Tech Giants See Significant Increase in Cloud Services Revenue This Week Due to AI Offerings. CIOs Say This Is Largely Due to Exploratory Efforts. Artificial Intelligence Drives Growth in Cloud Revenue for Major Tech Companies, Although Much of the Spending Still Reflects Preliminary Efforts by Customers.
"It's Still Early," Brian Kirkland, Chief Information Officer of Choice Hotels, commented on the company's ongoing work in the area of generative AI. "Most use cases are just exploratory." Both cloud companies and customers are questioning when the technology will gain more ground.
Since its public debut over a year ago, generative AI has not yet found broad acceptance in companies beyond a few specific use cases, as businesses familiarize themselves with the technology and other factors, including legal and data protection risks.
A recent CIO survey by Morgan Stanley revealed that most CIOs do not expect their first generative AI project to go into production until the second half of 2024 or later. Nevertheless, AI testing and proof-of-concepts have translated into more cloud usage.
Amazon.com, which announced its quarterly results on Thursday, reported that its new generative AI capabilities have been well-received by customers and are reflected in its financial results. The company reported that the revenue share of the Amazon Web Services segment has increased by 13 percent compared to the previous year.
On Tuesday, Google's parent company Alphabet also reported a 26 percent growth in its cloud division, with AI contributing. Also on Tuesday, Microsoft announced that its Azure cloud business had grown by 30 percent in the fourth quarter of 2023, with a 6 percentage point increase attributed to the demand for AI.
Microsoft attributed the majority of AI expenditures on Azure to the use of the latest OpenAI models, the use of third-party AI models such as Meta, Hugging Face, and Nvidia, as well as the utilization of its code generation tool GitHub Copilot, all running on its cloud infrastructure.
Google Cloud announced that companies like Anthropic can train and execute their own models on its infrastructure and that business users can utilize its basic model, Gemini. Google Cloud stated that the number of generative AI projects on its application platform Vertex AI has sextupled from the second to the third annual sample.
Both Google and Microsoft Indicated They Have Increased Their Investment Spending to Build Their AI Infrastructure. "We Have Moved from Talking About AI to Applying AI on a Large Scale," Microsoft Chief Executive Satya Nadella Said on Tuesday. For Many Cloud Customers, the Ability to Apply AI on a Large Scale Remains Somewhere in the Future.
The increase in AI usage in the cloud is largely due to corporate customers testing its applications, said Stefan Slowinski, global head of software research at investment bank BNP Paribas Exane. "They have not yet been deployed."
There is no or only a low risk since you haven't put it into people's hands yet. You are just testing." He added, "They are still very much in the trial phase, but of course that costs money, and that's why we see some increases in the infrastructure to accomplish that."
Generative AI Tests and Proof of Concept Range from Customer Service Numbers to Equipment Repair. Kirkland from Choice Hotels mentioned that the company is currently testing an AI model on Amazon Web Services that would generate summaries of interactions with customers on the Choice website for call centers if customers call the service hotline. Kirkland said it was too early to assess the results of the test, but this use case is just one example of what is possible with generative AI.
He also added that he was not in a rush to provide anything. Ultimately, he said, "We will get there." At CNH Industrial, a supplier of agricultural machinery and construction equipment, Chief Digital and Information Officer Marc Kermisch stated that the company is using OpenAI's GPT model and Azure infrastructure to test several applications.
One of them, which essentially serves as a search engine for equipment repair, is to be deployed this year, he said. But other proofs of concept have not yielded results, he added. A model that was used to query old sales data and make predictions about future demand has hallucinated when it did not find enough data, said Kermisch.
He added that the reason why proofs of concept often fail is that they cost more to go into production than they actually save in efficiency. The risk for hyperscalers is that only a few use cases survive the pilot phase, Slowinski said, either because they do not offer a clear return on investment or because companies are unable to develop sufficiently clear security controls.
John-David Lovelock, Research Vice President and Distinguished Analyst at the IT consulting and market research company Gartner, said that executives are suffering from "POC fatigue" or persistent frustrations about ongoing concepts that don't work. Certainly, some companies have positioned themselves as early adopters.
Online Furniture Retailer Wayfair Launched a Tool Last Year that Uses the Stable Diffusion AI Image Generation Model to Help Customers Reimagine Their Living Spaces. Not Included in Microsoft's Cloud Expenses was Microsoft 365 Copilot, Powered by the Same Technology Behind ChatGPT.
The company announced that the revenue for it would flow into the reporting of Office365 Commercial and would not be part of Azure. Microsoft declined to disclose revenue or data figures for the expensive tool that was launched at the end of last year.
Chief Financial Officer Amy Hood said in Microsoft's earnings conference: "Although it is still early days for Microsoft 365 Copilot, we are thrilled with the usage so far and continue to expect that revenue will increase over time."
