Investing.com - A standalone business bundling together Google's artificial intelligence accelerator chips and AI research lab could be worth over $700 billion, according to analysts at DA Davidson.
Alphabet-owned Google has been working on its so-called tensor processing units (TPUs), which can be utilized in training AI models and have a range of cloud computing infrastructure around them, for about ten years. The company has said one flavor its fifth generation of the chip rivals AI-darling Nvidia (NASDAQ:NVDA)'s popular H100 semiconductors.
DeepMind, meanwhile, is one of the key drivers of Google's push to develop its AI capabilities and keep pace with its big-name Silicon Valley peers in the arms race around the nascent technology.
In a note to clients, the analysts at DA Davidson led by Gil Luria flagged that, based on the current valuation of Nvidia and the top AI model providers, a business combining the two would have the potential value of as much as $60 a share.
Indeed, they noted that their analysis indicates Google would have generated $24 billion of revenue in 2024 had it been selling TPUs as hardware to Nvidia customers. As such, Google may have a "missed [an] opportunity", the analysts argued, adding that they are concerned by this because it "may point to a lack of ability to commercialize innovation" at the search engine titan.
They pointed to the history of the personal computer industry as an example of the risk they worry Google may be facing.
"We note that while Xerox (NASDAQ:XRX) invented the foundations of modern personal computing, and IBM (NYSE:IBM) has been issued the most patents around computing for decades, it is [Apple] and [Microsoft] that have derived almost all of the shareholder value from personal computing," the analysts wrote.
They gave Alphabet (NASDAQ:GOOGL) a "neutral" rating, but raised their share price target to $200 from $190 based on the "slight chance the company actually does choose to release shareholder value".
Alphabet is due to report its latest quarterly earnings on February 4.
(Reuters contributed reporting.)