AI infrastructure is treated as a new asset class.

Bloomberg’s reporting adds that xAI, which has already raised about $10 billion this year, is burning through roughly $1 billion a month as it builds infrastructure to compete with OpenAI and Anthropic. The Colossus 2 project—powered by the newly purchased Nvidia GPUs—is the cornerstone of Musk’s plan to create AI systems capable of “understanding the universe.” It’s a bold vision with a steep price tag, and one that’s increasingly financed through creative structures like the current SPV arrangement.

The timing also underscores the breakneck pace of the AI infrastructure race. OpenAI recently inked a multiyear deal to use AMD chips, while Meta Platforms secured a $29 billion financing package for new data centers and Oracle tapped $38 billion in debt for infrastructure expansion. In total, U.S. technology companies have raised around $157 billion in bond markets so far this year—roughly 70% more than in 2024—as investors chase anything linked to the AI boom.

Still, this frenetic capital formation has reignited bubble comparisons. The “AI circularity” theme—where vendors, financiers, and customers all depend on one another’s continued optimism—has drawn parallels to 1999. As Bloomberg noted, much of the current enthusiasm rests on multi-year spending commitments from firms still unprofitable and scaling costs as fast as revenue. If sentiment cracks, the unwinding could be sharp.

Jensen Huang, Nvidia’s CEO, brushed off those concerns in an interview with CNBC this morning, saying he’s “excited about the opportunities in xAI” and emphasizing that the industry is only at the start of a generational computing transformation. Investors, at least for now, seem to agree: demand for Nvidia’s chips remains overwhelming, and Musk’s fundraising prowess keeps xAI in the spotlight as one of the few challengers to OpenAI’s dominance.

In the end, Musk’s deal may prove to be both a symbol and a stress test for the AI era. By pairing aggressive leverage with hardware-backed collateral, xAI’s financing model could become the blueprint for how capital-intensive AI startups build infrastructure without crushing their balance sheets. Yet it also underscores the fragility of a market propped up by vendor-financed optimism. Nvidia may be minting profits from the AI revolution—but it’s also underwriting the bets that keep the revolution spinning.

If the cycle holds, Musk will get his chips, Nvidia will get its sales, and investors will get their returns. But if the music stops, the same SPV that enabled the AI gold rush could become its collateralized cautionary tale.