Last updated on April 8th, 2024 at 12:05 pm

The expense of AI services might push industry leaders towards subscription-based models

Google is reportedly considering introducing fees for AI-enhanced search features, marking a significant change to its revenue model. Experts suggest this shift is driven by the substantial costs associated with providing such services, prompting other major players in the industry to also explore subscription-based models.

According to the Financial Times, Google’s plans involve offering the new search feature exclusively to users of its premium subscription services. These services already require a subscription for access to artificial intelligence assistants in other Google tools like Gmail and its office suite.

Google is testing a new search experience in beta for select users, where its generative AI provides responses to queries in a conversational style similar to ChatGPT and other competitors. According to Heather Dawe, chief data scientist at UST, AI search is more computationally expensive than Google’s traditional search processes. By charging for AI search, Google aims to offset these costs.

The focus in AI often revolves around the substantial expense of computing power required to train advanced generative models. For instance, Amazon reportedly spent $65 million on a single training run last year, with predictions that the company will soon exceed the $1 billion mark in training costs.

Last week, OpenAI and Microsoft revealed plans for a $100 billion datacenter dedicated to AI training. Additionally, Mark Zuckerberg announced in January his intention to invest at least $9 billion in Nvidia GPUs alone for AI purposes.

However, according to analyst Brent Thill from Jefferies, the cost of training AI represents only a fraction of the total expenditure in the sector. Thill stated in a briefing note that “The majority of AI compute spend today is directed to the running, not training, of models, and 90%+ of AI compute spend today is being directed towards inferencing [the process by which an AI model is queried], as inferencing spend has been growing much faster than training as more models and tools get put into production.”

He further stated: “Some companies have opted to price new Gen AI features on a monthly basis, expecting that higher charges will offset usage expenses, while others have chosen a per-usage pricing model to mitigate cost risks. Some have also integrated these features into existing plans in hopes of driving user growth.”

Competitors in the AI search space are adopting similar subscription strategies. For instance, Perplexity, an AI-powered search engine, runs without advertisements but offers a $20 monthly “pro” tier that grants access to more advanced AI models and unlimited usage.

On the other hand, some companies continue to offer their products at a loss. Microsoft’s Bing provides AI features for free but requires the use of the company’s Edge browser. Arc, a browsing and search startup, currently offers its products for free to users and plans to generate revenue in the future by charging companies for business-oriented features.