Why is it so hard for Silicon Valley's biggest tech companies to build a car?

Even if the driverless cars tech companies are betting on arrive sooner than expected, how anyone makes money from them is still an open question.

Apple badly needs something new.

On this morning's earnings call, after Apple released its results, two exchanges were worth noting:

Analyst: Does Apple have a grand strategy for the next three to five years? Or are you just reacting to how the market changes? Tim Cook: The strongest new products in Apple's history are on the way, but we won't reveal them ahead of time. We have a clear sense of where the market is headed, and we can also respond quickly to what users need.

Analyst: What are Apple's plans for cars? Tim Cook: I won't comment on rumors. But Apple is always looking for ways to improve the user experience. Plenty of technology is clearly going to "revolutionize" the car experience, but I'm not going to announce anything today.

Analysts are in a hurry to ask about something new because Apple's current products can no longer sustain its long-term growth.

According to the financial results released this morning, Apple's revenue for fiscal year 2016 (October 2015 to September 2016) fell $18.1 billion from the previous fiscal year, a decline of 7.7%.

This is Apple's first full-year decline since Steve Jobs launched the iPod. It was mainly the result of weak iPhone 6s sales.

Helped by the iPhone 7 Plus (and possibly by the Samsung Note 7), Apple expects revenue of $76 billion to $78 billion for October through December this year, slightly above the $75.9 billion it booked in the same period last year.

The decline has stopped, but there isn't much growth either.

As a public company, Apple needs to keep growing to hold up its market cap of more than $640 billion.

And the iPhone, which drove eight years of growth, can no longer guarantee that Apple will post a better number every year. The global smartphone market is essentially saturated — most people who would buy an iPhone already have one, and they are replacing their phones more slowly.

The iPod and the iPhone worked the same way: a better product remade an entire industry, and Apple kept the profits for itself.

For the past three years, the car has looked like the next iPhone Tim Cook wants.

But Apple's car project hasn't gone well

A project called Titan was reported by several news outlets in early 2015. Even Apple's chief operating officer, Jeff Williams, when asked about Apple's next big business, answered that "the car is the ultimate mobile device."

Apple's investment in cars far exceeds what it spent on the Mac, the iPod, the iPhone, or the iPad. A Morgan Stanley report shows that Apple's R&D spending rose by as much as $4.7 billion between 2013 and 2015. By comparison, R&D spending rose by only about $200 million while Apple developed the iPhone from 2004 to 2006, by $600 million while it prepared the iPad, and by $2 billion for the Watch and its new internet services.

A campus larger than Apple's headquarters has gone up. In July 2015, Apple was reported to have bought land in north San Jose, possibly for electric-car testing. There are now eight parcels covering 58 hectares in all. One large parcel near the airport is thought to suit a test track for car production.

Engineers and executives from traditional automakers, suppliers, and companies like Tesla have joined Apple in large numbers. Even Tesla's vice president of engineering was pulled in.

Everything pointed to Apple building its own car and making money from that market through a combination of software and hardware, exactly as it did with the iPhone.

But over the past three months, things have turned sharply.

Personnel and direction changes at Apple's car project, as reported by the media

2015.02

The Wall Street Journal exposed Project Titan. The project's lead, Steve Zadesky, had worked at Ford and had helped develop the iPod and the first iPhone.

2015.09

The Titan team grew from 600 people to 1,800; the goal of producing an electric car in 2019 was set.

2016.01

Steve Zadesky left Project Titan for "personal reasons."

2016.05

Apple invested $1 billion in Didi. This was probably more about keeping up relations with the government than about the car project.

2016.07

Bob Mansfield, a former Apple vice president who had run the MacBook and iMac product lines, returned to Apple as Titan's new head. The project was restructured around developing a system platform, and the original goal of mass-producing a car in 2019 was pushed back to 2021.

2016.09

Project Titan laid off "dozens of people."

In the middle of this month, Bloomberg reported, citing people familiar with the matter, that the Titan team would cut "hundreds" of employees, who would either be reassigned or choose to leave.

