NVIDIA's AI chip empire, built on financial guarantees and revolving financing, is being methodically replicated by its fiercest rivals. Alphabet Inc. (GOOG) and Broadcom Inc. (AVGO) are using their own balance sheets as weapons, adopting NVIDIA's script to make aggressive inroads in the AI computing power market.
Alphabet's Strategy: Mirroring NVIDIA's Tactics
Alphabet is launching the most direct challenge yet to NVIDIA's dominance in AI chips by employing the latter's own playbook. According to a June 18 report, Alphabet is replicating NVIDIA's customer lock-in strategies by providing financial guarantees for data center projects and using revolving financing to leverage chip purchases. Backed by an $85 billion equity financing plan, the company is aggressively competing for external computing clients.
Broadcom's Parallel Path
Concurrently, Broadcom is pursuing a similar route. It has partnered with Apollo and Blackstone to establish a $35 billion AI computing power financing platform. By using its own credit to provide deficiency guarantees for senior bonds, Broadcom is bundling chip manufacturers, private credit, and AI computing demand into a new financing closed loop, directly targeting NVIDIA's over 90% market share in AI chips.
The core logic of this challenge is that in an AI race where computing power scarcity is a decisive variable, whoever helps clients solve financing problems wins chip orders. Analysis indicates the market significance of these moves: NVIDIA's long-standing business model of using financial guarantees to lower data centers' financing costs and using revolving investments to drive chip purchases is being systematically transplanted by Alphabet and Broadcom.
This trend not only signals a reshaping of the competitive landscape in the AI chip market but also foreshadows a new industry norm of deep integration between private credit and AI infrastructure financing, posing a substantive challenge to NVIDIA's market position.
Financial Guarantees for Chip Orders
Alphabet is systematically copying NVIDIA's core commercial strategy: using financial guarantees to help data centers secure lower-cost debt financing while ensuring part of its invested capital flows back through chip purchases via "revolving financing" arrangements.
A representative case is the Lake Mariner project on the southern shore of Lake Ontario in New York. Alphabet provided a $3.2 billion financial guarantee for this AI data center cluster, developed jointly by TeraWulf and cloud service provider FluidStack, which is backed by Alphabet. The computing power will be leased to AI giant Anthropic. Nazar Khan, TeraWulf's co-founder and CTO, stated, "These capital-rich firms all believe the market around computing power will create enormous value, and they don't want to be left behind."
Alphabet's financial guarantee initiatives extend further. According to the report, the company has also provided backing for another Anthropic project—the $7 billion River Bend project near Baton Rouge, Louisiana—and offered an additional $1.4 billion financial guarantee for an AI computing lease project in Colorado City, Texas.
On a broader strategic level, Alphabet recently reached a $5 billion agreement with Blackstone to form a new cloud services company, directly competing with NVIDIA-backed cloud providers CoreWeave and Nebius, which exclusively use NVIDIA hardware. Bernstein technology analyst Stacy Rasgon noted, "They are clearly more opportunistic and aggressive in monetizing their assets than a few years ago. But a few years ago, this opportunity simply didn't exist. Now, all we hear is that there isn't enough computing power."
Direct TPU Sales: From Internal Tool to Competitive Weapon
The commercialization path of Alphabet's self-developed AI chip, the Tensor Processing Unit (TPU), has undergone a three-stage evolution: from internal use to external availability, and now to direct sales.
According to reports, this journey began in 2013. Then-Alphabet AI researcher and current DeepMind Chief Scientist Jeff Dean conducted a "thought experiment" while studying speech recognition: deploying a speech model to 100 million users would require computing power equivalent to twice the number of servers Alphabet had at the time. His conclusion was, "We need to build specialized hardware."
Initially, TPUs were for Alphabet's internal use, powering AI functions for its search engine and other products. With the explosion of external computing demand, Alphabet began offering TPUs to external clients through its Cloud platform, driving rapid growth in its cloud business. In May of this year, Alphabet further announced plans to sell TPUs directly to customers and launched its first TPU product specifically optimized for inference workloads, expected to compete head-on with NVIDIA's new Groq 3 LPU.
Mark Lohmeyer, Vice President of Cloud AI and Computing Infrastructure at Alphabet, stated that chips optimized for inference, combined with Alphabet's improvements in cross-system chip coordination, have attracted new customer segments previously not considering TPUs. This includes long-time Alphabet Cloud user Citadel Securities, which recently began using TPUs for some research software workloads. Its CTO, Josh Woods, reported a 30% reduction in the cost of running key workloads and speed improvements of up to four times.
Broadcom's Bet on Deficiency Guarantees
Simultaneously, Broadcom is leveraging its own credit to gain market share in the AI chip arena, pioneering a new financing model that bundles chip manufacturers, private credit, and AI computing demand.
