Apollo and Blackstone have teamed up to raise $35 billion for a chip-focused financing package that will help Anthropic fund the next stage of its AI buildout, according to the Financial Times. The transaction is being positioned as one of the largest private credit fundraisings of its kind, and it underscores a broader shift in how frontier AI companies are paying for compute: not only through equity and traditional venture rounds, but increasingly through structured debt designed to match the economics of chips, power, and supply-chain lead times.
At the center of the deal is Anthropic, the Claude maker that has moved quickly from research-led momentum to large-scale deployment. While Anthropic’s models and product roadmap are often discussed in terms of capability and safety, the practical bottleneck for scaling is less visible: access to sufficient compute capacity, delivered on timelines that align with training cycles and inference demand. Chips are not just an input; they are a scheduling problem. They come with procurement constraints, manufacturing and logistics delays, and—crucially—pricing volatility. Financing that is tailored to those realities can be as important as the model itself.
This is where private credit enters. Unlike venture capital, which typically absorbs uncertainty in exchange for upside, private credit can be engineered to fund specific operational needs while providing lenders with a clearer path to repayment. In the AI context, that means structuring capital around the purchase and delivery of compute infrastructure—often including GPUs or other accelerators, networking components, and the systems required to run them. The “chip financing” framing suggests the money is intended to flow toward the hardware pipeline rather than general corporate spending, aligning the funding with the tangible assets and contracted obligations that underpin the buildout.
The scale—$35 billion—signals that the market is no longer treating AI infrastructure as a niche expense. It is becoming a core category of institutional lending. Apollo and Blackstone, both major players in private credit and alternative asset management, are effectively betting that the demand for AI compute will remain durable enough to support large pools of capital, even as the industry debates the pace of future model improvements and the sustainability of capex-heavy growth.
One reason this matters is that AI infrastructure is capital intensive in ways that traditional corporate finance often struggles to accommodate. Training runs require large, concentrated bursts of compute. Deployment requires ongoing inference capacity, which can scale rapidly as products gain users. Both phases depend on hardware availability and on the ability to secure power and data-center capacity. Even when a company has strong revenue growth, the timing mismatch between cash inflows and infrastructure outlays can be severe. Debt can bridge that gap—if it is structured correctly.
Private credit also offers a different risk profile than equity. Equity investors share in upside but dilute ownership and can be sensitive to valuation cycles. Credit investors, by contrast, focus on cash flows and collateral or contractual protections. In a chip financing arrangement, lenders may seek mechanisms that reduce uncertainty: covenants tied to deployment milestones, security interests in certain assets, or repayment schedules linked to contracted revenue streams. The exact structure is not fully detailed in the available reporting, but the emphasis on “chip financing” implies a level of specificity that goes beyond generic corporate borrowing.
For Anthropic, the strategic value is straightforward: accelerate infrastructure acquisition without waiting for slower fundraising windows. For lenders, the appeal is more nuanced. AI compute demand is not evenly distributed across the industry; it concentrates among companies that can convert hardware into usable models and services. Anthropic’s position—building Claude and expanding its enterprise and consumer footprint—makes it a credible candidate for sustained compute spend. In other words, the lenders are not simply financing chips; they are financing the conversion of chips into product capability and revenue.
This is also a bet on the durability of the AI supply chain. The last few years have taught the industry that compute availability is constrained by more than demand. It depends on semiconductor manufacturing capacity, packaging and testing throughput, and the ability of system integrators to deliver complete racks and clusters. Even if chip prices stabilize, lead times can remain long. Financing that supports procurement during those lead times can reduce the risk that a company’s roadmap slips because it cannot secure hardware when it is available.
There is another layer: the economics of AI are increasingly shaped by infrastructure efficiency. As models grow, the cost per token and the cost per inference become central to profitability. Companies that can secure compute at favorable terms—or at least avoid the worst-case pricing spikes—gain leverage. Private credit deals like this can be interpreted as a way to lock in financing conditions that are more predictable than spot-market procurement. If the financing is structured around specific purchases, it can also reduce the need for constant renegotiation as hardware costs change.
The involvement of Apollo and Blackstone adds credibility and signals institutional confidence. These firms have deep experience in underwriting complex credit structures across sectors, including leveraged finance, real estate, and infrastructure. Their participation suggests that AI infrastructure lending is moving from experimental to repeatable. When two of the most prominent alternative managers coordinate on a deal of this magnitude, it can influence how other lenders view the category—potentially encouraging more capital to follow.