The whole project's direction is no longer to build a car of its own. This team of more than 1,000 people will now concentrate on an in-car system built with automakers.

Apple's leadership has asked the team to prove by the end of the year that a self-driving system is feasible, and that will decide the car project's fate and direction.

Just today came news that Apple had opened an office in Canada to poach people from BlackBerry's QNX team. QNX is an operating system widely used in cars.

Apple's current direction is closer to an operating system like Windows or Android. The iPhone of the car world has been shelved, at least for now.

Google — the public company with the second-highest market cap after Apple, and at work on driverless cars since 2008 — hasn't had an easy time either.

Chris Urmson, the robotics expert who ran the project as head of Google's self-driving effort, left in August this year. Two other key machine-vision experts went with him.

Like Apple, after building a large batch of retrofitted test cars, Google now wants to pull its core software out on its own so it can be ported onto traditional automakers' models and gather the data it needs.

If that plan works, it would amount to Google creating an Android for an auto industry more than a century old.

When it comes to cars, manufacturers have no intention of partnering with tech companies as equals

Tech companies have teamed up with traditional automakers to "digitize" the car before.

In 2007, Microsoft and Ford began jointly developing SYNC, an in-car entertainment system. It connected a user's device to the car over Bluetooth, USB, and the like, and SYNC later added vehicle health monitoring, emergency call connections, traffic monitoring, navigation, and messaging.

But at the moment, the major automakers are all developing driverless technology on their own or by buying stakes in smaller tech companies, hoping to steer the future themselves.

Their thinking isn't hard to follow. After all, the manufacturers that partnered with Google and Apple before are not doing well now.

According to figures from Canaccord Genuity, only Samsung and Apple turned a profit among publicly listed phone makers in the second quarter of this year.

Apple took 75% of the entire smartphone market's profit and Samsung 31% — the two shares add up to more than 100%, which means every other phone maker, most of them building Android phones, was losing money.

Google got 2 billion users out of Android and made some money from its app store, search, and video ads.

But the companies that partnered with Google earliest have had a hard time. HTC has posted losses for 15 straight months; Motorola was sold to Lenovo; Sony went through restructuring, layoffs, and asset sales, and the business it has left that comes closest to phones is supplying camera modules to other phone makers.

Apple has been no better. Its component suppliers have to hand over all their cost figures, and Apple then decides how much profit they may keep from working with it — a number that falls every year.

The Japanese screen supplier JDI, for instance, saw its losses in August this year widen to 1.5 times what they were a year earlier. Smaller, more Apple-dependent component suppliers such as Dialog Semiconductor, Cirrus Logic, and Skyworks have watched revenue growth slide from a blistering 45% to 150% all the way to negative growth in the latest quarter.

Carriers that partnered with Apple to sell subsidized phones haven't gained much either. They lost pricing power in their dealings with Apple, but to keep customers they have had to pay out large subsidies.

In 2015, the combined profit of China's three big carriers was only 70% of Apple's profit alone, and this year it will be far lower. Among U.S. carriers, which have worked with Apple longer, there is even a saying: "An apple a day keeps profit away" — a play on "an apple a day keeps the doctor away."

Ford CEO Mark Fields said in an interview last year that Ford would not work with companies like Google and Apple: "We don't want the in-car system to end up as a business like phones."

Neither CarPlay nor Android Auto can get Apple and Google into the car market

Ford's thinking is common across the auto industry. So Apple's CarPlay and Google's Android Auto have both been confined to simple infotainment, walled off from the car's driving systems.

In 2014, Audi, GM, Toyota, and Hyundai formed the "Open Automotive Alliance" (OAA) with Google, meant to "bring Android safely and seamlessly into the car."

Two years on, and even with support for more than 500 apps, Android Auto has not taken control of the car. Its main job is still audio and video. Users still connect a smartphone to the car and use the car's screen to check messages, look at maps, play media files, and be entertained.