Broadcom, Apollo, and Blackstone jointly announced the formation of the "AI XPV Platform" last week. The inaugural transaction is valued at $35 billion, financing the expansion of over 1 gigawatt of computing infrastructure for Anthropic, marking one of the largest private credit Special Purpose Vehicle (SPV) deals to date. The core vehicle is an SPV led by Apollo's Atlas SP Partners, which purchases chips and leases them to Anthropic, using the lease payments as a source of debt repayment.
The debt structure is divided into three tiers: $600 million in A1 notes sold to banks at a spread of 100 basis points over Treasury rates; $24 billion in A2 notes sold to institutional investors with a 5.75% yield; and $4.5 billion in subordinated notes not backed by Broadcom, offering a yield as high as 8.5%. Additionally, Atlas SP Partners provided an $800 million equity layer. The key to low-cost financing for the senior bonds lies in the "deficiency guarantee" provided by Broadcom—if Anthropic fails to meet its obligations and proceeds from chip disposal are insufficient to cover principal and interest, Broadcom will cover the losses for A1 and A2 tier investors.
As recently as March this year, Broadcom CEO Hock E. Tan was cautious about using Broadcom's balance sheet for such guarantees but subsequently changed his stance. The driving pressure for this shift is that NVIDIA has already employed similar supplier financing methods to accelerate chip sales; failure to follow suit could leave Broadcom behind in the AI chip competition. Tan positioned this collaboration as "the first of many transactions to come" and plans to provide financing for over 20 gigawatts of computing power for frontier AI labs through this platform by 2028, with potential chip procurement possibly reaching $700 billion.
NVIDIA's Moat: CUDA Ecosystem and Vendor Lock-in
Despite the challenges from Alphabet and Broadcom, NVIDIA's market position remains resilient. Behind its over 90% share of the AI chip market lies a powerful ecosystem barrier composed of plug-and-play interconnected hardware and the easy-to-use CUDA programming library.
Reports indicate that some emerging cloud service providers fear that deviating from NVIDIA's complete hardware stack might risk losing their chip allocation, a dilemma referred to in the industry as vendor lock-in. Bessemer Venture Partners partner Adam Fisher stated, "Not all NVIDIA cloud service providers would say this; some say NVIDIA gives them everything they need. But there are also those who desperately want other options but cannot get them from other suppliers."
Regarding competitors' challenges, NVIDIA CEO Jensen Huang has publicly expressed equanimity. In a podcast in April, he stated that NVIDIA holds a significant lead over Alphabet and other custom chip (ASIC) manufacturers, questioning the cost advantage of TPUs: "I'd love to hear them prove the cost advantage of TPUs; it makes no sense to me." He also emphasized that Anthropic is Alphabet's only significant external TPU customer.
However, Amin Vahdat, Alphabet's CTO for AI infrastructure, holds a different view. He stated that he does not focus on confrontation with NVIDIA or any competitor, noting that NVIDIA is both a competitor and an important partner, as Alphabet data centers also use NVIDIA GPUs. "For me and for us, this is not a zero-sum game. The market demand is large enough."
Trillion-Dollar Capital Expenditure Drives New Financing Landscape
The moves by Alphabet and Broadcom reflect the broader industry backdrop of rapidly膨胀的 AI infrastructure financing demand.
According to a Morgan Stanley forecast, capital market financing for the U.S. AI sector is expected to reach $400 billion, potentially exceeding $1 trillion by 2028, matching an estimated $1.8 trillion in capital expenditure needs over the next two years. Traditional banks are already showing significant strain in digesting large-scale AI-related debt, making private credit a crucial alternative channel.
Alphabet announced plans this month to raise $85 billion in equity financing, primarily to support AI infrastructure construction. Other recent comparable cases include Meta completing a $27.3 billion SPV transaction for its Hyperion data center in Louisiana, with private credit provided by Blue Owl and arranged by Morgan Stanley, supported by a Meta guarantee-like structure. Amazon completed an issuance of approximately C$14 billion (about $10 billion) in the Canadian bond market, setting a record for the largest single issuance in the history of the Canadian dollar bond market.
For Anthropic, this arrangement is the clearest signal yet of its shift towards building its own computing power supply, reducing reliance on cloud service providers like Alphabet or Amazon. According to reports, Anthropic has arranged for Alphabet to provide backing for leases at five of its data center facilities, ensuring physical space for chip installation.
This series of arrangements indicates that in the battle for AI computing power, the interests of chip manufacturers, tech giants, and private capital are becoming intertwined at an unprecedented depth and speed. The "financial guarantee for market share" model pioneered by NVIDIA has become a new paradigm that the entire industry is racing to emulate.
Comments