AI Strategies Reshape Amazon and Salesforce

Amazon.com Inc. is accelerating investment in artificial intelligence across its Amazon Web Services (AWS) cloud business, even as investors react cautiously to soaring capital expenditures. The company has indicated it could spend as much as $200 billion in capital expenditure in 2026, largely tied to building AI infrastructure and expanding capacity.

According to commentary cited in recent reports, a slowdown in AWS cloud growth and higher AI-related spending have weighed on Amazon’s share price. The stock is up about 8% over the past year but has fallen 6.3% since Jim Cramer highlighted it as a long-term idea on his Mad Money show in January 2025. It also dipped 5% in February after the 2026 capex outlook was disclosed, prompting DA Davidson to downgrade the stock to Neutral from Buy.

Despite market concerns, Cramer has framed Amazon as a “multi-year move,” arguing it should not be judged on quarterly performance. Other commentators have noted that while Amazon has significant AI potential, some alternative AI stocks may offer higher near-term upside and lower downside risk.

On March 13, AWS and Cerebras Systems unveiled a collaboration to deliver what they describe as the world’s fastest AI inference solutions on Amazon Bedrock, with launch expected in the coming months. The offering is based on a “disaggregated inference” model that splits workloads between AWS Trainium-powered servers and Cerebras CS-3 systems.

Under this design, AWS Trainium handles the highly parallel prefill stage of AI inference, while Cerebras CS-3, which offers significantly higher memory bandwidth than traditional GPUs, is dedicated to the memory-intensive decode stage. The systems are connected via AWS’s Elastic Fabric Adapter and secured with the AWS Nitro System, aiming to deliver high-speed data transfer with enterprise-grade isolation.

This marks the first time a cloud provider has integrated Cerebras hardware into a disaggregated inference service. AWS plans to broaden the offering later in 2026 by running leading open-source large language models and its own Amazon Nova models on the combined architecture.

Amazon CEO Andy Jassy has tied these infrastructure moves to a much larger long-term AI opportunity. In an all-hands meeting reported by Reuters, he said he expects AI to more than double AWS’s cloud computing addressable market estimates, implying potential annual sales in the ballpark of $600 billion, about double prior internal estimates.

Jassy emphasized that Amazon is “not just spending the $200 billion of capex because we’re hoping AI is going to be big,” but because faster AWS growth requires near-term investments in land, power, buildings, chips, servers and networking gear. AWS revenue has already risen from a mid-$40 billion run rate in 2020 to $128.7 billion in 2025, up 19% year over year.

Salesforce, another of Jim Cramer’s highlighted AI stocks, is facing a different market reaction as it pursues AI-led growth and large-scale capital returns. The company’s shares are down 29% over the past year and more than 39% since Cramer discussed the stock in February 2025, with most of the decline occurring in 2026; the stock is down 21% year to date.

Cramer has attributed much of the turbulence to concerns over AI’s impact on the broader software sector, while maintaining that Salesforce’s Agentforce AI business is performing well. He previously cited positive client feedback on Agentforce and argued in early 2025 that weakness in the stock could present a buying opportunity.

Operationally, Salesforce reported strong Q4 fiscal 2026 results, with revenue of $11.2 billion, up 12% year over year, and earnings per share of $3.81, ahead of expectations. Full-year revenue reached a record $41.5 billion, up 10%, and remaining performance obligations climbed 16% to $35.1 billion, indicating robust contracted revenue.

On March 16, 2026, Salesforce launched its largest-ever accelerated share repurchase, prepaying $25 billion for about 103 million shares as part of a broader $50 billion capital return plan. The transaction is expected to reduce shares outstanding by roughly 14%, and the board also approved a further $25 billion open-market buyback authorization.

The program, funded in part by new debt, followed the company having already returned $14 billion to shareholders in the prior year. Despite the strong earnings backdrop and management’s capital-return signal, Salesforce shares fell 1.5% on March 17 to around $195, as investors weighed leverage concerns. Analysts currently maintain an overall Moderate Buy rating, with average targets near $280, while some have trimmed targets on specific product softness.

Salesforce has guided for fiscal 2027 revenue between $45.8 billion and $46.2 billion, slightly below some analyst projections, and forecast EPS of $13.11 to $13.19. The company points to AI offerings such as Agentforce and Data Cloud, which together generated $2.9 billion in recurring revenue and helped end a period of single-digit growth, as central to its strategy.