None of that has much to do with how the car drives; the so-called "operating system" is about as meaningful as propping a smartphone on your dashboard.

Plenty of automakers even make buyers choose between CarPlay and Android Auto at the point of purchase, one or the other.

Technology is changing the car, and three trends stand out right now

Electrification

Automakers everywhere are changing their "energy mix" — more and more cars no longer run on gasoline, or not on gasoline alone. Japanese makers like gasoline-electric hybrids and hydrogen fuel cells; European makers lean toward clean diesel and plug-in hybrids; and the new car companies typified by Tesla have chosen the technologically independent path of the pure electric vehicle.

European automakers' diesel plans have been hit by Volkswagen's diesel-emissions cheating scandal, and with environmental rules tightening, European makers are expanding their electric-vehicle work.

At this year's Paris Motor Show, electric cars have become the mainstream. Europe's best-known automakers — Mercedes-Benz, Volkswagen, BMW, GM's Opel, Renault, and Citroën — no longer showed the clean-diesel cars that used to take center stage. The stars were electric cars, either close to mass production or planned for the longer term.

Ride-hailing replaces car ownership

Uber has long argued that once self-driving technology matures, nobody will need to buy a car — you will simply summon a driverless car cruising the streets to take you where you want to go.

Some research already shows that Uber has changed some Americans' willingness to buy a car. Real estate developers have even teamed up with it to replace parking spots with Uber.

Autonomous driving

Driverless cars are the other big trend. If new energy is something only European and Japanese automakers care deeply about, autonomous driving is something automakers everywhere are pouring money into.

Trends in the auto industry are pushed along by many things — environmental rules, for instance, drove the development of electric and hybrid cars. Driverless cars were mainly started by Google and Tesla: Google was the first to promote self-driving research as a formal corporate vision, and its test cars were soon on the road; Tesla, a new company, suddenly began selling a reasonably finished "autopilot" car faster than any traditional automaker.

Electrification was supposed to be Apple's opening

Apple's products have always aimed at the high end, and an electric car would have solved several of Apple's problems at once.

A high-end car in the traditional sense has to check off a few boxes: a smooth powertrain, meaning the car is fast; precise mechanicals, meaning it stays stable and easy to handle at speed; and a premium cabin, which usually means quiet, with good workmanship and materials like leather — though under the push toward sustainability, the materials are negotiable.

If Apple built an electric car, the mechanical and the quietness requirements would be largely met, because without an engine's vibration or a transmission's shift shock, the car's NVH system (noise, vibration, and harshness) loses two major sources of noise and vibration. That is Tesla's situation too, which means building a car would be no harder for Apple than it was for Tesla.

Electric cars are also easier to build. Tesla is the best example. Its first car, the Roadster, can be read as a modified Lotus sports car, but its second, the Model S, was already a fairly finished product.

Hemmed in by patents and designs tied to the internal-combustion engine, no company in history has gotten a gasoline-car business off the ground this fast.

Tesla has never solved its capacity problem, and even in its most important market, the U.S., it accounts for only 1% of annual vehicle shipments.

With far more money than Tesla, Apple might have had a chance to reach the mass market with an electric car, had it held to its 2019 mass-production plan.

But that assumed driverless cars were twenty or thirty years away.

Driverless cars may arrive far faster than anyone imagined

Autonomous driving is not a very new idea. In the 1980s, the driverless cars designed under Carnegie Mellon University's Navlab program were conceptually close to today's — both use sensors to see other cars and pedestrians.

But before Tesla announced its Autopilot feature in 2014, no automaker had shipped a commercialized system like it.

The U.S. National Highway Traffic Safety Administration (NHTSA) has proposed a formal classification for autonomous driving. The standard put forward by SAE International is similar:

• Level 1: The driver must control the vehicle at all times
• Level 2: The vehicle has individual automated features, such as electronic stability control or automatic braking
• Level 3: At least two automated features work together, such as adaptive cruise control combined with lane-departure warning
• Level 4: The driver can hand over control of key vehicle functions entirely under specific conditions; when the car detects a situation that needs the driver, it gives them ample time to take control back
• Level 5: The vehicle can drive itself completely, from start to stop, with no driver intervention, including parking. The vehicle is allowed to drive itself with no driver at all.

