Tesla is on a roll. Are million-mile batteries and self-driving cars next?

It’s been a banner year thus far for Tesla, which saw continued profits and a skyrocketing—albeit, contentious—valuation, despite COVID 19-related sales declines in the auto industry.

Now, Tesla investors and industry experts are gearing up for the firm’s first Battery Day, to be live-streamed September 22, after its annual shareholder meeting. There, CEO Elon Musk is expected to unveil breakthrough battery technology and automotive architecture, including a longer-lasting “million-mile” battery that could drop electric vehicle (EV) prices to those of gasoline cars and contribute to the electric grid. And then there’s the question of whether Musk will stand by earlier promises to deliver Level 5, fully autonomous vehicles (AVs) this year, especially in the wake of a Consumer Reports review this month that found Tesla’s current self-driving capabilities falling short.

But will Tesla be able to deliver on those promises, from game-changing batteries to true AVs in 2020? The rumors that swirl around the company, its erratic founder, and inflated valuation cloud a more complicated truth. To achieve its goals, Tesla will have to pull together an assemblage of research, data collection, and acquired technologies to enable a battery that could revolutionize EVs, as well as beat out a host of self-driving competitors that are in many ways closer to achieving driverless vehicles than Tesla.

An overvalued company

Already producing the world’s leading EV, Tesla has also become the most valuable auto company in the world, despite selling millions fewer cars than traditional car manufacturers. It’s the first automaker to sell more than 1 million EVs, with its bestselling Model 3 topping 500,000 units. Tesla stock—which was going for $2213 a share before its 5:1 split on August 31—has enjoyed a 414% price increase this year, with a market cap of $412 billion, a figure that many analysts and even Musk himself deem overvalued. The stock’s performance has polarized analysts and investors on whether to value Tesla as a car or technology company.

“Tesla is hugely controversial,” says Lawrence Fossi, an attorney, Tesla blogger who goes by Montana Skeptic, and former portfolio manager with a short position in Tesla. “So why are people buying this stock? Some think it’s a growth company and the greatest disruptive transportation innovator in the past century. So, to them, the stratospheric price to earnings ratio [of 1137] is justified. Others are momentum traders riding the share price exuberance. And there’s a religious aspect of Tesla. It has this charismatic CEO and intensely devoted fan base that attacks its enemies. It’s a true cultural stock.”

Other reasons for heightened investor expectations have included aggressive marketing and incentives, ramped up battery research, increased demand in China, faster Model Y production, product fandom, and a long-term company vision that extends decades into the future, encompassing a solar and energy storage business that Musk believes will eventually outpace electric car sales. Musk has also expressed interest in licensing software and selling powertrains and batteries to accelerate sustainable energy. (Tesla previously supplied batteries to Mercedes and Toyota, but the deals fell apart due to disagreements over quality control.)

“A lot of retail investors want to support a company that is trying to change the world for the better by driving the shift to electric vehicles,” says Sam Abuelsamid, a principal research analyst with Guidehouse Insights. “Others invest because they believe in Elon. Of all his enterprises, like SpaceX, The Boring Company, and Neuralink, the only one that’s a public company is Tesla. So they buy Tesla stock as a proxy for all the other things Elon is doing—because they love seeing these rockets take off and land in perfect synchronization.”

“If you take out those credit sells, the business itself actually lost money,” says Abuelsamid. “They probably actually lost a lot more than what their filings actually indicate because there’s a lot of dubious accounting. They counted warranty services as goodwill repairs, so it doesn’t count against their net income. Based on just their core revenue, it’s probably not worth more than $5 or $10 billion.”

Charging up for Battery Day

This mixture of doubt and intrigue has fueled excitement about Battery Day. Musk is expected to give a tour of the company’s cell production at its Fremont plant, and announce cheaper, more efficient, and longer-lasting battery technology that can be recycled for energy storage. “It’s gonna blow your mind,” he teased last spring.

