But why this change of heart? Shouldn’t Alibaba Cloud, given its e-commerce roots so similar to Amazon’s, naturally follow the AWS model? The answer, it turns out, lies in a nuanced understanding of evolving market demands and the unique strengths of each cloud provider.
Azure, Microsoft’s cloud offering, boasts impressive profitability, with operating margins significantly higher than both AWS and Google Cloud. This financial success, however, is deeply intertwined with Microsoft‘s sprawling enterprise software ecosystem.
Microsoft leverages its existing dominance in areas like Outlook, Teams, and Power BI to create a tightly integrated and compelling cloud package for enterprise clients. As one Fortune 50 enterprise noted, quoted in this article, Azure’s pricing is difficult to negotiate against because of their reliance on Microsoft software. This ecosystem advantage is, unfortunately for Alibaba, difficult to replicate.
While Chinese tech giants like Tencent and Alibaba itself are striving to build similar ecosystems with products like Tencent Meeting and DingTalk, they have yet to achieve the same level of comprehensive integration and user stickiness. It’s not a matter of desire, but of insurmountable existing advantages.
AWS, the undisputed market leader in cloud computing, built its empire on providing scalable infrastructure for customers to build their own solutions. However, this “build-it-yourself” philosophy is increasingly out of sync with the demands of today’s fast-paced business environment.
Clients now crave pre-built, readily deployable solutions, particularly in the realm of artificial intelligence. This is where AWS has faltered, especially compared to Azure and Google Cloud, who have been quicker to embrace and integrate AI tools. The partnership between OpenAI and Microsoft Azure, for example, gave Azure a significant early advantage in the AI race. AWS, lacking its own large language model or equivalent to Microsoft’s Copilot, has struggled to keep pace.
This shift in the competitive landscape is reflected in growth rates: AWS’s growth is hovering around 15-20%, while Azure and Google Cloud are experiencing growth rates of 35%+ and 30%+, respectively.
Google Cloud’s resurgence is fueled by a potent combination of factors: cutting-edge AI models, advanced hardware, and a strategic focus on enterprise solutions. Google’s Gemini and its stake in Anthropic, providing access to both Claude and Gemini, offer customers a compelling choice of leading large language models.
Furthermore, Google’s Tensor Processing Units (TPUs) provide a significant cost advantage in AI training scenarios. These TPUs are not just marginally better, but, as the article notes, roughly 30% more cost-effective than GPUs in training scenarios. Even Apple leverages Google’s TPUs for some compute tasks, a testament to their superior performance.
Alibaba’s Echo: Compute, Models, and Chips
Alibaba Cloud’s strategic shift mirrors Google’s three-pronged approach. First, they are aggressively expanding compute capacity, rivaling ByteDance and surpassing Tencent in capital expenditure on computing power. Second, they are heavily investing in large language models, with Qwen (千问) emerging as a top-tier LLM in China. Finally, they are developing their own chips (PingTouGe) and GPU pooling systems to optimize resource utilization and reduce reliance on external vendors.
These GPU pooling systems have already yielded impressive results, increasing GPU utilization from 13.3–33.9% to 48.1% in trials and cutting the required number of H20 GPUs by a staggering 82%.
Alibaba Cloud’s decision to emulate Google Cloud’s strategy signals a significant shift in the cloud computing landscape. It’s not about simply following the biggest or the richest player, but about aligning with the most forward-looking and adaptable approach. Whether this strategy will propel Alibaba Cloud to new heights remains to be seen, but one thing is certain: the cloud wars are far from over, and the next chapter promises to be filled with innovation, competition, and strategic realignments.
