‘I want to see red’: General Motors is teaching its engineers to fail fast and ‘speak fearlessly’ as they take the EV battery fight to China

Fitting in like a bright Lego set is the $900 million investment in the Battery Cell Development Center (BCDC), a massive light-gray rectangle, and the Innovation Center, which resembles a blue-and-yellow battery pack sitting on top of a historic building.

If GM is a giant campus, longtime GM battery engineer Andy Oury is the cheerful science teacher, polo shirt tucked into his blue jeans, driving around in his large (all-electric) Silverado. Just the weekend before, he said, he drove it 220 miles to Traverse City, where he powered his camper off the truck all weekend and towed a wood chipper around town. 

Oury joined GM in 1999 and became interested in batteries about a decade later, as GM began work on the Chevrolet Volt. This plug-in hybrid convinced him GM was serious about electric transportation, he told Fortune. Oury reports to Kurt Kelty, a battery lifer who spent over a decade each at Panasonic and Tesla. About a decade ago, Oury said, he noticed on LinkedIn that Kelty was leaving Tesla, and he called his boss at the time: “Hey, you’ve got to try to hire this guy.”

Kelty, now GM’s vice president of battery and sustainability, told Fortune that the battery operations and research labs have been a massive investment at a time when Detroit is largely reeling from a failed bet on EVs. The federal EV tax credit expired at the end of September 2025, and GM took about $7.6 billion in EV-related charges last year, although that was a fraction of Ford’s and Stellantis’ similar charges in their EV divisions ($19.5 billion and $26.5 billion, respectively). GM says these charges are not comparable as it is sticking with EVs as its end-game scenario.

The battery plant is still very much a work in progress, expected to come online as fully operational later this year. Anselmo “Mo” Gallegos is the director of battery cell development, responsible for rapidly scaling up science class into an industrial process. If Andy is the cool teacher, then Mo is the brainy grad student, grinning from behind his glasses or safety glasses, and he’s also prone to food metaphors. Every battery maker has a proprietary formation protocol, the final charging process that brings a cell to life, he said: “a recipe to get the perfect donut at the end of it.”

courtesy of General Motors

Walking me around the ovens and mixers, some of them a tangle of pipes and wires that recall the bombastic 1960s art of Marvel Comics great Jack Kirby, Gallegos casts himself as a chef seeking just the right mix of ingredients. Batteries, he said, are “a mix of chemistry, mechanical interfaces [and] dark, dark black magic.” The ovens are where the proverbial cake can be overbaked. “Is it going to stick? Is it going to crack?” he said. “You get your cake out of the oven and sometimes it’s got cracks along the top. We hate that.”

Mo’s team is ramping up to make a couple thousand full-size cells a day, and Gallegos’ job is to work out the process so that every one of them comes out smooth, no cracks. Before building the BCDC, GM couldn’t build a cell like this at all. Engineers drew one on paper, took it to a partner like Samsung or LG and asked, in Oury’s words, “pretty please, will you put this together?”

What the plant is built to prove is a big bet on chemistry. Much of the world’s cheap EV battery supply runs on lithium iron phosphate, or LFP, a chemical that China dominates. GM is betting that its next generation of electric trucks and full-size SUVs will run on something else: lithium manganese-rich, or LMR, cells, which swap out much of the nickel and cobalt for cheaper, more plentiful manganese and extra lithium. GM’s LMR cells use about 30% to 40% nickel, Oury said, compared with roughly 85% in the high-nickel cells currently being used. GM says LMR can deliver roughly a third more energy density than leading LFP cells at comparable cost — and Gallegos and Oury are scrambling to start commercial production with partner LG Energy Solution in 2028.

Kelty explained that GM has been working on LMR for more than a decade and considers it homegrown. “We’re not licensing this from China,” he said. “We’re developing this here.” If it works, GM gets a cheaper, longer-range battery for its biggest vehicles that doesn’t depend on Chinese technology. If it doesn’t, GM will have wagered on a chemistry that, by Kelty’s own account, most of the industry considers a long shot. 

courtesy of General Motors

“Our data showed one thing,” Kelty said: “that this actually made sense.” The question inside GM became why the company wasn’t already moving. “It took a voice saying, the reason we’re not doing LMR is because it’s us. It’s our own broken mentality. So let’s fix that,” Oury recalled. The hurdle, he said, was “actually more momentum-based or emotion-based rather than data and business-based.”

