Following NVIDIA’s lead as a 'role model,' both Google and Broadcom have started emulating its approach, moving toward a closed-loop AI chip strategy.

$NVIDIA (NVDA.US)$The AI chip business empire built on financial guarantees and revolving financing is now being replicated one by one by its strongest competitors.$Alphabet-A (GOOGL.US)$and $Broadcom (AVGO.US)$ They are leveraging their own balance sheets as weapons, following NVIDIA's 'playbook' to aggressively break into the AI computing market.

Google is directly challenging NVIDIA’s dominance in AI chips by replicating its strategy more explicitly than ever before. According to a June 18 report by The Wall Street Journal, Google has fully adopted NVIDIA’s customer-lock-in tactics—providing financial guarantees for data center projects and using revolving financing to drive chip procurement. Backed by an $85 billion equity financing plan, Google is aggressively pursuing external computing clients.

Meanwhile, Broadcom is pursuing a similar strategy—partnering with Apollo and Blackstone to establish a $35 billion AI computing financing platform. Broadcom is providing shortfall guarantees for senior notes using its own creditworthiness, bundling chip manufacturers, private credit, and AI computing demand into a new closed-loop financing model aimed squarely at NVIDIA’s over 90% market share in AI chips.

At the heart of this challenge lies a fundamental logic: when computing power scarcity becomes the decisive variable in the AI race, whoever can help clients solve their financing needs will win chip orders. Analysts note that the market significance of these moves is clear: NVIDIA’s long-standing business model—using financial guarantees to lower data centers’ financing costs and leveraging revolving investments to drive chip procurement—is now being systematically replicated by Google and Broadcom.

This trend not only signals a reshaping of competitive dynamics in the AI chip market but also heralds a new industry norm: deep integration between private credit and AI infrastructure financing, posing a substantive challenge to NVIDIA’s market share of over 90%.

Google Replicates NVIDIA’s Playbook: Financial Guarantees in Exchange for Chip Orders

According to reports, Google is systematically replicating NVIDIA’s core business strategy—using financial guarantees to help data centers secure lower-cost debt financing, while simultaneously implementing 'revolving financing' arrangements that channel part of its invested capital back through chip procurement.

The most representative example is the Lake Mariner project, located on the southern shore of Lake Ontario in New York State. Google provided a $3.2 billion financial guarantee for this AI data center cluster, which is being developed by$TeraWulf (WULF.US)$and FluidStack, a cloud service provider backed by Google, with computing capacity leased to AI giant Anthropic. Nazar Khan, Co-Founder and Chief Technology Officer of TeraWulf, stated: "These well-capitalized companies all firmly believe that the market centered around computing power will generate immense value, and they don’t want to be left behind."

Google’s financial guarantee commitments extend far beyond this. According to reports, sources familiar with the matter revealed that Google also provided backing for another Anthropic project—the $7 billion River Bend project near Baton Rouge, Louisiana—and issued an additional $1.4 billion in financial guarantees for an AI computing leasing project in Colorado City, Texas.

On a broader strategic level, Google recently struck a $5 billion agreement with Blackstone to form a new cloud services company directly competing with$CoreWeave (CRWV.US)$and$NEBIUS (NBIS.US)$two cloud service providers exclusively using NVIDIA hardware. Stacy Rasgon, technology analyst at Bernstein, noted:

"They are clearly more opportunistic than they were a few years ago and are far more actively monetizing their own assets. But just a few years ago, this opportunity simply didn’t exist. Now, all we hear is that there isn’t enough computing power to go around."

Direct TPU Sales: From Internal Tool to External Competitive Weapon

Google’s commercialization path for its self-developed AI chip, the Tensor Processing Unit (TPU), has undergone a three-stage evolution—from internal exclusivity to external availability, and now to direct sales.

According to The Wall Street Journal, this journey began as early as 2013. Jeff Dean, then a Google AI researcher and now Chief Scientist at DeepMind, conducted a 'thought experiment' while researching speech recognition: if Google were to deploy a speech model to 100 million users, it would require computing power equivalent to twice the capacity of Google’s entire server fleet at the time. His conclusion was clear: 'We need to build specialized hardware.'

Initially, TPUs were used exclusively within Google to support AI development for its search engine and other products. As external demand for computing power surged, Google began offering TPUs to external customers via its Cloud platform, significantly accelerating the growth of its cloud business. In May of this year, Google announced plans to directly sell TPUs to customers and unveiled its first TPU specifically optimized for inference workloads—a move expected to position it in direct competition with NVIDIA’s new Groq 3 LPU.

Mark Lohmeyer, Vice President of Google Cloud AI and Compute Infrastructure, stated that chips specifically optimized for inference, combined with Google’s improvements in cross-system chip coordination, have attracted a new customer segment that had not previously considered TPUs. This includes Citadel Securities, a long-standing Google Cloud user, which recently began deploying TPUs for certain research software workloads. According to Chief Technology Officer Josh Woods, this shift has reduced operating costs for critical workloads by 30% and accelerated performance by up to fourfold.

