Search engines with built-in artificial intelligence models seem to be the next great internet revolution. Google has already taken the first steps with its chatbot Bard –although it has been expensive-, while Microsoft faces unprecedented success in Bing thanks to its integration with ChatGPT. However, while all may seem like gains, a recent report from Reuters confirms that these artificial intelligences bring more costs than benefitsat least today.
It has been John Hennessy himself, president of Alphabet -Google’s parent company-, who has answered the questions of the aforementioned medium. Here, he claims that “an exchange with the AIs known as great linguistic models costs probably 10 times more than a standard keyword search«. This could translate into “several billion additional dollars” for Google.
How many billions are we talking about, precisely? According to a Morgan Stanley analysis for Reuters, the additional cost generated by a chatbot like ChatGPT could be up to 6 billion dollars. About 10% of Google’s annual revenue of $60 billion. Meanwhile, reports from companies such as SemiAnalisis place this number at 3,000 million dollars.
Percentage of Google queries handled by AI | Estimated cost to Alphabet based on the number of words generated per query | |||
25 words | 50 words | 75 words | 100 words | |
10% | 600 million dollars | $1.2 billion | $1.8 billion | $2.4 billion |
twenty% | $1.2 billion | $2.4 billion | $3.6 billion | $4.8 billion |
30% | $1.8 billion | $3.6 billion | $5.4 billion | $7.2 billion |
40% | $2.4 billion | $4.8 billion | $7.2 billion | $9.6 billion |
fifty% | 3 billion dollars | 6 billion dollars | 9 billion dollars | 12,000 million dollars |
This is the reason for the large costs generated by ChatGPT in Google
The reason is simple to understand. Regular Google searches work by scanning a huge, pre-built index. These entries are subsequently scanned, classified and categorized according to their relevance and positioning. This all happens in a split second, and even Google itself lets you see how much it’s taken right under its search bar.
Meanwhile, searches carried out in ChatGPT or similar models have a much more complicated background. Each of them involves triggering the daunting neural network model, attempting to mimic a human brain. This happens every time you search, and their answers generate a lot of text that then has to be categorized to offer the most objective information. In addition, its chatbot nature means that people will want to interact with it for much longer, contrary to what is seen in traditional searches.
How does Google plan to fix these extra costs? At the moment, it seems that the company intends to release a lightweight version of its Bard chatbot to the market, something we saw in ChatGPT recently. Being much smaller in scale, it also requires less computing power, allowing more users to use it.
Microsoft has a bit of an advantage in this section
For their part, those from Redmond are in a much more privileged position, and it all has to do with the low popularity of your search engine. While Google has a 93% market share in search, Microsoft Bing barely reaches 3%. This translates into fewer people using ChatGPT at the same time, resulting in an incremental cost much lower than that of Google with its 8.5 billion daily searches.
Another problem that both companies face is making their integration with ChatGPT profitable. According to John Hennessy, Google is optimistic, and believes that this problem will only last a couple of years at the most. After all, they already did it once with YouTube, and created a real money machine.