Driverless Cars, Robots, Other Future Bets — Pivoted or Killed. Google Is Still an Ad Company | Top 15 Annual Reports

How to make money from artificial intelligence is a question Google, too, has to answer.

At 5 p.m. on March 15, Go world champion Lee Sedol resigned to the "machine" after 280 moves. Across the five-game match, Google's AI algorithm AlphaGo won four.

Few had predicted that before play began.

Go is held to be the hardest board game in the world: the possible moves outnumber the atoms on Earth, so no computer can brute-force its way to the best answer. When IBM's AI beat the chess grandmasters in the 1990s, Go was named as the one challenge computers could never clear.

For years, news like this built Google's image as a technology pioneer. Besides Go there were contact lenses that measured blood sugar, walking robots you couldn't knock over, and those endearingly round driverless cars. Such stories ran far more often than the advertising business that brings in nine-tenths of Google's revenue.

In 2015 Google reorganized into Alphabet — a name that nods both to the alphabet and to a big bet — and these future-tinged projects became known as Google's "other bets." Their job was to explore what Google might be beyond advertising.

But this year these once-ambitious projects were shut down, sold off, or pivoted. Almost none of it went smoothly.

The Party Is Over

In January, the secretive division inside Alphabet was renamed X. It took a new logo and recast itself as "Alphabet's incubator," pushing mature frontier projects to "graduate" into standalone Alphabet businesses rather than stay science-fiction concepts.

People slowly grasped the other side of that message: frontier projects that had failed to turn commercial would no longer get Alphabet's backing. What awaited them was shrinkage, pivots, or the axe.

From early in the year, the star executives running Alphabet's frontier projects left in a wave. First came Mike Cassidy, head of the high-altitude balloon Wi-Fi effort Project Loon; in August, Chris Urmson, who ran Google's self-driving car project, also left; in November, drone-project executive David Vos quit alongside his boss, Sean Mullaney.

The contraction reached beyond the X Lab moonshots with no path to profit. Alphabet subsidiaries that were missing their numbers got hit too.

In March, smart-home subsidiary Nest was reported to have fallen short of Alphabet's sales targets. Three months later, Nest CEO Tony Fadell resigned.

In May, Boston Dynamics, the robotics company Alphabet had bought three years earlier, was sold to Toyota — the end of Google's main bid to enter the commercial robotics market.

In September, telecom subsidiary Access shelved Fiber, its ultra-high-speed broadband project, because it was expanding slowly and earning back its money even more slowly. CEO Craig Barratt resigned and the team was deeply cut — the end of Google's bid to become a broadband provider.

ATAP, the unit meant to build frontier products that could go commercial within five years, also changed beyond recognition. Its head, Regina Dugan, was poached by Facebook, and its three-year modular-phone project was killed.

Alphabet does not invest in new projects out of charity. The point is to find growth beyond advertising.

The cuts we kept reading about this year didn't necessarily mean these frontier projects had flopped. They meant the parent company, Alphabet, could no longer bankroll them the way it once had, no matter the cost in money or time.

Take the robotics division. Boston Dynamics, the company Alphabet sold, was less a "robotics company" than a robotics research institute. It began as a frontier project funded by the Pentagon and lost military backing once practical use proved far off.

Google bought it in 2013 and never rushed to make products from it, just explored new robot forms — like the can't-knock-it-over robot mule that went viral on video sites at the time. Fun, yes, but it couldn't go far from a power cord, and no one could say what it was actually for.

After Boston Dynamics was sold, what stayed in the robotics division were Perception, an industrial-robot project, and Redwood Robotics, a robotic-arm project. Both looked far more commercially useful in the near term than Boston Dynamics' machines, even if they also looked more conventional.

"Google is not a conventional company. We do not intend to become one." Larry Page, Alphabet's founder, wrote that in his shareholder letter when the company went public in 2004.

Now everything has turned conventional.

Innovation Spending Shrinks Because the Advertising Ceiling Is Close

Since it was founded in 1998, Google's main source of money has worked the same way: offer free, useful internet services, gather user data, then show users targeted ads. Last year 89 percent of Google's revenue came from advertising.

Google is still the world's largest advertising company, and its revenue is still growing.

But the ceiling on that growth is close. Over the past five years, the ad money Google earns per user has barely moved.

What has driven Google's revenue is the growth in the total number of internet users. But outside war-torn and poor regions, fewer and fewer people are left to come online.

