Microsoft and Alphabet released their quarterly earnings reports at the same time. What do they say about AI progress?
Google Search “and other” revenue rose from $39.6 billion to $40.4 billion last quarter, representing growth of about 1.87 percent year over year. This compares to 24.28 percent growth in the same quarter in 2022 and 30.11 percent growth in Q1 2021.
However, as with the entire digital industry, there were tailwinds from the Corona crisis that are now partially reversing. Google’s search business actually declined slightly in Q4 (-1.6%) and is now growing again. The numbers do not indicate that Google Search’s revenue growth is exhausted or that ChatGPT with Bing is an acute threat.
Overall, Google was able to beat analyst expectations and the stock reacted slightly positively after the earnings call. Google seems to be weathering the current market period well. Nevertheless, Google CEO Sundar Pichai is looking for the next growth horizon – in addition to the cloud business, that could be AI.
AI transformation on by with the mobile revolution
Speaking to journalists and investors, Google CEO Sundar Pichai described AI as an opportunity comparable to the shift from desktop PCs to mobile devices more than a decade ago. Google has been investing in AI for ten years and is in an excellent position, Pichai said.
He pointed to the launch of Bard with the latest code generation updates, the PaLM API, the integration of generative AI in Workspace, AI-specific training offerings in Google Cloud, and the merger of Google AI and Deepmind to form Google Deepmind.
Bard is an add-on to Google Search, Pichai said, but the capabilities of large language models will gradually be integrated natively into search. AI could be more helpful in a “category of queries, maybe in which there was no right answer, and they are more creative,” Pichai said. But even in existing search categories, LLMs could “do some heavy lifting for the users” and “guide them.” Pichai added that it’s still early days and there’s a lot of innovation to come.
Pichai sidestepped a question about the higher cost of search with language models: Google has always paid attention to the costs of compute, he said, and is very experienced at becoming more efficient in hardware, software, and models. “This is not new,” Pichai said.
Bing app grows four times faster
Microsoft also reported its quarterly earnings, beating analyst expectations for revenue growth and profit. Search and news advertising revenue, still a small line item for Microsoft, grew 10 percent and 13 percent, respectively, with two percentage points coming from the acquisition of the Xandr advertising platform.
This is the lowest growth in this segment in four quarters. In Q3 ’22, search and news grew 23% and 25%, respectively. Bing’s user growth from its ChatGPT-like chatbot does not yet appear to be driving significant incremental search revenue or offsetting the post-Corona decline from weak advertising.
According to Nadella, Bing now has more than 100 million daily active users, a figure Microsoft already announced in early March, so it doesn’t seem like there’s been much additional growth since then. Since the launch of Bing and Edge, daily installs of the Bing app have quadrupled. A number that is almost meaningless without more context.
According to Microsoft, both Edge and Bing have gained market share in the US, but again, the company does not provide specific numbers.
DALL-E 2, which has been integrated into the Bing chatbot for about a month, has generated more than 200 million images so far, according to Microsoft. That’s the equivalent of about 50 million queries (one query results in four images).
OpenAI reported last November that in DALL-E’s native web interface, three million users generate four million images per day, which translates to about 120 million images or 30 million queries per month. So DALL-E’s usage through the Bing chatbot may be slightly higher than OpenAI’s native usage, but it’s not a huge difference.