Stephen Colbert kicked off the interview like so: "You are the author of a book called Free, and it costs $26.99. I love a bait and switch. Well done, my man!" Then Colbert turned to the camera, pointed at us all, and said "Suckers!"

For me, this launched the golden era of the free internet business model, with all the incredible benefits and riches it created, along with all the contradictions and problems.

Google Search was free as long as you occasionally clicked on ads and let the company scoop up mountains of data about you for ad targeting. Facebook was exploding as a free way to connect with friends in 2009, and the social media ad avalanche was just beginning. YouTube was emerging as an online video giant on the back of endless clips uploaded in a content free-for-all.

Our paid AI future

Today, we're at the start of a similarly exciting new technology wave with generative AI. This time, though, many products and services won't be free.

Amazon's Alexa voice-powered service is a great example. This technology was always provided free, but now the company is working on a revamped version that is powered by a new large language model. This offering, known as "Remarkable Alexa," will be a paid subscription, according to Business Insider's Eugene Kim.

Even Google, the master of free online services, is considering paid subscriptions for some of its new AI offerings. With AI, CEO Sundar Pichai sees an opportunity to evolve Search and the Google Assistant service over the next decade.

"Over time, there will be newer paths, just like we have done on YouTube," he told analysts late last year. "I think with the AI work, there are subscription models as a possible path as well."

Software code spotted recently suggests that the advanced version of Google's Bard chatbot will be a paid subscription, possibly through the Google One cloud storage offering, according to Engadget.

So, why will generative AI offerings be paid from the start? I asked Oren Etzioni, an AI expert and a veteran of the online Search business. He shared a few thoughts:

The broad services (Search, Facebook, WhatsApp) are still free/ad-supported  (if the product is free, then you're the product).Paid/subscription offerings have emerged (Spotify, YouTube Premium, etc.)GenAI is more expensive to run and potentially higher value

It's true that training and running large language models is incredibly expensive right now. Just the required GPUs cost many thousands of dollars each and you need thousands of these chips.

Indeed, Amazon's former hardware and devices boss, Dave Limp, has said the company would have to start charging for the more advanced version of Alexa, given the high cost of running AI models.

Ads might not work as well

But if cost is the issue, then why did Google do Search as a free ad-supported product years ago? That still required massive investment in data centers, networking and a host of other expensive ingredients.

One possible answer is that ads may not work as well in this new generative AI future. Take Google Search as an example. Traditional search results have lots of space on the page for links and ads that users click on. Google has made more revenue and profit in recent years by stuffing search result pages with more and more ads. If AI chatbots are providing straight answers to users in the future, that leaves less space for ads. In extreme scenarios, the best generative AI technology will give users one single answer directly, and there may be no ads to click on at all in that case.

So how will Google keep making enormous profit? Getting users to pay for its powerful new AI offerings via subscriptions is one way to keep the earnings machine running.

Pressure on Big Tech companies

More broadly, Etzioni thinks the most important reason for this paid AI future may be that tech companies are under pressure to demonstrate revenue growth (amid a recent advertising slowdown) and strong profit margins.

They have responded so far with substantial layoffs. "Seeking new revenue streams is a natural next step," he told me.

There are certainly concerns about a looming growth problem in the tech industry. The "Magnificent 7," which includes Google, Apple, Amazon, Microsoft and Meta, generated annual revenue growth of 15% from 2013 through 2019, and raised that to 18% in 2021 and 2022. Wall Street sees that falling to 11% a year from 2023 through 2025, according to Goldman Sachs research.

And the valuation of these big tech companies is off the charts, compared to the rest of the US market. The Magnificent 7 has a price-to-earnings multiple of 27 versus 17 for the other companies in the S&P 500, Goldman researchers noted recently.

Slowing growth and relatively high valuations mean big tech companies have to pull on other levers to keep investors happy. As Etzioni notes, that will likely be the profit lever.

Charging for new generative AI services is one way to create new earnings. As long as these products are good enough to persuade users to pay.