Google is selling broad access to its most powerful artificial-intelligence technology for the first time as it races to make up ground in the lucrative cloud-software market.
Google is selling broad access to its most powerful artificial-intelligence technology for the first time as it races to make up ground in the lucrative cloud-software market.
Thomas Kurian, chief executive officer of Google Cloud, outlined the offerings to thousands of customers at Google’s annual cloud conference in San Francisco on Tuesday, while making widely available a number of tools that can help draft emails or summarize lengthy documents stored in the cloud.
Hi! You're reading a premium article
Thomas Kurian, chief executive officer of Google Cloud, outlined the offerings to thousands of customers at Google’s annual cloud conference in San Francisco on Tuesday, while making widely available a number of tools that can help draft emails or summarize lengthy documents stored in the cloud.
The new products will intensify competition with Microsoft, the second-largest cloud provider, and OpenAI, the ChatGPT creator it has backed with billions of dollars. Both companies already sell access to the latest AI technology behind the popular ChatGPT bot, but Google’s launch on Tuesday puts it ahead of Microsoft in making AI-powered office software easily available for all customers.
Google also used the event to showcase customer uses of its technology, including a planned GE Appliances app that uses Google’s AI models to create recipes based on the food in a customer’s fridge.
The efforts are Google’s latest attempt to spark growth in the cloud business, an important part of CEO Sundar Pichai’s attempts to reduce dependence on its cash-cow search engine. Recent advances in AI, and the computing resources they require, have added extra urgency to turn the technology into profitable products.
Google’s infrastructure and software offerings produce $32 billion in annual sales, about 10% of total revenue at parent company Alphabet. Its cloud unit turned a quarterly operating profit for the first time this year.That still leaves Google firmly in third place in the cloud behind Amazon and Microsoft.
Kurian has likened generative AI to the invention of cloud computing decades ago, a development that allowed files to be stored on remote servers rather than local devices and accessed through the internet.
Google will make widely available its current large PaLM 2 model, which powers many of the company’s generative-AI features. It was previously only available for handpicked customers. The company also will make available AI technology developed by Facebook-owner Meta Platforms and the startup Anthropic, in which it is an investor.
The company also announced Tuesday the general availability of a suite of AI-powered tools for corporate Gmail accounts and other workplace software products. The tools will cost an extra $30 a month per person, matching the price Microsoft will charge for a competing offering in its flagship 365 suite, which is still in testing with select businesses.
Google’s ability to deliver on its cloud pitch will largely depend on whether it can secure enough chips to meet the demands of its growing customer base. The generative AI boom has placed greater pressure on executives who help divide computing resources between customers and Google’s own products, such as its namesake search engine.
Kurian said Google Cloud has enough resources to meet customer needs. Google has shifted the mix of chips in its data centers toward processors that can handle AI applications, he said.
Requests for AI-computing resources have nearly exceeded long-range expectations the cloud business set more than a year ago, Kurian added.
Many businesses are still early in experimenting with generative AI, and it is uncertain most will find productive uses for the tech. Users have also reported diminished performance recently in popular products such as ChatGPT.
“When the technology is really useful, people tend to stay with it," Kurian said in an interview before the event, citing data showing that workers using an AI writing tool in Gmail were sending roughly 30% more emails than normal.
Pichai said in an interview that Google was aggressively adding resources needed to run AI software, while adding a note of caution about the company’s ability to plan for the future.
“It’s tough to fully predict the demand," Pichai said. “I think we will be OK, but does it keep me up at night? Yes."
Pichai said Google will offer Gemini, its answer to the AI system behind OpenAI’s breakout ChatGPT, to cloud customers once the model is developed. Google co-founder Sergey Brin has been working almost full time at the company with researchers developing the model, The Wall Street Journal reported.
An early developer of generative AI and specialized chips that power the technology, Google has found itself playing catch up at times since the release of ChatGPT in November. Google Cloud, the company’s largest division by head count, has played a central role in its response.
Google Cloud oversees data centers holding the valuable chips that power large algorithms, pieces of hardware that have grown in importance as AI systems suck up greater quantities of data. Those include TPUs, a kind of processor Google began using in 2015 to develop its AI models.
Kurian also announced Tuesday, alongside Nvidia CEO Jensen Huang, that a large cluster of Nvidia’s newest AI chips, the H100, will become generally available to Google Cloud customers sampling a growing range of its AI offerings. Google’s TPUs also will become accessible for a wider range of tasks.
Character Technologies, whose chatbot app has attracted more than 20 million registered users, struck a deal in May to access Google Cloud TPUs and GPUs, the graphics chips used in AI systems. CEO Noam Shazeer said the company could still use more chips.
“The computing capabilities are improving a lot, but the applications are improving even more," said Shazeer, who pushed for Google to release a chatbot to the public before leaving the company in 2021. “There will be trillions of dollars worth of value and product chasing tens of billions of dollars worth of hardware."
0 Comments