Dive Brief:
- Google’s AI strategy requires cloud infrastructure optimization, according to the company’s CEO Sundar Pichai. “We have significant multiyear efforts underway to create savings, such as improving machine utilization and finding more scalable and efficient ways to train and serve machine learning models,” he said during last week’s Q1 2023 earnings call.
- Google Cloud turned a profit last quarter for the first time since the company began reporting separately on the unit, posting $191 million in operating income on $7.4 billion in revenues for the three-month period ending March 31.
- A data center build-out is also underway, according to CFO Ruth Porat. “We expect the pace of investment in both data center construction and servers to step up in the second quarter and continue to increase throughout the year,” Porat said during the earnings call.
Dive Insight:
Clearing space for the storage and compute needed for large language models — and the generative AI tools they power — is a priority for Google.
“We are making our data centers more efficient, redistributing workloads and equipment where servers aren’t being fully used,” Pichai said. “This is important work as we continue to significantly invest in infrastructure to drive our many AI opportunities.”
Google Cloud is the third-largest hyperscaler by market share, behind AWS and Microsoft. The cloud provider captured 10% of the worldwide market, according to Synergy Research Group, a one percentage point drop from its 11% share the prior quarter.
In addition to their battle for cloud dominance, the hyperscalers are now jousting for the top position in enterprise-grade generative AI.
After Microsoft’s early success with ChatGPT, Google and AWS responded with cloud-based generative AI tools. All three cloud providers are now embedding the technology in their products.
Pichai outlined three areas of opportunity for Google Cloud during the earnings call:
- Large language models
- Tools for developers, creators and partners
- Enterprise-grade AI tools across organization sizes
The run cost of generative AI, even once it's past the development stage, is a potential roadblock to broader adoption. Massive data centers housing racks upon racks of servers are needed to support the user-friendly interfaces of ChatGPT, Bard and similar LLM innovations.
“Cost of compute has always been a consideration for us,” Pichai said. “For us, it’s a nature of habit to constantly drive efficiencies in hardware, software and models across our fleet.”
Improving external procurement and managing its real estate portfolio are two other areas the company has targeted for savings, Pichai said.
The company has also reduced its workforce. Pichai announced the elimination of 12,000 positions in January and said the company will continue to shift its approach to hiring.
“We are meaningfully slowing the pace of hiring in 2023, while still investing in priority areas, particularly for top engineering and technical talent,” Pichai said last week.