For its EVs, Tesla uses lithium-ion batteries with nickel-cobalt-aluminum cathodes, which it produces with Panasonic, and nickel-manganese-cobalt (NMC), which it buys from South Korea’s LG Chem. However, it’s working toward mass-producing its own EV battery cells through its Roadrunner Project. Last year, it acquired both battery manufacturer Hibar Systems and Maxwell Technologies, which makes supercapacitors and dry cell battery technology. In April, Tesla signed a deal with South Korean Hanwha Corp, which makes battery formation equipment. And since 2016, it’s been funding battery research by a Canadian team led by lithium-ion battery pioneer Jeff Dahn.

Last year, Musk began hinting at a “million-mile” battery that will enable greater range at a lower cost, and last a million miles—with an eye toward lowering EV car prices to those of gasoline cars. This year, Reuters reported that Tesla developed this battery with Chinese battery manufacturer Contemporary Amperex Technology (CATL) and Dahm’s lab. (CATL confirmed it had a battery designed to last 16 years and 1.24 million miles, though it’s uncertain whether this is the same battery.)

Meanwhile, Tesla followers have been searching for clues pointing toward other potential battery designs. Tesla has filed a number of patents for methods to slow battery degradation, improve efficiency and reliability, and reduce energy density loss during production. Some of Dahn’s research found that hybrid lithium-ion/lithium-metal cells deliver a 20% higher energy density than conventional lithium-ion cells. The Maxwell acquisition has folks speculating about combining supercapacitors, which store and discharge energy more quickly, with more durable dry cells. And rumors swirled when silicon nanowire firm Amprius Technologies—whose technology encases lithium in hairlike silicon wires for higher energy density—moved across the street from Tesla’s Kato Road facility in Fremont.

However, scaling battery production is also crucial to lowering costs. “Buying raw materials in volume is the main way to drive down battery costs, but volumes tend to be limited by Tesla’s ability to sell cars,” says Ed Niedermeyer, Autonocast podcast host and author of Ludicrous: The Unvarnished Story of Tesla Motors. “The plan for the Model 3 was supposed to be 800,000 to a million units in 2020, and we’re at about half that right now. They’re not supplying batteries to other automakers. Tesla has a reputation in the industry of not playing well with others, so I’m very skeptical that there’s going to be another deal with Tesla supplying batteries to another automaker.”

Another goal is eliminating cobalt, a staple of EV batteries because it doesn’t overheat easily and can store and transfer more energy than other elements. But mining it presents environmental and humanitarian hazards. Last year, a human rights organization sued Tesla and other tech companies for benefiting from forced child labor in cobalt mining in the Democratic Republic of the Congo. 

CATL reportedly plans to supply Tesla’s China operations next year with a nickel-manganese-cobalt battery with 50% nickel and 20% cobalt. And this year, Tesla struck a deal to use CATL-manufactured lithium-iron-phosphate (LFP) batteries in Model 3 cars for the Chinese market. LFP batteries are cheaper—potentially driving down car prices—as well as safer, charge faster, and last longer, which could improve car resale values. However, they have lower energy density and range for a given physical size compared to NMC batteries, which can travel as much as 400 miles on a single charge.

Meanwhile, Battery Day may unveil other innovations, including a structural overhaul to the Model Y to launch in Europe and China and built at Tesla’s Berlin Gigafactory. During an appearance at the plant, Musk teased a “radical redesign of the core technology of building a car,” while idiosyncratic software entrepreneur John McAfee followed up with, “Tesla will manufacture cars without axles and without frames.”

When a Twitter follower asked how he knew, McAfee (in a since-deleted tweet) replied, “I’m John Fucking McAfee. Google me.”

The barriers on Tesla’s road to autonomy

With Tesla’s stock and battery tech pulling focus lately, it’s uncertain whether Battery Day will include any self-driving progress. But the company’s future remains intertwined with its ability to live up to its promises to deliver driverless cars, something it’s currently struggling with. In July, Musk backed up predictions made last year when he mentioned via video at a Chinese artificial intelligence conference that Tesla was “very close” to achieving Level 5 autonomous driving technology this year, despite some projections that Level 5 won’t happen before the late 20s, if at all.

The Society of Automotive Engineers assigns six levels to categorize the amount of human intervention needed to drive a car, from zero (fully manual) to five (no human input in all situations). Level 3 is conditional automation requiring driver control at the upper limits of the system. Level 4, high automation, no longer requires a driver or human supervision, but has a limited operating domain and can stop if something goes wrong. Level 5, full automation, operates without drivers or constraints. Alphabet’s Waymo, Volkswagen/Ford’s Argo, and GM’s Cruise are a few of the companies testing Level 4 vehicles in various parts of the country in limited geographic areas. A 2020 Guidehouse report anticipates a modest deployment of Level 4 vehicles through 2025, with global volumes growing after, to some 13 million in 2030.