Here is another, more telling figure: in 2015, 56 percent of Google's ad revenue came from just two markets, the UK and the US, and 44 percent from everywhere else. That split hasn't changed in three years.

From 2013 to now, the global internet user base outside China has grown by 547 million — and almost none of that growth was in the UK or the US. So the new users barely add to Google's ad revenue.

If most of the new users Google gained over these three years can only afford a few-hundred-yuan Android phone, how much is their data worth to advertisers?

And then there is Facebook. Search ads and the banner ads beside web pages have always been Google's main source of money.

As users spend more and more time on their phones, more of that time goes to social apps — a business led by Tencent in China and owned almost entirely by Facebook everywhere else.

That means Google has fewer ad slots to show each new user than it did in the desktop era.

For Alphabet, the picture is not encouraging. In the third quarter of this year, the number of paid-ad clicks on Google across PC and mobile rose 33 percent — but the price per thousand clicks fell 11 percent. Ads no longer fetch a high price.

So Alphabet Has Started Cutting Investment in New Businesses

Most of the changes to new projects you saw this year trace back to CFO Ruth Porat.

She moved over in March 2015 from Morgan Stanley, where she had been CFO and earned the title "the most powerful woman on Wall Street." Under the pressure she put on the company's finances and management, Google's share price rose 40 percent within a year.

Last year she single-handedly drove Alphabet's restructuring, splitting the company into its core advertising business (Google) and the "Other Bets" subsidiaries. She also began cutting costs company-wide, from capping hiring targets to trimming the free meals Google's staff once bragged about.

Employees unhappy with her privately called her "Ruthless Ruth," but investors loved her. In February of this year, at Porat's urging, Alphabet began disclosing the revenue and losses of "Other Bets" in its earnings reports, stripping the cover these cash-burning frontier projects used to enjoy.

The frontier projects beyond advertising exist to find new growth. But if you treated all those many subsidiaries as a single company, its financials would look dreadful:

Last year these new projects cost Alphabet $3.5 billion. This year, under Porat's belt-tightening, that figure may drop somewhat.

On this kind of trajectory, Alphabet can no longer pour money into frontier technology the way it once did.

After the Pivot, Alphabet Eyes Two Traditional Businesses: Cloud Computing and Premium Hardware — Both Led by Google Itself

In late November last year, Alphabet's subsidiary Google poached Diane Greene, co-founder and former CEO of cloud-computing provider VMware.

This was not a routine hire but the start of an organizational overhaul. In October of this year, Google placed its enterprise communication service Google for Work, its cloud office software, Cloud Platform, and even enterprise-grade Android phones, tablets, and Chromebook computers all under Greene. It was the biggest restructuring Google went through this year.

In November, two of Silicon Valley's most famous AI researchers joined the cloud division: Fei-Fei Li, head of Stanford University's AI lab, and Jia Li of Snapchat's research group, who will use artificial intelligence to improve Google's cloud products.

On building better products, Google says it now has the data to prove it: 60 percent of Fortune 500 companies use paid Google cloud services — though most of them mainly use email and online documents.

Even so, in cloud services Google has not closed the gap with Amazon and Microsoft.

It isn't only a matter of market share.

In the cloud business, the better product doesn't necessarily decide the outcome. You also need a full sales team for enterprise customers, strong after-sales and support, and years of accumulated client relationships — none of which resembles the advertising-sales model Google has spent its 18 years perfecting.

The other new business is learning from Apple: selling high-margin phones. But here, as with the cloud, Google has put itself in a market that plays by entirely different rules.

On October 3 this year, Google launched the Pixel, a phone it designed itself.

Only a small part of the event was about the AI assistant. Most of it looked like an ordinary smartphone hardware launch: specs, hardware design, a better camera, a charge to 70 percent in 15 minutes.

Even on price it went head to head with Apple. The Pixel starts at 32GB and costs exactly the same as the iPhone 7.

From the website copy to the search keywords, the message was the same: Google can make a very good phone too.

The phone is sold mainly in the US for now. It caught the wave of the Note 7's withdrawal and its early sales were decent, but how it does in wider markets has yet to be proven.

As with cloud computing, selling phones is something Google doesn't know well. Supply chain, carrier deals, marketing, sales channels, ongoing hardware R&D — every one of these is something Google has to learn afresh.

And Google's corporate structure was never built to sell hardware. Its teams are organized by product: Android is one division, Gmail is one division, YouTube is one division. Each has its own goal — make a better software product, not sell more phones.