Before Tesla, quite a few cars already had automatic braking or collision warning, achieving Level 2 "autonomous driving." But with relatively simple technology, Tesla's Autopilot was more capable than any commercialized car's self-driving feature before it, reaching roughly Level 3.

Autopilot opened up in October 2015. Spurred by it, more and more automakers have brought out Autopilot-like technology this year.

Judging by what automakers are aiming for now, they are not developing along the autonomous-driving levels at all. Cars carrying lidar, high-precision digital maps, and image-recognition AI are aiming for full autonomy, not some transitional Level 4 version.

Almost every traditional automaker is developing autonomous driving in-house. Self-driving cars from Toyota, Ford, Volvo, and Mercedes-Benz have all begun road testing, and most makers have moved their target dates for commercial autonomous driving forward by about a decade, to around 2025:

• Nissan expects to launch a car that can drive itself on multi-lane highways in 2018
• Volvo expects to launch a car that "actively prevents occupant injuries" in 2020
• GM, Mercedes-Benz, Audi, BMW, and Renault all expect to launch cars that can drive themselves part of the time around 2020
• GM, Fiat, and Daimler plan to launch commercialized, fully autonomous cars in 2025

In a driverless world, Apple has no business model yet

For automakers, the driverless car is a new selling point the wider environment has forced them to develop faster. For Apple, whether "driverless cars" can be a good business has no answer yet.

Research from Columbia University's Earth Institute and PwC both predict that mature autonomous driving will shrink the U.S. car market sharply, because ordinary people will no longer need to buy a car.

Columbia's Earth Institute puts the total number of cars in the U.S. market at around 20 million; PwC's forecast is more aggressive, as low as 2.4 million.

A better smartphone helps manufacturers sell phones to more people. With driverless cars it is exactly the reverse — a better driverless car means fewer people need to own one.

As personal purchases fall, Uber-style driverless fleets will need large numbers of driverless cars. But those cars exist to carry passengers for money, and just as taxis are usually not the best cars, fleet cars won't chase quality either. A market like that cannot meet Apple's demand for profit margins.

Second, Apple started very late on the AI research that autonomous driving requires. As a frontier field, AI research depends on universities and research institutes and needs scientists around the world to talk and work together — which runs straight into Apple's culture of secrecy.

Only last week did Apple get a positive example: Russ Salakhutdinov, an AI professor at Carnegie Mellon University, announced he was joining Apple to lead its future AI research while keeping his university post.

For Apple that is a breakthrough. For companies like Google and Microsoft it has long been ordinary.

Google started on driverless cars first, and still hasn't found a way to make money

Google's driverless-car project began earliest among tech companies. The research team led by Sebastian Thrun grew out of a project at Stanford University's AI lab, which won a $2 million grant from the U.S. Department of Defense as far back as 2005. In 2007, the team was folded into the Google X lab.

Since then, Google's driverless-car team has used test cars retrofitted from ordinary vehicles to complete 2 million miles of real-world road testing, and has built an entirely new "pod"-shaped, Google-branded self-driving prototype. But Google's vision was never to build cars to sell to consumers: John Krafcik, head of Google's self-driving effort, has said that building cars is too hard and Google won't do it.

So where can this project be commercialized? Package the self-driving module and sell it to automakers?

That model might have worked a little earlier — Mobileye, which supplied Tesla's self-driving module until this May, ran exactly that kind of licensing-fee business.

But automakers are now developing their own driverless technology through acquisitions and in-house research. Volvo even wants to turn itself into a supplier, packaging and selling its own self-driving solutions, and traditional suppliers such as Delphi have the same idea.

Unless Google were the only company able to develop driverless technology, the technology itself cannot be the barrier that lets Google charge for it.

