Nvidia internal memo: Major US bank looking for Amazon cloud alternative, AWS responds

Capital One is reportedly exploring alternatives to Amazon Web Services. This change will help the US banking major manage rising expenses associated with its artificial intelligence (AI), a report claims. According to an internal Nvidia document obtained by Business Insider, the bank is concerned about its cloud-computing relationship with Amazon and is "looking to control costs."In an internal memo, the Nvidia employee revealed that the bank expects demand for specialised chips and reasoning models to increase significantly. The employee also noted in an internal email that Capital One believes “the costs in AWS will soon get out of hand,” leading the financial institution to discuss different infrastructure options with Nvidia representatives at a recent tech conference.In a statement to Business Insider, a Capital One spokesperson denied the claims and wrote: “We continue to be committed to AWS as our predominant strategic cloud partner."Meanwhile, an AWS spokesperson said: “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. 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.”As per Nvidia’s memo, the discussions included the possibility of building an “AI factory,” an in-house data centre that enables a company to train and run its own models instead of renting computing power from a third party. Nvidia suggests that banks can use this private infrastructure for tasks such as fraud detection, customer support and algorithmic trading.Another option mentioned in the alleged meeting was "neoclouds," which are newer cloud providers explicitly focused on AI workloads. These companies, such as CoreWeave and Lambda, often use Nvidia hardware and offer more specialised services than the broad computing range provided by Amazon. Nvidia has also been working with these startups to reduce its dependence on large cloud companies.The Capital One situation shows a critical pattern in the AI boom. Companies are rushing to adopt generative AI while also trying to reduce rising cloud costs. This new technology is up-and-coming, but creating and running AI models and AI applications can be costly.Major companies also regularly review their cloud spending, including different setups that more and more use multiple providers. 43% of companies use more than two public cloud providers, according to a recent RBC Capital report.Capital One isn't the only company complaining about AWS costs. A previous report from October claimed that AI startups are increasingly shifting traditional AWS spending to AI models and developer tools, according to internal Amazon documents from March and July.The documents said that 90% of startups in Radical Ventures' portfolio were building primarily on competing clouds due to AWS costs. Amazon also noted internally that more customers were seeking neocloud providers, which let them rent GPU power in small amounts and pay only for what they use.At that time, an AWS spokesperson told Business Insider that it was “using old data to reach outdated conclusions,” pointing out that“AWS remains the top choice for startups to build because we offer the best core services as well as the most innovative and powerful generative AI offerings.”