Still, the deal raises questions about how AI financing will evolve. One concern is whether private credit could become a dominant funding channel for frontier AI, potentially changing incentives. Equity investors often push for aggressive growth and accept dilution as the price of scaling. Credit investors, however, may prefer steadier execution and clearer repayment pathways. That could influence how companies plan their capex cycles, how quickly they expand deployments, and how they manage risk.
Another question is how lenders evaluate technology risk. AI performance improvements can be rapid, but so can shifts in competitive dynamics. A company might invest heavily in compute and then face a scenario where model demand grows slower than expected, or where a competitor’s approach changes the market. Credit underwriting must therefore account for demand uncertainty. In practice, lenders may rely on a combination of factors: existing customer traction, contracted enterprise usage, the company’s ability to monetize inference, and the likelihood that compute investments translate into ongoing service capacity rather than one-time training runs.
The “chip financing” framing also hints at a broader trend: the financialization of the AI supply chain. Historically, supply chain financing focused on inventory and receivables. Now, it is extending to high-value, long-lead components that determine whether AI systems can be built at all. This is not merely a funding story; it is a structural shift in how capital markets interact with technology procurement. Chips are becoming a financial asset class in their own right—not because they are traded like commodities, but because the ability to secure them is increasingly mediated by sophisticated financing.
From an industry perspective, the deal may also affect how other AI companies approach capital strategy. If Anthropic can scale compute through private credit at a scale previously associated with large infrastructure projects, competitors may seek similar arrangements. That could intensify competition not only for talent and model research, but for financing capacity. In turn, lenders may develop specialized AI infrastructure funds, partnerships with data-center operators, and frameworks for underwriting compute procurement.
There is also a potential second-order effect on data centers and power. Compute is not just chips; it is electricity, cooling, and physical space. Large AI buildouts often require coordination with hyperscalers, colocation providers, or dedicated facilities. If chip financing accelerates hardware acquisition, it can increase pressure on power availability and on the timeline for data-center readiness. That could create new financing opportunities for grid upgrades, transformer capacity, and facility expansion—areas where private credit has traditionally been active.
In that sense, the $35 billion figure is not only about chips. It is about the entire stack required to run modern AI systems. Even if the financing is labeled “chip,” the downstream impact likely includes systems integration, networking, and the operational readiness needed to deploy clusters. The lenders’ exposure may therefore extend beyond semiconductor procurement into the execution risk of building and operating infrastructure.
A unique angle on this deal is how it reflects the maturation of AI from a research frontier into an industrial process. Early AI funding cycles were dominated by equity and venture capital, with investors underwriting the possibility of breakthrough performance. As companies move into production, the bottleneck becomes less about whether the model can work and more about whether the company can sustain compute at scale. Private credit is well suited to that phase because it can be aligned with operational milestones and cash generation.
That does not mean the risks disappear. AI infrastructure lending still faces uncertainties: regulatory changes, export controls affecting chip availability, shifts in demand, and the possibility that hardware generations become obsolete faster than expected. Lenders mitigate these risks through structure—shorter tenors, refinancing options, or covenants that allow adjustments if procurement plans change. But the fundamental challenge remains: the technology cycle is fast, and credit investors must ensure that the financed assets retain value or that repayment does not depend on optimistic assumptions.
For Anthropic, the immediate benefit is speed. In AI, speed is not just about being first; it is about maintaining momentum across training and deployment. If compute capacity is secured earlier, the company can iterate more frequently, test more variants, and respond to user feedback with shorter cycles. That can translate into better product quality and stronger retention—factors that matter for monetization and, indirectly, for creditworthiness.
The deal also highlights how the AI ecosystem is increasingly interconnected with capital markets. When Apollo and Blackstone raise a pool of private credit specifically for chip financing, they are effectively translating investor appetite for yield into real-world infrastructure capacity. That can help stabilize the funding environment for AI companies, but it also ties AI growth to the health of credit markets. If credit spreads widen or liquidity tightens, the cost of financing could rise, potentially slowing infrastructure expansion. Conversely, if credit conditions improve, AI capex could accelerate further.
In the near term, investors and industry watchers will likely focus on how Anthropic uses the financing: whether it supports training expansions, inference capacity for new products, or both. They will also watch for signs of how the deal influences Anthropic’s competitive posture—whether it enables faster iteration,