Could it instead be given to small automakers free, the way Android is?

Android can be free because Google makes its money mainly from the app store and ads. In cars, no one has yet shown how to make money from services.

Set driverless cars aside: building a car is hard enough on its own

The Silicon Valley company that has gone furthest in building cars is Tesla.

It was the first to build an electric car it could charge a high price for, and the first to sell driving technology still in testing to its own customers.

But its real test has only just begun — the cheaper Model 3.

Tesla's Fremont factory has a target annual capacity of 500,000 cars, but weighed down by battery capacity and production-line upgrades, Tesla delivered only 24,500 cars in the third quarter, with more than 400,000 Model 3 orders still waiting to be built.

Beyond capacity, Tesla's quality problems have surfaced as more cars get delivered. A new report this week from Consumer Reports, the most authoritative independent consumer-testing magazine in the U.S., found Tesla's cars highly unreliable: out of 29 automakers ranked, Tesla came in 25th.

The main problem is the new Model X. Consumer Reports said the car has trouble with its doors, brakes, locks, power equipment, in-car electronics, and climate control, and it scored last for reliability among more than a dozen luxury mid-size SUVs. Unusual features on the Model X — falcon-wing doors, powered middle-row seats, a panoramic windshield — have brought higher failure rates.

For an automaker, building a high-end car is much easier than building a low- or mid-range one. Rolls-Royce, for instance, makes some of the most luxurious cars in the world, each one a hand-built work of art. But a product like that cannot be turned out by the millions each year, the way a family sedan is, and still be reliable.

Once annual production runs into the hundreds of thousands, reliability becomes the first priority. For someone buying a family car like the Toyota Corolla, not breaking down, saving fuel, and driving smoothly are what matter most. Tesla still has to prove all of that with the Model 3.

Even selling cars carries policy risk

At the end of last month, Tesla sued the state of Michigan for the right to open company-owned stores and sell Tesla cars in the state.

Michigan, Texas, Connecticut, and Utah all flatly refuse to let Tesla sell direct within their borders, because of America's "franchise dealer laws." Grown out of 19th-century federal law, these are state laws that many U.S. states passed to protect vulnerable car dealers, requiring automakers to sell only through franchised dealerships (4S stores) rather than opening their own.

On that foundation, every state has formed a "car dealers association" to protect dealers' interests. It is precisely this group that pushed through the legislation banning Tesla sales in some states.

Tesla differs from an ordinary car in that it needs little maintenance and is more like a piece of electronics; selling directly through its own stores is the best way to maximize profit and control the buying experience.

Tesla now runs company-owned stores in 23 U.S. states, but in states that don't allow direct sales, such as Texas, its stores can only display cars — staff aren't even allowed to tell customers the website address where they can place an order.

If Apple ever built a car of its own, Apple, so used to running its own stores, would run into the same problem.

The hottest new business in the car market has nothing to do with tech companies either

In that same U.S. market Silicon Valley companies most need to win over, nearly half of the best-selling vehicles recently have been high-performance pickups and SUVs:

The top three are three perennial favorites in the U.S. market: the Ford F-Series, the Dodge RAM series, and the Chevrolet Silverado series. According to Citibank's figures, pickups and large SUVs contributed more than 67% of Ford's and GM's profit in 2016.

In the first three quarters of 2016, these big vehicles accounted for 19% of all U.S. car sales.

Since 2005, meanwhile, pickups have been moving toward the relatively "high-end" part of the market — average prices have risen from around $30,000 to more than $46,000. By comparison, average prices for mature sedan models rose only 10% over the same 10 years.

The hottest and most profitable market in the auto industry has nothing to do with the sharing economy, autonomous driving, or electrification and sustainability.

Just as the direct-sales fight is confined to a few states, pickups aren't hot everywhere in the world either.

But the U.S. is still the world's largest car market, and the car market most open to new technology. Apple, Google, Tesla — all of them have to win it first.

And a market like that may genuinely not suit an Apple car.

Cover image / New York Times