Google’s Bard AI could cost 10x as much as traditional keyword search

AI has taken the world by storm in recent months, and the next steps are still to come. For Google, though, further expansion of AI in Search by way of Bard could represent a huge increase in cost.

Bard was announced earlier this month as Google’s AI model for Search products, allowing users to obtain information in a style similar to that of ChatGPT and, now, Microsoft’s “new” Bing. But Bard is yet to actually make it to market, with testing underway behind the scenes and no public access available.

Reuters reports that Google could see drastically increased costs from using Bard for Search when the tool does go live.

Alphabet chairman John Hennessy said in an interview with the publication that a “large language model” like Bard could inflate the costs of Search by 10 times compared to a traditional keyword search. That expense would apparently go down “quickly” as the product is fine-tuned. Analysts still believe the tech could end up chipping into Google’s bottom line, though, even with advertisements attached.

An estimate by Morgan Stanley pins down a cost of roughly 1/5 of a cent for a Google Search query. That obviously adds up quite quickly with the 3.3 trillion searches performed on Google last year alone, but the number would skyrocket when using AI. It’s estimated that Google would see costs rise by $6 billion by 2024 if just half of searches are performed by the ChatGPT-like Bard are answered with 50 word responses. The chart below illustrates how costs may scale.

The added cost of AI-enhanced search boils down largely to computing costs, including the cost of chips, maintenance, and electricity. SemiAnalysis estimated that Bard could cost up to $3 billion in use, but noted that the amount was limited by the number of Tensor Processing Units Google has.

Notably, though, Reuters cites a source close to Google that says it is “early” to pin down the exact costs of generative AI in Search, as efficiency and usage will play key roles in finding the true costs.

Microsoft has already launched its ChatGPT-powered Bing in a public preview, and in response to a question of the costs, Microsoft’s CFO said that the influx of new users and increased advertising revenue “outweigh” the added costs of AI search.

While cost is surely a concern for Google with Bard, it seems the company’s bigger focus is on getting the technology right. John Hennessy, cited earlier, recently said that AI is still “one to two years” away from being a truly useful tool, and that Bard is still churning out wrong answers at times. But clearly, Google is pushing to bring the product to life under increased competition, an effort employees have called “rushed” as early errors contributed to a 9% stock dip for Alphabet.

More on Bard & AI:

• Google announces Bard, powered by LaMDA, to compete with ChatGPT
• Google working to bring Bard AI chat to ChromeOS
• Report: Google asking employees to help train Bard by rewriting wrong answers