“Right now, you cannot buy anything more than a Level 2 vehicle, which is partial automation that requires human supervision,” says Abuelsamid, who co-authored the report.

Those classifications make Tesla’s Level 5 pronouncements and the monikers of its autonomous driving products all the more curious. While futuristic ideas are a staple of the Musk mythos, implying more autonomy exists than in reality has contributed to fatalities. Tesla’s Autopilot and Full Self-Driving Capability (FSD), the latter of which drivers can purchase for an additional $8,000, enable automatic lane changes, parallel parking, slowing for stop signs and traffic lights, and retrieval from a parking spot to where the driver is standing. But even Tesla states they require an alert driver prepared to take over at any moment, which is not complete autonomy. It further maintains that its new vehicles contain all the hardware needed for autonomous driving, once the company readies the necessary over-the-air software updates. Once full autonomy happens, Tesla owners could earn extra money by adding their cars to a shared network of robotaxis.

“Reliable automated driving systems need redundancy in their systems with multiple kinds of sensors that overlap, but also have different capabilities,” says Abuelsamid. “Tesla has only one radar sensor on the front of the car, eight cameras, and 12 ultrasonic sensors. They’re relying almost entirely on cameras, which are not very good at detecting distance. Lidar and radar are much better at that.”

By comparison, Waymo cars possess 360-degree imaging radar, near and far-sensing lidar, 29 high-definition cameras, and HD 3D mapping, which enable a richer data quality. But the Level 4 test cars are confined to smaller areas of operation.

Ultimately, Level 5 autonomy has to work everywhere. This brings us to a philosophical difference in how AV carmakers are building their neural networks, the computerized “brain” that recognizes patterns of real-world situations based on data observations. One approach is to incrementally build upon increasing amounts of limited datasets created by driving test AVs around select neighborhoods, staffed with human safety drivers ready to take over in case something goes wrong. The other involves sorting through a wider array of more varied data. Most AV companies favor the former, while Tesla is opting for the latter, gleaning real-world data and driver behavior from the millions of cars sold around the globe.

“Tesla has more data to draw from, but is limited in the quality of data it can get from fewer, lower-cost sensors,” says Niedermeyer. “But it needs a colossal amount of data because it’s making these decisions about the entire world. ‘Mainstream’ autonomous drive developers reduce the complexity of this problem by operating in a limited domain like a single city, using HD maps, and adding sensors like lidar, which makes them far less likely to make the wrong inference and endanger people.”

Limiting complexity also makes it easier for researchers to figure out why errors occur as they develop the system. On a finer note, Tesla’s real-world data, which it says comes from three billion miles (compared to its AV competitors’ tens of millions), are actually splices of information from those miles.

“Musk makes it sound as though they’re capturing all the data from all those miles and using that to train their automated driving systems when they are actually specific snippets,” he says. “They run experimental algorithms in the Autopilot computer in parallel with what is actually running in the car, and look for discrepancies between what their algorithm thinks the car should have done versus what the driver actually did. When they see those discrepancies, they’ll capture maybe two or three seconds of video to send back to Tesla for training the autonomous system, which doesn’t provide enough context to really be useful.”

In the short run, industry insiders have worried about Tesla possibly rushing its autonomous driving technology to market, especially given the industry’s lack of regulations and Autopilot-related fatalities. Last year, a National Transportation Safety Board investigation into another crash found Autopilot made it too easy for drivers to disengage, while the Consumer Reports review called its Full Self-Driving Capability a misnomer.

“Autonomous vehicles have the potential to have huge safety benefits and enable mobility,” says Abuelsamid. “But if you put it out there before it’s mature, you run the risk of killing more people instead of saving lives, and losing trust.”

As Tesla tries to rethink transportation and look to an autonomous green future, Battery Day could provide a more concrete roadmap of how it will get there—and whether it will feed the skeptics or live up to investors’ dreams along the way.