Broadcom Bets on Shortfall Guarantees: Trading Credit for Market Share

Meanwhile, as previously reported by Wall Street News, Broadcom is sacrificing its own creditworthiness to gain market share in the AI chip sector, pioneering a novel financing model that ties together chipmakers, private credit, and AI computing demand.

Last week, Broadcom, Apollo, and Blackstone jointly announced the formation of the 'AI XPV Platform,' with an initial transaction valued at $35 billion to finance Anthropic’s expansion of over 1 gigawatt of AI computing infrastructure—making it one of the largest private credit special purpose vehicle (SPV) transactions to date. The core structure involves an SPV established by Atlas SP Partners, an affiliate of Apollo, which purchases chips and leases them to Anthropic, using lease payments as the primary source of debt repayment.

The debt structure comprises three tranches: $600 million in Class A1 notes were sold to banks at a rate of 100 basis points above U.S. Treasury yields; $24 billion in Class A2 notes were sold to institutional investors with a yield of 5.75%; and $4.5 billion in junior notes, which carry no Broadcom guarantee, offer a yield of 8.5%. Additionally, Atlas SP Partners contributed $800 million in equity. The key enabler of low-cost financing for senior debt is Broadcom’s 'shortfall guarantee'—if Anthropic defaults and proceeds from chip liquidation are insufficient to cover principal and interest, Broadcom will compensate losses for Class A1 and A2 investors.

As recently as March this year, Broadcom CEO Hock E. Tan remained cautious about using Broadcom’s balance sheet to provide such guarantees, but has since changed his stance. The shift was driven by the pressing reality that NVIDIA had already adopted similar supplier financing mechanisms to accelerate chip sales, and Broadcom risked falling behind in the AI chip race if it did not follow suit. Tan characterized this collaboration as “the first of many future deals” and plans to finance over 20 gigawatts of computing capacity for cutting-edge AI labs through this platform by 2028, potentially supporting chip purchases worth up to $700 billion.

NVIDIA’s Moat: The CUDA Ecosystem and the 'Jensen Jail'

Despite pressure from both Google and Broadcom, NVIDIA’s market position remains highly resilient. Behind its dominance—holding over 90% of the AI chip market share—is a formidable ecosystem barrier built on plug-and-play interconnected hardware and the easy-to-use CUDA programming libraries.

According to reports, some emerging cloud service providers worry that deviating from NVIDIA’s full hardware stack could jeopardize their chip allocation—a predicament the industry refers to as the “Jensen Jail.” Adam Fisher, a partner at Bessemer Venture Partners, stated:

“Not all NVIDIA cloud service providers would say this; some will say NVIDIA gives them everything they need. But others are eager for alternatives yet can’t get what they require from other vendors.”

In response to challenges from competitors, NVIDIA CEO Jensen Huang has expressed calm confidence. On a podcast in April this year, he asserted that NVIDIA maintains a significant lead over Google and other custom chip (ASIC) manufacturers, and questioned the cost advantage of TPUs: “I’d really like to hear them demonstrate TPU cost advantages—it makes no sense to me.” He also emphasized that Anthropic is the only major external customer for Google’s TPUs.

However, Amin Vahdat, Google’s Chief Technology Officer for AI Infrastructure, holds a different view. He stated that he does not focus on competing against NVIDIA or any other rival, noting that NVIDIA is both a competitor and an important partner, as Google’s data centers also use NVIDIA GPUs. “For me and for us, this isn’t a zero-sum game—the market demand is large enough.”

Trillion-Dollar Capital Expenditures Reshape AI Financing Landscape

The aforementioned moves by Google and Broadcom reflect the broader industry trend of rapidly escalating financing demands for AI infrastructure.

According to Morgan Stanley, capital market financing in the U.S. AI sector is projected to reach $400 billion and could exceed $1 trillion by 2028 to meet the estimated $1.8 trillion in capital expenditure requirements over the next two years. Traditional banks are already showing significant strain in absorbing large-scale AI-related debt, making private credit an important alternative channel.

Google announced this month its plan to raise $85 billion in equity financing, primarily to support AI infrastructure development. Recent comparable cases include:

$Meta Platforms (META.US)$ A $27.3 billion special-purpose vehicle (SPV) transaction was completed around Hyperion’s data center in Louisiana, with Blue Owl providing private credit, Morgan Stanley acting as lead arranger, and Meta offering quasi-guarantee support;$Amazon (AMZN.US)$also completed an issuance of approximately CAD 14 billion (equivalent to about USD 10 billion) in the Canadian bond market, marking the largest single issuance in the history of the Canadian dollar bond market.

For Anthropic, this arrangement represents its clearest signal yet of a strategic shift toward building its own computing capacity and reducing reliance on cloud service providers such as Google or Amazon. According to The Information, Anthropic has arranged for Google to backstop leases for five of its data center facilities, securing physical space for chip installations.

These arrangements collectively indicate that in the race for AI computing power, the alignment of interests among chipmakers, technology giants, and private capital is advancing with unprecedented depth and speed. NVIDIA’s pioneering model of exchanging financial guarantees for market share has become a new industry-wide paradigm widely emulated across the sector.

Want to select stocks or get a stock diagnosis? Curious about the opportunities and risks in your portfolio? For all your investment questions,Just ask Futubull AI!

Editor/joryn