Amazon Web Services (AWS) is in advanced negotiations for a massive cloud computing deal worth approximately $475 million over five years to provide IBM with access to NVIDIA graphics processing units (GPUs) through its cloud infrastructure, according to internal Amazon documents obtained by Business Insider.
Under the proposed agreement, IBM would leverage AWS’s Elastic Compute Cloud (EC2) servers equipped with NVIDIA’s cutting-edge AI chips to power its artificial intelligence operations. The deal specifically focuses on AI training workloads, with IBM Research already having begun training some of its AI models on EC2 servers using AWS’s machine-learning platform SageMaker.
A source familiar with the negotiations confirmed that talks are ongoing, though cautioned that no final agreement has been signed and negotiations could still fall apart. Both AWS and IBM declined to comment on the potential deal.
This partnership would significantly expand the existing collaboration between the two tech giants in the AI space. In May 2024, IBM announced it was increasing its use of AWS infrastructure across its Watson AI platform, signaling a deepening relationship between the companies.
The negotiations underscore the continued intense demand for NVIDIA GPUs, which have become the gold standard for AI training and inference workloads. Interestingly, AWS has been developing its own AI chips—Trainium and Inferentia—and has been actively encouraging cloud customers to adopt these homegrown alternatives. Amazon CEO Andy Jassy recently touted AWS’s proprietary AI chips as more appealing due to their superior price performance and energy efficiency compared to competitors. However, it remains unclear whether the IBM deal would include access to these Amazon-developed chips, suggesting that NVIDIA’s dominance in the AI chip market remains largely unchallenged.
The deal would provide another significant boost to AWS’s rapidly growing AI business. During an analyst call last month, Jassy revealed that the company’s artificial intelligence business is on track to generate “multibillion dollars” in revenue this year, growing at “triple-digit percentages year over year.” He emphasized that Amazon’s AI business is “growing three times faster at its stage of evolution than AWS did itself,” highlighting the explosive growth trajectory of AI services.
In the most recent quarter, AWS reported $27.5 billion in revenue, up 19% from the previous year. To support this growth, Amazon is investing heavily in infrastructure, with Morgan Stanley estimating over $75 billion in capital expenditure this year, with even more planned for next year, primarily focused on cloud infrastructure including GPUs, networking equipment, and data centers.
Key Quotes
AWS’s AI chips were more appealing because of their price and energy efficiency compared with other offerings.
Amazon CEO Andy Jassy made this statement recently, promoting AWS’s homegrown Trainium and Inferentia chips. However, the fact that IBM appears to be seeking NVIDIA GPUs specifically suggests that AWS’s proprietary chips may not yet be competitive enough for demanding AI training workloads.
The company’s artificial-intelligence business was on pace to generate ‘multibillion dollars’ in revenue this year and growing at ’triple-digit percentages year over year.’
Andy Jassy shared this during an analyst call last month, highlighting the explosive growth of AWS’s AI services. This IBM deal would further accelerate that growth trajectory and solidify AWS’s position as a leading AI infrastructure provider.
Amazon’s AI business is ‘growing three times faster at its stage of evolution than AWS did itself.’
This statement from Jassy provides crucial context for understanding the scale and velocity of AI adoption. Given that AWS itself experienced phenomenal growth to become Amazon’s most profitable division, this comparison underscores the unprecedented demand for AI infrastructure and services.
Our Take
This deal negotiation reveals a fascinating paradox in the AI infrastructure market: AWS is simultaneously trying to convince customers to use its proprietary AI chips while still relying heavily on NVIDIA hardware to close major deals. This suggests that AWS’s chip development efforts, while promising, haven’t yet achieved feature parity with NVIDIA’s offerings for demanding AI training workloads.
The $475 million figure also provides a valuable benchmark for understanding AI infrastructure costs. If IBM—a company with substantial technical capabilities—finds it more economical to rent GPU access through the cloud rather than purchase hardware directly, it signals a fundamental shift in how enterprises approach AI infrastructure.
Most significantly, this deal could trigger a domino effect where other large enterprises follow IBM’s lead, accelerating the concentration of AI computing power in the hands of a few major cloud providers. This consolidation raises important questions about competition, access, and control in the AI ecosystem that regulators and industry observers should monitor closely.
Why This Matters
This potential deal represents a critical inflection point in the cloud AI infrastructure market, demonstrating how even tech giants like IBM are increasingly relying on cloud providers for access to scarce AI computing resources. The $475 million price tag underscores the enormous costs associated with AI development and the premium companies are willing to pay for NVIDIA GPU access.
The negotiations highlight NVIDIA’s continued dominance in the AI chip market, even as major cloud providers like AWS invest billions in developing their own alternatives. Despite AWS’s push for its Trainium and Inferentia chips, IBM’s apparent preference for NVIDIA hardware suggests that proprietary chips still have significant ground to cover before matching NVIDIA’s performance and ecosystem.
For the broader AI industry, this deal signals that cloud-based AI infrastructure is becoming the preferred model over building in-house capabilities, even for established technology companies. This trend could accelerate consolidation in the AI infrastructure market, with AWS, Microsoft Azure, and Google Cloud capturing an increasing share of AI workloads. The deal also validates AWS’s strategy of massive infrastructure investment, potentially pressuring competitors to increase their own capital expenditures to remain competitive in the rapidly expanding AI services market.
Recommended Reading
For those interested in learning more about artificial intelligence, machine learning, and effective AI communication, here are some excellent resources:
Recommended Reading
Related Stories
- Jensen Huang: TSMC Helped Fix Design Flaw with Nvidia’s Blackwell AI Chip
- Amazon to Invest Additional $4 Billion in AI Startup Anthropic
- EnCharge AI Secures $100M Series B to Revolutionize Energy-Efficient AI Chips
- Wall Street Asks Big Tech: Will AI Ever Make Money?
Source: https://www.businessinsider.com/aws-negotiating-huge-cloud-deal-supply-ibm-with-nvidia-gpus-2024-11