NEW YORK (AP) -- Elon Musk promised in 2019 that driverless Tesla "robotaxis" would be on the road "next year," but it didn't happen. A year later, he promised to deliver them the next year, but that didn't happen either.

Despite the empty pledges the promises kept coming. Last year in January, Musk said, "Next year for sure, we'll have over a million robotaxis."

Would you settle for 10 or 12?

Musk appears to be on the verge of making his robotaxi vision a reality with a test run of a small squad of self-driving cabs in Austin, Texas, starting Sunday. Reaching a million may take a year or more, however, although the billionaire should be able to expand the service this year if the Austin demo is a success.

The stakes couldn't be higher, nor the challenges.

While Musk was making those "next year" promises, rival Waymo was busy deploying driverless taxis in Los Angeles, San Diego, Austin and other cities by using a different technology that allowed it to get to market faster. It just completed its 10 millionth paid ride.

Boycotts related to Musk's politics have tanked Tesla's sales. Rival electric vehicle makers with newly competitive models have stolen market share. And investors are on edge after a $150 billion stock wipeout when Musk picked a social media fight with a U.S. president overseeing federal car regulators who could make the robotaxi rollout much more difficult. The stock has recovered somewhat after Musk said he regretted some of his remarks.

Tesla shareholders have stood by Musk over the years because he's defied the odds by building a successful standalone electric vehicle company -- self-driving car promises aside -- and making them a lot of money in the process. A decade ago, Tesla shares traded for around $18. The shares closed Friday at $322.

Musk says the Austin test will begin modestly enough, with just 10 or 12 vehicles picking up passengers in a limited area. But then it will quickly ramp up and spread to other cities, eventually reaching hundreds of thousands if not a million vehicles next year.

Some Musk watchers on Wall Street are skeptical.

"How quickly can he expand the fleet?" asks Garrett Nelson, an analyst at CFRA. "We're talking maybe a dozen vehicles initially. It's very small."

Morningstar's Seth Goldstein says Musk is being classic Musk: Promising too much, too quickly.

"When anyone in Austin can download the app and use a robotaxi, that will be a success, but I don't think that will happen until 2028," he says. "Testing is going to take a while."

Musk's tendency to push up the stock high with a bit of hyperbole is well known among investors.

In 2018, he told Tesla stockholders he had "funding secured" to buy all their shares at a massive premium and take the company private. But he not only lacked a written commitment from financiers, according to federal stock regulators who fined him, he hadn't discussed the loan amount or other details with them.

More recently, Musk told CNBC in May that Tesla was experiencing a "major rebound" in demand. A week later an auto trade group in Europe announced sales had plunged by half.

Musk has come under fire for allegedly exaggerating the ability of the system used for its cars to drive themselves, starting with the name. Full Self-Driving is a misnomer. The system still requires drivers to keep their eyes on the road because they may need to intervene and take control at any moment.

Federal highway safety regulators opened an investigation into FSD last year after several accidents, and the Department of Justice has conducted its own probe, though the status of that is not known. Tesla has also faced lawsuits over the feature, some resulting in settlements, other dismissed. In one case, a judge ruled against the plaintiffs but only because they hadn't proved Musk "knowingly" made false statements.

Musk says the robotaxis will be running on an improved version of Full Self-Driving and the cabs will be safe.

He also says the service will be able to expand rapidly around the country. His secret weapon: Millions of Tesla owners now on the roads. He says an over-the-air software update will soon allow them to turn their cars into driverless cabs and start a side business while stuck at the office for eight hours or on vacation for a week.

"Instead of having your car sit in the parking lot, your car could be earning money," Musk said earlier this year, calling it an Airbnb model for cars. "You will be able to add or subtract your car to the fleet."

Musk says Tesla also can move fast to deploy taxis now because of his decision to rely only on cameras for the cars to navigate, unlike Waymo, which has gone a more expensive route by supplementing its cameras with lasers and radar.

"Tesla will have, I don't know," Musk mused in an conference call with investors, "99% market share or something ridiculous." Given Waymo's head start and potential competition from Amazon and others, dominating the driverless market to that extent could be a reach.

But Dan Ives, a Wedbush Securities analyst and big Musk fan, says this time Musk may actually pull it off because of Tesla's ability to scale up quickly. And even skeptics like Morningstar's Goldstein acknowledge that Musk occasionally does gets things right, and spectacularly so.

He upended the car industry by getting people to buy expensive electric vehicles, brought his Starlink satellite internet service to rural areas and, more recently, performed a gee-whiz trick of landing an unmanned SpaceX rocket on a platform back on earth.

"Maybe his timelines aren't realistic," Goldstein says, "but he can develop futuristic technology products."