Capital One is actively exploring alternatives to Amazon Web Services (AWS) as the financial giant grapples with escalating AI infrastructure costs, according to an internal Nvidia document obtained by Business Insider. The revelation highlights a growing tension in the AI industry between rapid adoption and cost management.
An Nvidia employee documented discussions with Capital One representatives at a recent tech conference, noting that the bank is “looking to control costs” as its AI needs expand. The internal email specifically stated: “They see their need for GPUs and reasoning models growing and the costs in AWS will soon get out of hand.” This concern is driving Capital One to evaluate alternative infrastructure solutions for its AI workloads.
The discussions centered on two key alternatives: AI factories and neoclouds. An AI factory represents an in-house data center that companies can build to train and run AI models independently, eliminating reliance on third-party compute rentals. For financial institutions like Capital One, this infrastructure supports critical operations including fraud detection, customer support, and algorithmic trading. Neoclouds, meanwhile, are emerging cloud providers—often powered by Nvidia hardware—that specialize in AI workloads rather than the broader computing services offered by AWS. Leading neocloud players include CoreWeave, Lambda, Crusoe, and Nebius.
This situation reflects a broader industry trend where companies are simultaneously racing to adopt generative AI while attempting to mitigate ballooning cloud costs. According to a recent RBC Capital report, 43% of companies now use more than two public cloud providers, indicating a strategic shift toward multi-cloud environments to optimize spending.
Capital One maintained its commitment to AWS, with a spokesperson stating: “We continue to be committed to AWS as our predominant strategic cloud partner.” AWS defended its pricing strategy, emphasizing its philosophy to “work relentlessly to take cost out of our own cost structure and to pass those savings back to our AWS customers in the form of lower prices.”
The bank isn’t alone in its concerns. Business Insider previously reported that AI startups are increasingly delaying traditional AWS spending in favor of rival clouds, with 90% of startups in Radical Ventures’ portfolio building primarily on competing platforms due to AWS costs. Nvidia declined to comment on the Capital One discussions.
Key Quotes
They see their need for GPUs and reasoning models growing and the costs in AWS will soon get out of hand
An Nvidia employee wrote this in an internal email after meeting Capital One representatives at a tech conference, revealing the bank’s concerns about escalating AI infrastructure costs on Amazon’s cloud platform.
We continue to be committed to AWS as our predominant strategic cloud partner
A Capital One spokesperson provided this statement to Business Insider, attempting to reassure stakeholders about the bank’s relationship with Amazon while not denying explorations of alternative infrastructure options.
Our pricing philosophy is to work relentlessly to take cost out of our own cost structure and to pass those savings back to our AWS customers in the form of lower prices
An AWS spokesperson defended Amazon’s pricing strategy in response to concerns about rising AI costs, emphasizing the company’s commitment to affordability and efficiency for customers.
It’s easy to lower prices, it’s much harder to be able to afford to lower prices, and at AWS we work really hard at that
The AWS spokesperson continued their defense of Amazon’s pricing approach, highlighting the company’s operational efficiency as a competitive advantage in the increasingly cost-conscious AI infrastructure market.
Our Take
Capital One’s exploration of AWS alternatives represents a watershed moment that could fundamentally reshape AI infrastructure economics. The irony is striking: Nvidia, whose GPUs power the AI revolution, is actively helping customers reduce dependence on its largest cloud customers. This reveals Nvidia’s strategic pivot to diversify its customer base beyond hyperscalers.
The emergence of neoclouds as viable alternatives suggests the AI infrastructure market is maturing rapidly. These specialized providers can offer more competitive pricing for AI workloads because they’re not subsidizing broader cloud services. For enterprises, this creates leverage in negotiations with AWS and validates multi-cloud strategies.
Most significantly, this signals that AI cost optimization is becoming as important as AI capability. Companies that master both technical implementation and cost management will gain competitive advantages. Capital One’s willingness to consider building AI factories indicates that for large enterprises, the economics may favor ownership over rental—a paradigm shift in cloud computing philosophy.
Why This Matters
This development signals a critical inflection point in the AI infrastructure market as enterprise adoption scales. The fact that a major financial institution like Capital One is openly exploring AWS alternatives underscores how AI costs are becoming a strategic concern that could reshape cloud computing dynamics.
For the AI industry, this represents both a challenge and an opportunity. While hyperscalers like AWS have dominated cloud computing for years, the specialized demands and costs of AI workloads are creating openings for neocloud providers and in-house solutions. This could accelerate the fragmentation of the cloud market and intensify competition among providers.
For businesses, Capital One’s situation serves as a cautionary tale about AI cost management. As companies rush to implement generative AI applications, they must simultaneously develop strategies to control infrastructure expenses or risk budget overruns that could undermine AI initiatives. The emergence of neoclouds offering flexible, pay-as-you-go GPU access provides alternatives that may better align with AI-specific workloads.
Longer-term, this trend could democratize AI infrastructure access by breaking the hyperscaler oligopoly, potentially enabling more companies to deploy sophisticated AI systems affordably.
Related Stories
- Nvidia CEO Jensen Huang Reveals Public Speaking Struggles Despite AI Success
- Goldman Sachs Hires Google’s Melissa Goldman as Tech Head for AI Push
- CEOs Express Insecurity About AI Strategy and Implementation
- Big Tech’s 2025 AI Plans: Meta, Apple, Tesla, Google Unveil Roadmap
- Meta and Nvidia Billionaires’ Wealth Soars $152B in AI Boom