Microsoft’s early AI lead has become a test of faith

Unlock the Editor’s Digest for free

Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.

Remember when Microsoft was riding the crest of the AI wave? 

All of two years ago, its close partnership with OpenAI was thought to make it the best-positioned of all the Big Tech companies on AI. Today, things could hardly look more different, and not just because its ties with OpenAI have been loosened.

The software company’s shares have fallen 18 per cent over the past two years, in the midst of an AI boom that has lifted the wider stock market by a third. At only 22, its price/earnings ratio is at its lowest point for years.

There’s a touch of the AI revolution devouring its own children about the predicament Microsoft finds itself in. Providing the computing platform for ChatGPT propelled it to the forefront of generative AI. But as AI disrupts the software business, there are concerns about how much growth Microsoft will see from the next generation of AI-powered applications and how much those new services will cost to deliver.

The only inescapable certainty amid this upheaval: Microsoft’s capital spending is going through the roof, soaring from 12 per cent of revenue in 2022 to about 50 per cent this year.

On top of the general angst, there have been more specific concerns about execution. Copilot, Microsoft’s AI assistant that was meant to make workers more productive, has been slow to catch on. And when the company turned from its reliance on OpenAI to building more of its own models, investors worried about the costs, as well as whether the models would be competitive.

At the same time, the frontier AI labs have made inroads into the company’s markets. Microsoft’s GitHub had an early lead in applying AI to software coding, but coding agents from Anthropic and OpenAI have since stolen the limelight. Anthropic’s Cowork, the first of a new generation of AI agents designed to do routine office tasks, is the kind of service that might have been expected from Microsoft itself.

At the heart of this is the fact that AI enables new types of software, often supported by new business models, and is opening the market to competition from new directions. 

Microsoft’s CEO Satya Nadella spelled out the implications in an essay last month. AI-powered applications had little in common with traditional software, he said. Instead of being static tools for workers, they have the potential to evolve and become more useful as they absorb what Nadella called the “workflows, domain knowledge, and accumulated judgment” of workers inside a business. Getting the most out of the software will mean training internal AI models and then creating a “loop” of learning to constantly improve the performance of human workers.

In effect, this amounts to a new layer of software, coordinating the output of both models and workers to improve results. Not surprisingly, other companies, including the AI labs, are bidding for this ground.

The need to adapt processes to this new type of software will make it far more complicated to sell and implement. Last week, Microsoft said it had set up an internal business division for the purpose, with the aim of putting 6,000 engineers into its customers’ premises to help.

It is far from alone in recognising this need. Days before, Amazon Web Services announced plans to create an army of what the tech world calls “forward deployed engineers”. OpenAI and Anthropic have forged alliances with private equity firms to push AI deeper into the business world. 

Nadella, meanwhile, had a warning for Microsoft’s customers. If companies don’t learn how to train their own AI models to enhance internal operations, the general-purpose models from the AI labs will take on the job. The result, he said, would be “a world where every company across every sector is ceding value to a few models that eat everything they see”.

There is obvious self-interest in this, with Nadella positioning Microsoft as the antidote to the problem he diagnoses. But his words carry a warning for the software company itself. If the general-purpose AI models win, its business will be hollowed out.

Microsoft has many of the technology components in place to deliver on Nadella’s vision. And when it announces earnings this month, Wall Street will be hoping for further signs of the pick-up in growth it registered in the previous quarter. But it will take some heavy lifting before the full returns from its massive AI spending become apparent.

[email protected]