Alphabet has a data quality advantage in AI – share price decline 'overdone' argues Jefferies

This post was originally published on this site

Alphabet’s (NASDAQ:GOOGL) stock hit is overdone, according to Jefferies analysts, who said the firm is positive on Google’s AI advantages.

Jefferies hosted an AI expert who emphasized to them that the quality of large-language AI models is highly dependent on the quality of data — an area the analysts state Google has a “huge lead given usage of Search, Chrome, and Android.”

The analysts reiterated a Buy rating and a $130 price target on Alphabet shares in the note.

One of the key takeaways from the meeting for Jefferies was that Google “excels in large language models (LLMs).”

“GOOGL wrote the seminal research paper on transformer concept that led to the latest LLM architectures and generative AI. GOOGL has its own LLMs, including BERT (based on 340M parameters), LaMDA (137B), and PaLM (540B), which compare to OpenAI’s GPT-3 (175B; ChatGPT is based on GPT-3.5),” the analysts explained.

“But good LLMs depend more on good data, where GOOGL has a huge moat,” he added. “Our expert emphasized that regardless of which LLM model is used, it needs to be trained on good data, where GOOGL has a huge lead given its massive volume of search requests, #1 Chrome web browser, and #1 Android mobile O/S.”

They also explained that Google excels in processing and ingesting this data. While in contrast, OpenAI was trained on a limited, publicly available data set.

As a result, Jefferies believes switching consumers will be difficult, and they expect “no rev shift in foreseeable future, with GOOGL unveiling substantial AI advances gradually.”

The analysts concluded that the 12% decline in the stock in the last two days is “overdone, and 9.8x EV/EBITDA near 7x 15-yr trough is attractive.”