Apple is very different. Apple divides its units by function, and the whole company runs toward one goal: build better hardware and sell it. Every generation of iPhone draws on the energy of every executive on Apple's top team.

This creates Apple's own problems — it isn't good at internet services. But it also ensures Apple can build better hardware than its rivals and charge a higher price.

That is the treatment the Pixel will not get inside Google.

Unless Google makes some major changes, or AI becomes the one feature consumers care about when choosing a phone, selling phones won't be easy for Google.

There Is Also the Question of How to Make Money from AI

After the AlphaGo match in March, everything at Google seemed to hook onto artificial intelligence.

In May, at Google's annual I/O conference, the company unveiled several new AI products — among them Google Assistant, an AI assistant you can use on phones, TVs, watches, even home devices. This year Google wants to make AI the force driving all its products.

This shows up not only in hardware but in its internet services. Google Assistant works like an all-purpose AI: you can chat with it, ask it about the weather and your schedule, and ask it about anything under the sun.

On this year's second-quarter earnings call, Google CEO Sundar Pichai said the whole company was already using AI to improve its products. "Machine learning is an engine that can take us into the future. Google now has more than 100 teams using machine learning — from Street View and voice search to email, and much more."

Only two things matter most in AI: the algorithms, and the volume of data used to train them. On both, Alphabet is far ahead of every other company.

But when it comes to actually making money from AI, Alphabet has fallen behind its rivals.

The clearest and most important case is the driverless car.

Google announced its self-driving car project back in 2010. By 2015 it had built a prototype from scratch — a cute, round-bodied two-seater with no steering wheel and no foot brake.

This year, more than a hundred thousand semi-autonomous cars that know how to overtake, dodge obstacles, and park themselves took to the roads.

They came from Tesla.

Soon after, Uber's driverless trucks and self-driving cars also began commercial trials.

It wasn't only tech companies that beat Alphabet to commercial use; the traditional automakers looked set to move faster too. General Motors, Ford, Toyota, and the German carmakers all announced driverless-car plans, most with timelines inside five years.

Meanwhile, almost every executive who had been with the self-driving project from the start had left Alphabet. The project was rushed to "graduate," dropping the idea of building its own car and instead partnering with Fiat Chrysler, a traditional automaker whose own technology is not especially strong.

The reason behind all this can be summed up in something Steve Jobs said at a 1997 launch event, answering a question from the audience, after he returned to Apple:

"I've always found that you've got to start with the customer experience and work backwards to the technology. You can't start with the technology and then try to figure out where you're going to sell it."

Tesla, Uber, and the traditional automakers all trail Google in AI technology. But all of them pushed self-driving forward by asking how to build a better car.

Their early cars couldn't drive themselves for real and had some accidents. But they got on the road faster and kept refining them, working out what it actually takes to build a better self-driving car — and without asking too much of consumers' habits.

Google was different. It used self-driving technology to rethink, from scratch, what a car should be, without considering how to commercialize it or whether users would accept it any time soon.

While hundreds of thousands of rival cars with some self-driving ability drove the roads every day, Google was still at the laboratory stage.

AI is an important technology, but it is more like the protocols behind the rise of the internet in the 1990s — vital, yet not a selling point.

People use email because it solves a communication problem, not because of POP3 or IMAP. People use Taobao or Amazon because shopping is convenient, not because of HTTP.

In the same way, if AI is to create enormous commercial value, it will be because it solves real problems, not because the technology itself is advanced.

Sometimes a company can find a technology first and work out how to make money later. Google's own existence is an example: it began as a university research project, and advertising was figured out only a few years after Google search was born.

That kind of success runs in Google's blood. To this day, not one of its businesses has gone any direction other than backward from technology to product.

And to this day, not one of its products makes money by any means other than search advertising.

When the restructuring was announced last year, a Google executive said: "We don't want to face a future where Google is just an ad company with a lot of hobbies."

But turning those hobbies into the main business may require Google to break its habit of always working backward from technology to product.

About the Top 15 Annual Reports

2016 was a turbulent year around the world. Judging by how companies performed, the market leaders had a harder time than the challengers chasing them.

The Top 15 companies QDaily follows — the ones leading changes in how people live — had a rough year, and their shared trouble came down to a single word: ceiling.

Whether launching new products (like Apple's watch and car) or reaching new users and channels (like IKEA moving into e-commerce), these industry leaders were urgently — and perhaps blindly — searching for their "next" growth engine.

Breaking through that ceiling has proved harder than expected. The 2016 Top 15 year-end reports are not all good news.