How Microsoft Plans to Squeeze Cash Out of AI
Microsoft has invested $13 billion in OpenAI, providing the money-losing start-up with the huge amounts of capital and computing power necessary for its continued operations. In exchange, Microsoft gets access to OpenAI’s technology for use in its own products as well as a real and reputational stake in the AI boom. “We are below them, above them, around them,” Satya Nadella said in March about OpenAI.
Microsoft’s investment has been cast as a strategic masterstroke, through which stodgy old Microsoft sent Google into a panic and established itself as a player in tech’s Next Big Thing. Good for it! And the company’s stock price. But Microsoft’s long-term plans to make money with OpenAI’s technology — and to monetize AI in general — are still taking shape. There was a Bing relaunch, which promised clean, simple responses to questions, which Nadella suggested was existentially threatening to Google’s core business; one year later, it was still error-prone, its interface was cluttered, and its market share was still hovering in the low single digits. Last year, the company started charging for subscriptions to Copilot, its AI assistant software, in hopes that enterprise customers would shell out for the promise of increased employee productivity in Microsoft Office and Github; according to the company, uptake has been solid.
AI subscriptions are, so far, the tech industry’s favorite idea for making money from AI. This is conceptually simple — your customers are paying you for access to a new product. The problem is that compute-heavy cloud services like ChatGPT and Copilot remain extremely expensive to run, meaning that in some cases even paying customers might be costing them money. Computing costs are likely to fall, and AI-model efficiency could improve, but, much like the basic assumption that there’s a huge market for these things just waiting to be tapped, these are bets and not particularly safe ones.
This week, Microsoft announced that it would be integrating AI more deeply into even more of its products, including Windows, which, among many other chatbot-shaped things, is set to get a feature called Recall, described by the company as “an explorable timeline of your PC’s past.” This feature, which will be turned on by default for Windows users, records and “recalls” everything you do on your computer by taking near-constant screenshots, processing them with AI, and making them available for future browsing through a conversational interface. You can watch Nadella try to explain the feature and its appeal to the Wall Street Journal’s Joanna Stern:
Microsoft’s AI Recall explained:
— Joanna Stern (@JoannaStern) May 21, 2024
➡️ Takes screenshots of everything you do in Windows 11 and uses local small language model to help you search through it
➡️ Stores screenshots locally. You adjust how long they are stored for and how much drive space to dedicate
➡️ Runs… pic.twitter.com/5KWIbzWoPH
Like smartphones, personal computers already collect and produce vast amounts of data about their users, but this is a big step in the direction of surveillance — constant, open-ended, and mostly unredacted — offered in exchange for a strange feature that Microsoft’s CEO is quite insistent its users will enjoy. Nadella attempts to preempt any concerns by pointing out that the AI models powering Recall run locally — that is, on the user’s device, not in the cloud. This is, at best, a partial solution to a problem of Microsoft’s own creation — a problem Windows users didn’t know they had until this week.
On-device AI processing is interesting to Microsoft for other reasons, too. In a world where AI services are expensive to run, installing them in every popular Microsoft product represents a real risk. In a world where the processing necessary to run chatbots, generate images, or surveil your own computer usage to the maximum possible extent occurs on users’ devices, the cost of deploying AI is vastly lower.
For Microsoft, that is — if it expects to fully utilize these new features that are becoming increasingly integral to the core Window product, customers will have to buy new machines, some of which Microsoft also showed off this week. According to The Verge:
“All of Microsoft’s major laptop partners will offer Copilot Plus PCs, Microsoft CEO Satya Nadella said at an event at the company’s headquarters on Monday. That includes Dell, Lenovo, Samsung, HP, Acer, and Asus; Microsoft is also introducing two of its own as part of the Surface line. And while Microsoft is also making a big push to bring Arm chips to Windows laptops today, Nadella said that laptops with Intel and AMD chips will offer these AI features, too.”
These PCs will come with a “neural processor,” roughly akin to a graphics card, which is a separate hardware feature that can handle AI-related processing tasks more quickly and with lower power use than existing CPUs and GPUs. In conjunction with Microsoft’s shift to more efficient mobile processor architecture for laptops and desktops — something Apple committed to years ago, selling huge numbers of laptops in the process — AI is being used to make the case to its customers that this is the next stage of the upgrade cycle. It’s time to get a new PC, says the company that makes the software that powers most PCs and that sells PCs of its own.
Microsoft, like many other tech giants, says it’s all in on AI, but its approach includes hedges against AI deflation, too. Maybe customers flock to new AI features, in which case Microsoft will have shifted computing expenses back to its billions of customers, improving margins on subscription products and selling lots of Windows licenses in the process. If they don’t, though — if people keep using their Windows machines in approximately the same way they have for decades — Microsoft makes money anyway and leaves its cloud computing capacity free to sell to other firms that want to try their luck building AI tools. Multi-market sector domination with considerable leverage over different but overlapping groups of customers: Not such a bad deal!