By chance on a Sunday morning, Microsoft CEO Satya Nadella penned a lengthy X post on the “Reverse Information Paradox” of artificial intelligence adoption and utilization within companies: the idea that although companies pay to use AI, they also risk revealing proprietary information.
Nadella argues that to best respond to this paradox, organizations should not prioritize the AI models themselves, but the “harness” around the model — the infrastructure that allows an AI model to operate as an agent and execute more complex, long-running tasks. However, this argument presents a false choice for organizations and customers: Either become dependent on a particular AI company or become dependent on the infrastructure that surrounds the AI model (and in this case, that infrastructure is Microsoft).
Recommended Stories
This false binary completely overlooks a third option that combines the best of both worlds and doesn’t require any lock-in. Instead of choosing between a model and a system, why not both? Why not utilize cutting-edge, innovative AI models that run on secure, interoperable cloud technologies?
Nadella’s position is a fascinating, if predictable, move when you consider Microsoft’s evolution on AI, especially the ongoing challenges with its AI offerings. When his company has embedded Copilot into virtually every product it offers, from Microsoft 365 to Edge, it makes sense to argue that it’s what surrounds the model that’s the most important factor to consider (as compared to a competitive, cutting-edge model itself) — and do so just before Microsoft was scheduled to report its Q4 earnings.
Microsoft has faced serious questions from investors and analysts about its AI strategy during the past two quarters. Following its most recent earnings results, some analysts expressed continued concern about high spending linked to AI investment, with a financial institution noting that free cash flow came in “meaningfully below” their estimate. To address these concerns, Microsoft must demonstrate ongoing and significant returns on its AI investment and strategy. As Microsoft attempts to achieve these returns, it’s interesting to see Nadella argue for the importance of the AI “harness” Microsoft offers over the models themselves.
Microsoft’s AI approach effectively creates a layer that leverages the company’s existing monopoly to push an AI tool alongside its more widely used products. Unfortunately for Microsoft, bundling Copilot with existing tools doesn’t automatically mean users will embrace it. In fact, Microsoft customers generally haven’t made the jump from casual Copilot user to a paying customer: A scant 1% of Microsoft 365’s millions of users actively utilize Copilot weekly. Only around 6% of these users are willing to pay for Copilot, and some organizations are only using 10% of the seats they paid for.
Other companies have proven that a different, less restrictive approach to AI is possible by launching their own interoperability efforts.
For example, Cisco partnered with a group of other companies to develop an industry-standard agent interoperability language, with the goal of allowing different AI agents to communicate quickly and easily. Anthropic has a similar initiative, the Model Context Protocol, which creates a universal, open standard to allow AI agents to connect to external ecosystems. Efforts like these should be supported and held up as examples of what a robust, competitive AI future could look like.
In contrast, Nadella’s “harness” argument, when viewed through the lens of Microsoft’s bundling behavior and ongoing adoption challenges, seems even more disingenuous. This situation is why scrutiny of anticompetitive behavior is so important — companies like Microsoft will try to argue that their lock-in activities benefit users, even though the benefit is actually entirely for the company itself.
For anyone who remembers the early days of the internet, watching Microsoft bundle AI to box out competitors around an emerging technology should feel very familiar. This is the very same process by which Microsoft bundled Internet Explorer to crush Netscape during the “First Browser War” of the ’90s. This time, however, regulators are showing signs that they are onto Microsoft’s game and acting faster to prevent another casualty like Netscape. The Federal Trade Commission is already investigating Microsoft’s AI and cloud business practices, which could mean an antitrust lawsuit is on the horizon. Considering Microsoft’s long history of anticompetitive behavior, this investigation is warranted.
COMEDY WITH A WARNING LABEL: NEW MEXICO WANTS THE BABYLON BEE TO SPOIL ITS OWN JOKES
Microsoft is just one part of a much larger tech ecosystem, where customer choice and healthy competition should be the ultimate goal. Open, interoperable technology is only possible when companies with a history of anticompetitive behavior are sufficiently scrutinized, and action is taken to course-correct.
Nadella gives readers a false choice between two less-than-ideal lock-in scenarios, and he may be attempting to shift the narrative around Microsoft and AI. However, customers and, more importantly, regulators, have been here before, and they don’t want a repeat of the licensing lock-in of the 1990s. It’s a false choice: The choice is between restrictive, anticompetitive practices vs. innovation and interoperability.
David Linthicum is a globally recognized thought leader, innovator, and influencer in artificial intelligence, cloud computing, and cybersecurity. Until early 2024, he served as Managing Director and Chief Cloud Strategy Officer at Deloitte. Currently, David continues to provide key thought leadership, architecture, and technology leadership, serving government agencies and technology companies, including those that compete with Microsoft.
