By almost any measure, Google is the master of applied artificial intelligence. Research papers and conferences are replete with the company's technical achievements in the field of deep learning, today's dominant form of AI.
And AI is also finding its way into operations and products at Google, by helping to reduce data center costs for the company's cloud computing operations, on the one hand, and by developing new categories of offerings, such as its Waymo self-driving automobile unit.
There must be a value for all of this, but what exactly is it? It doesn't appear to show up in the valuation of Google parent company, Alphabet (GOOGL) . The stock valuation, at a recent price of $1,214.65, is an extremely modest multiple of 5.2 times projected revenue of $162 billion this year, based on a market cap of $839 billion. At five times revenue, the stock is not reflecting the value of its tangible assets, which totaled about $121 billion in cash and cash equivalents as of last quarter; if those assets, net of long-term debt, is taken into account, the multiple is actually even cheaper, at 4.5 times.
But AI as a corporate asset, would most likely be an intangible asset from a balance sheet perspective. That's the view taken by noted economist Erik Brynjolfsson and his colleagues at MIT's Sloan School of Management in recent research.
The way they look at things, AI is a form of intangible capital that comes on top of the investment in computers. And it's hard to measure, because AI, in the form of algorithms for language translation -- think, Google Translate -- isn't a single thing, it's a whole set of capabilities and technologies. As Brynjolfsson and colleagues write, "AI requires developing datasets, building firm-specific human capital and implementing new business processes."
Those things can all be of much greater value, ultimately, than simple computer systems. Take the example of Google's Cloud Computing service. That service provides AI capabilities such as natural language processing and image recognition as a service to Google's cloud customers. That generates cloud revenue for Google, just as it does for Amazon (AMZN) and Microsoft (MSFT) . In theory, one could work backward from the value of cloud to arrive at some estimate for what AI contributes. But Google doesn't disclose how much revenue it makes off of its cloud business, so it's impossible at present to calculate how much AI may be contributing.
On the balance sheet, at least, the closest comparable for AI would be the intangible assets of Alphabet, which totaled $1.9 billion as of last quarter. That figure mostly reflects acquisitions Google has done, and so it is unlikely to reflect the value that organic development of AI is adding to the company's assets.
Perhaps the best immediate proxy is to look at a productivity figure, such as revenue per employee, since what AI does is make work more efficient. On that basis, Google's far more efficient. The company had 107,00 employees at last count. Dividing that into its projected $162 billion in revenue equals $1.5 million in revenue per employee annually. Amazon has a comparable number of approximately $427,000 this year per employee, based on Wall Street projections for revenue. But the figures are muddied by the fact that Amazon reports an employment figure, 653,000, that also includes part-time contract workers.
In any event, it's quite likely Google is more efficient on a per-capita basis than Amazon -- and a lot of other companies, for that matter -- but it's also not entirely certain to what extent that kind of number reflects its operations being more efficient as a result of AI. The mystery will continue until Google itself, or someone else, can quantify what the outcome of its prolific AI outputs are in dollars and cents.
It's an important valuation question because over time, the intangibles of AI can have a big impact on the efficiency and therefore the worth of a company. As Brynjolfsson and his colleagues have written, looking back to the IT revolution of the '80s and '90s that ushered in PCs and networking, "the intangible assets associated with the last wave of computerization were about ten times as large as the direct investments in computer hardware itself."
"We think it is plausible that AI-associated intangibles could be of a comparable or greater magnitude," they conclude.
Maybe the simpler way to put it for investors is that Alphabet stock is really, really cheap right now, by most conventional measures, and those measures probably don't even reflect what may be some of its greatest long-term value, the datasets and the algorithms that make up AI.