Entering the hardware business is nothing unusual for an Internet giant like Google. But in launching its Pixel phone last week, Google GOOG is setting itself up to compete with its own direct customers. Unlike its previous Nexus smartphones, where most manufacturing and marketing is carried out by third parties—including Huawei, LG, and ASUS—the Pixel phone is completely Google’s own creation. From design to development to branding to marketing, Google is behind it all. But what’s peculiar is its determined pursuit in this less-profitable, asset-intensive side of business.

Several studies, business academics, and consultants have shown the limits of asset builders, or companies that accumulate physical assets in order to market and distribute physical products. Examples include Ford F , Marriott MAR , and FedEx FDX . Dividing the market values of these companies by their respective revenues shows that asset builders consistently fall behind network orchestrators, who maintain user networks in which participants interact (Uber, Airbnb, and Alibaba BABA , to name a few). Because network orchestrators do not need to invest in physical infrastructure as fast as they grow—Uber does not need to buy vehicles when it expands into a new city—their return on investment tends to be higher, with a faster growth rate.

This is how Android, even as a latecomer, surpassed Apple’s iOS in market share. By mobilizing hardware heavyweights like Samsung, countless app developers such as Snapchat, and content providers including Sony SNE Music, Google established Android as an alternative to Apple’s AAPL . Over the years, Google has also honed in on the most potent business model of the Internet age. From Search to Gmail and Google Maps, the Silicon Valley paragon always gives away products for free, rapidly expands its customer base, drives consumer usage, mines user data, and then sells advertisements.

So what could be a logical rationale for Google to suddenly mimic Samsung, Huawei, and HTC?

In 1995, Intel announced it was entering motherboard manufacturing. A motherboard is essentially an oversized electric circuit board laid inside a desktop PC, a sort of physical foundation of the machine. Ever since the ’80s, Intel has concentrated its resources in advancing microprocessors, leaving the Taiwanese motherboard manufacturers to dominate this low-margin sector of the business.

At the time, Intel had successfully brought forth Pentium processors. But to the dismay of the company, IBM IBM and Compaq—the world’s two largest PC makers—showed little interest in Intel’s latest innovation. The PC makers were busy selling computers using the old product architecture—the dated 80486 microprocessors alongside the existing motherboard for which the Taiwanese suppliers were not willing to let go. Without any new product in sight, consumers were not aware of the benefit of the new Pentium processor.

To break out from this circular situation, Intel went ahead with the goal of selling 10 million of its own motherboards, targeting smaller PC manufacturers who so far lacked the technological sophistication to work out their own system designs, but were eager to attain market advantages by adopting the latest technology. In short, Intel INTC was creating new competition in order to push the industry forward. To accomplish this goal, it had to enter a financially less attractive operation—motherboard manufacturing.

The gamble paid off. Other PC manufacturers, including HP HPE , Gateway, and Dell, quickly followed suit and sourced Pentium-compatible motherboards from Intel to leapfrog IBM and Compaq. They all wanted to bring out the next generation of PCs faster. To the relief of many Taiwanese suppliers, after a year with six million motherboards being made, Intel pulled back and stopped producing them. Why? The firm had since regained control over product pacing on microprocessors. Selling motherboards makes sense only when it helps Intel sell more microprocessors.

So what is it that Google wants to seek control over so desperately?

Anyone who has tried the Pixel phone will notice how similar it feels to other top-of-the-line smartphones. Certainly newsworthy is the supercharging speed that allows users to gain seven hours’ worth of battery life after a mere 15-minute charge. Still, Pixel lacks other features, like waterproofing and stereo speakers, touted by Apple’s iPhone 7. In short, Google has yet to differentiate itself function by function based on hardware design. The real differentiation lies elsewhere: the integration of artificial intelligence (AI) into the mobile device.

Baked into Pixel is Google Assistant, similar to Microsoft’s MSFT Cortana, Apple’s Siri, or Amazon’s AMZN Alexa. But because Google has acquired considerable knowledge about the world, having spent the last 18 years parsing data for search queries, the conversational assistant appears to be far more intelligent than all other contemporaries. When asked, “Play the Shakira song from Zootopia,” the assistant knew to play the correct track, whereas Amazon’s Alexa fumbles at this sort of cryptic request. Google Assistant is also capable of offering detailed advice on local businesses and directions with traffic predictions thanks to Google Maps.

 

From Google Docs to Gmail, Google Photos to Google Calendar and YouTube, the more people who use Google’s services, the more powerful the assistant will become. Your favorite holiday will display when you casually say, “Show me my pictures of a nice beach from last summer.” For end users, Pixel will, in short order, elevate the switching cost to other providers to new heights. It should not be a surprise that in a few more years, Google Assistant could evolve to remind you to buy something specific for your friend’s birthday, autonomously reserve a table at your favorite restaurant, and arrange a special drink before dinner. Google Assistant would then alert you when to head home and order an Uber on your behalf so that you get enough sleep for an important business meeting the next day.

To get to this point, Google needs to push the industry forward at a time when its traditional partners are busy with their own agenda and conflicting demands. Integrating AI into smartphones, as promising as it may seem, does not occupy the same priority at Samsung, for example, when the company is busy fixing its battery-explosion problem found in Galaxy Note 7.

To integrate AI into people’s daily lives, the input devices are irrelevant in the long run. Tablets, smartphones, or any consumer wearables are simply user interfaces. Thus, earlier this month, Google also went ahead to launch Google Home, the company’s answer to Amazon Echo. In building a Wi-Fi speaker, a sort of Siri-in-a-tube that answers users’ spoken questions, Google is now in the race of making devices that connect the homes of millions of Americans.

If history is a guide, however, all of these hardware furies will be short-lived. Like Intel, Google’s priority is only to advance the paradigm of consumer electronics to fully embrace AI at its very core. Once consumer demand is established, and the industry standard has emerged, Google is likely to retreat back to its prior comfortable confines to focus on mastering one core product: the almighty, know-it-all algorithm.

Howard Yu is professor of strategy and innovation at IMD business school, based in Switzerland and Singapore. In 2015, Professor Yu was featured in Poets & Quants as one of the Best 40 Under 40 Professors. He received his doctoral degree at Harvard Business School.