Microsoft Corporation (NASDAQ:MSFT) is getting into semiconductor design.
The announcement that Microsoft is hiring engineers to work on artificial intelligence (AI) designs in its Azure cloud has yet to impact the stocks of graphics chip maker Nvidia Corporation (NASDAQ:NVDA), but maybe it should.
That’s because Microsoft is the last of the five huge cloud companies, after Apple Inc. (NASDAQ:AAPL), Facebook, Inc. (NASDAQ:FB), Alphabet Inc (NASDAQ:GOOGL)(NASDAQ:GOOG) and Amazon.com, Inc. (NASDAQ:AMZN), to start work on its own semiconductors. Google is already on the third generation of its Tensor Processing Unit chips.
Because microprocessors are so expensive to make, nearly all U.S. semiconductor companies, like Nvidia, are “fab-less.” They design chips, they don’t make them. They’re software companies. And if hardware is software, it makes sense for cloud software giants to get into the business.
The Cloud Upgrade Cycle
When clouds were first deployed early in this decade, they depended entirely on inexpensive mass-produced hardware, along with open source software, for virtualization and distributing the computing load across many machines.
Now the largest cloud companies are looking to upgrade for a new generation of artificial intelligence applications, and they are sucking up all the talent they can find, or create, to do it. Google has helped create courseware on AI for Coursera, creating a machine learning specialization built around its TensorFlow.
Microsoft has reorganized its engineering staff, putting 8,000 engineers to work on AI, talking up ethics and growth to new computer science graduates. A March memo from CEO Satya Nadella mentioned AI no fewer than 18 times.
So far, the clouds’ hardware mavens are working on specialized designs for specialized purposes.
Google’s chip designers are working around TensorFlow , its own machine learning software library that uses a common framework and a Python front-end.
Facebook is looking to analyze video, so it can halt Facebook Live feeds that show people killing themselves or others. The company said it is specifically trying to reduce its dependence on Intel Corp. (NASDAQ:INTC) and Qualcomm, Inc. (NASDAQ:QCOM).
Apple has been working the problem from the other side, designing AI chips for use in devices like the iPhone and Macintosh. International Business Machines (NYSE:IBM) has been making chips for decades.
A New Market
These new competitive threats to mainstream chip-making have flown under the radar because the business is so white hot. General purpose graphics processors like the Nvidia GEForce have immense markets in gaming, cryptocurrency, and self-driving cars, as well as cloud upgrades. So far cloud chip efforts are focused on specialized designs for specialized applications, aiming to harness the raw graphics power Nvidia brings to the market.
But the cloud is becoming the market, as companies like Microsoft show that it is far more energy efficient to put workloads into its data centers than for companies to build, or even maintain, their own.
When chip companies approach this market down the road, they’re going to be dealing with companies that have as much chip design expertise as they do, and the financial wherewithal to get their custom designs produced.
No one can see this cloud upgrade market ending or even slowing, but all markets do that, as demand for what they’re delivering wanes. When it does, will the cloud giants continue buying chips from others, or will they just design their own?
It’s something for long-term investors to think about.
Dana Blankenhorn is a financial and technology journalist. He is the author of the historical mystery romance The Reluctant Detective Travels in Time, available now at the Amazon Kindle store. Write him at [email protected] or follow him on Twitter at @danablankenhorn. As of this writing he owned shares in AMZN and MSFT.