March 5, 2025 – At the Mobile World Congress in Barcelona, Vishal Sharma, Vice President of Amazon’s General AI, stated that “almost no business unit at Amazon is untouched by AI.” Additionally, he dismissed the notion that open source models can reduce computational requirements.
During a startup conference, Sharma explained that Amazon is leveraging its proprietary basic model to advance AI deployment in various domains, including Amazon Web Services, warehouse automation robots, and the Alexa smart assistant.
“We now have approximately 750,000 robots performing tasks ranging from item picking to autonomous operations. Alexa is arguably the world’s most widespread household AI product… It can be said that generative AI has penetrated every business segment of Amazon,” Sharma noted.
Furthermore, Amazon is collaborating with Anthropic, an AI startup it invested $8 billion in, to build a large AI computing cluster based on the Trainium 2 chip. Meanwhile, Elon Musk’s xAI has unveiled Grok 3, which is trained using a massive datacenter in Memphis equipped with about 200,000 GPUs.
When queried about whether the surge in AI computing resources will be a long-term trend, Sharma opined, “Computing power will remain a central topic of discussion in the future.”
Regarding the emergence of numerous open source AI models from China, Sharma did not perceive them as a challenge to Amazon. Instead, he suggested that Amazon is willing to run DeepSeek and other models on AWS. “We advocate for open choice… As long as it benefits our customers, we will actively embrace new technologies,” he said.
According to a TechCrunch report, when asked if the release of ChatGPT in late 2022 had left Amazon unprepared, Sharma disagreed. “I don’t agree with that. Amazon has been deeply involved in AI for 25 years. Take Alexa for example, it runs at least 20 types of AI models… Our language models have been supported by billions of parameters for a long time. This is not a new field for us; we have been continuously investing in it,” he stated.
