Barclays highlights a "big shift" happening in artificial intelligence

Investing.com - Artificial intelligence is in the midst of a "big shift" that will see models move from being trained to becoming more sophisticated reasoning systems, according to analysts at Barclays.

Unlike older AI models, this already-trained form of the technology is capable of using logical rules to evaluate and analyze new information -- all without being given examples of the desired result.

AI "agents," for example, are able to harness their high-end computing power to pursue goals and complete tasks autonomously on behalf of human users. As a result, they have become crucial pillars of many companies’ AI agendas.

In a note to clients on Friday, the Barclays analysts said that Microsoft (NASDAQ:MSFT), which has pushed to be at the forefront among mega-cap tech names in the AI sector, appears to have started to make this transition towards so-called "inference" models.

"The shift to agents at the application layer and reasoning at the model layer has dramatic impacts to underlying technology and compute requirements going forward," the Barclays analysts led by Ross Sandler wrote.

In particular, growth in AI "pre-training," or the process of teaching a model on a large, general dataset before it is then honed on a specific task, would stall beginning next year, the Barclays analysts said.

Meanwhile, the bank argued that the cost of using inference computing power would end up being "far lower" than prior forecasts "as the underlying state-of-the-art models would end up smaller."

Although they do not expect AI researchers will stop training these cutting-edge models, the analysts said "the idea of spending $10 billion on a pre-training run on the next base model, to achieve very little incremental performance, would likely change."

3rd party Ad. Not an offer or recommendation by Investing.com. See disclosure here or remove ads.

The comments come as investors, spurred on by the rise of a low-cost AI model from China’s DeepSeek earlier this year, have raised concerns around the necessity of -- and eventual returns from -- massive capital expenditures by Big Tech players like Microsoft and Facebook-owner Meta Platforms (NASDAQ:META) on building out their AI operations.

Executives at these firms have, for the most part, backed their AI spending plans, arguing that it is crucial for the business to capture value from the epoch-making technology.

The Barclays analysts recommended that, should the evolution towards reasoning models continue and pre-training growth begin to ease, investors "go long" on hyperscalers because their free cash flows would start to rise.

"As we move from the models stage to the product adoption stage, and as inference compute unit costs decline, the application stocks should outperform" as well, the analysts said.

Is MSFT truely undervalued?

With MSFT making headlines, investors are asking: Is it truly valued fairly? InvestingPro's advanced AI algorithms have analyzed MSFT alongside thousands of other stocks to uncover hidden gems with massive upside. And guess what? MSFT wasn't at the top of the list.

Unlock ProPicks AI