Google challenges AWS and Microsoft in the cloud with an AI offensive

The company, which presents a battery of new features in cloud infrastructure, language models and productivity, announces that companies will be able to connect Gemini AI to the search engine.

Google is redoubling its AI offensive to gain market share in the battle it is waging with AWS and Microsoft in the cloud computing business. The company sees this technological revolution as an opportunity to boost its business in the cloud, where it faces AWS (Amazon) and Microsoft . Google announced today during the opening day of its Google Next event, which is being held this week in Las Vegas, a battery of innovations in infrastructure (with its first chip based on Arm designs ), natural language models, development assistants code, security and intelligent productivity.

Google Cloud is the third largest provider of public cloud services, behind AWS and Microsoft . “We are the provider that is growing the fastest,” remarked Thomas Kurian, CEO of Google Cloud , in a meeting with the media prior to the start of the event. Revenue grew 26% to $9,192 million last year, in which it left the red ink behind after posting an operating profit of $864 million. On an annualized basis, Google Cloud is a $36 billion business. “Enterprise adoption of AI expands the size of the market we can capture,” says the executive.

Kurian emphasizes that “more than 60% of start-ups and close to 90% of generative AI unicorns are Google Cloud customers.” Anthropic (a lab in which Google is a shareholder), AI21 Labs, Contextual AI, Essential AI and France’s Mistral are training or running their large language models on Google’s cloud infrastructure. In addition, he says that “hundreds of clients”, such as Deutsche Bank, Puma, Bayer, Mercedes Benz, or Verizon, use Google’s generative AI technology in their businesses. The executive has revealed that companies such as ING, Best Buy, Etsy or Home Depot are creating personalized intelligent agents capable of understanding and processing video, audio and text, which have been built on the basis of their AI platform.

Within the framework of this offensive, Google has announced that companies can now connect their Gemini language models with their search engine , which enriches the quality of the responses generated, which are not restricted only to the information with which the models were trained. . “In this way, AI hallucinations are significantly reduced,” Kurian said.

Additionally, they can use a wizard to create generic AI agents connected to their organization’s data and Google Search. These capabilities are available in its Vertex AI enterprise platform, which allows companies to fine-tune large models with their enterprise data, and connect them to different corporate systems, applications and databases.

The company also announced the public availability of the preview version of Gemini Pro 1.5, a new generation of its mid-size natural language model that matches the quality of the most advanced (Ultra 1.0), but requires less computing power. Additionally, it improves in terms of understanding the context. With this model, companies can, for example, create chatbots that remember conversations and manage complex tasks; analyze hours of video content by locating specific details, or create assistants capable of reasoning through long documents, comparing details in contracts or summarizing research.

AI models

Google Cloud is not limited exclusively to the Gemini. The company offers in Vertex AI access to more than 130 AI models , including Claude 3 from Anthropic, Llama 2 from Meta and Mistral. There are also other proposals from the company’s AI laboratories, such as the Gemma family of open models, which has now been expanded with a version (CodeGemma), designed to help programmers generate code; as well as Image 2 , Google’s image generation AI, which now also allows you to convert text into animated images and include an invisible digital watermark.

Infrastructure

The company also announced the availability of its TPU v5p accelerator chip , in its strategy to advance its own alternative chips to Nvidia, which dominates this market. It has also unveiled Axion, its first general-purpose processor for data centers based on Arm designs instead of x86, which deepens its strategy of creating its own processors, and will be available at the end of the year.

In parallel, it has announced cutting-edge infrastructure based on Nvidia chips, such as the A3 Mega, and has highlighted that it will give its customers access to the expected new Nvidia AI superchips, known as Grace, in early 2025. Blackwell.

Productivity

In the battle it is waging with Microsoft in the field of business productivity tools, the company has also announced AI news in Workspace. The tech giant has introduced Vids , a new app (launching in June for businesses) for collaborative video creation.

In addition, it delves into its strategy of monetizing AI applied to productivity with two new commercial offers, both for $10 per month per user . The first adds AI capabilities to video conferencing and messaging apps, with smart meeting note-taking and instant translation features. Meanwhile, the second offering uses advances in AI to automatically protect information in Google Drive.

SOURCE