When GM made the call, Kelty said, the team had supporting data to get it perhaps 60% to 70% of the way to the decision. “What we had to do was make that last leap and say, ‘You know what? We’re going to get the next 30% or 40% of the data and make it there.’” He called it a “testament to the trust in the team.” GM hasn’t fully abandoned LFP; in July 2025, it announced it would convert its Spring Hill, Tenn., cell plant to produce LFP cells, but GM chose LMR for its biggest vehicles. 

It’s been a big change for America’s No. 1 automaker, which has long produced, to recall Alfred P. Sloan’s famous phrase, “a car for every purse and purpose”  — and whose employees pride themselves on being A students who know all the answers. Leaders at GM told Fortune they’re trying to grow a different kind of high achiever: someone who challenges inherited constraints, works from incomplete information, admits what won’t get done and uses early failures to learn faster, without missing core deliverables. But it’s not easy to accept a few Bs when you’re used to being an A student. 

courtesy of General Motors

Red light, green light

When Kelty joined GM in February 2024, he began a new quarterly process with his team. At the start of each period, he spends an hour one-on-one with each direct report, agreeing on goals. He describes his job in three parts: “set the direction, set the priorities, and then remove obstacles.” At the end of each quarter, the work is marked red, yellow or green.

“At the end of the quarter, when we look at it, there should be some red lights on there,” Kelty said in an interview with Fortune. “And if there isn’t, then to me that means that they didn’t stretch far enough. So I want to see red.”

The goals have to be measurable, he said, citing something he might assign in batteries: “We’re going to make the cells, the cells are going to cycle 500 times and they’re going to have at least 90% retention by the end of the quarter.” His teams now run their goals through AI to check that each one has a metric attached, he added. “I want to make sure that we have some stretch targets in there,” Kelty said. “I don’t want just goals that they’re doing already.”

At GM, Kelty said he found a team that treated the quarterly scorecard like a report card. “The team was just paranoid that everything had to be green because that’s kind of the GM culture,” he said, adding that he comes from a culture where “if everything’s green, you obviously are not stretching hard enough.”

At his first staff meeting, Kelty asked his new direct reports to bring their top priorities for the year and the top two things they would not do. Several called Oury to ask whether their new boss was serious. As a longtime GMer, Oury said the instinct was to never say no, “because it always feels like you have the resources to get it done.” But the truth, he said, is that “no company, no matter how big they are, has the resources to do everything. So having that — building that muscle of saying no and prioritizing — is super important.”

Kelty puts it more bluntly: “If you don’t have a list of things you’re not going to do this quarter, then I assume you have too many people.”

It took about a year for the team to buy in, Kelty said. What helped was an example. At a meeting with his leaders, he shared an anonymized quarterly review, full of red lights. “They didn’t hit hardly anything on their targets, but I gave them the highest ranking of all of you,” he recalled telling them. As he put it to Fortune, “The person that got the most red actually was the highest-performing person on my team.”

Some reds are delays. Some are experiments that didn’t work out. Either way, Kelty said, “Let’s move on.”

The view from the CFO’s office

The finance chief, Paul Jacobson, has steered the company into a position where Kelty has room to dig deep and experiment. The former Delta Air Lines CFO has held the same position at GM since 2020 and on his watch, GM has generated annual cash of about $10 billion since 2021, compared to an average of $3 billion for the five years prior, while buying back roughly a third of its shares since late 2023. 

He said he learned from the airline sector to take out what he called “self-imposed cyclicality” from piling-up inventory and price slashes during soft demand. That cushion pays for bets like Kelty’s, he said. “What we really want to do is create a continuous stream of healthy cash flows.” The company recorded $185 billion in revenue and $2.7 billion in net income in 2025, with 2.85 million cars sold domestically, a 17.2% market share. 

Victor J. Blue/Bloomberg via Getty Images

Jacobson has run a bit of a stretch experiment in public. At GM’s 2021 investor day, he laid out 10-year targets that were unusual for an automaker to put on paper, including $20 billion in annual digital revenue by 2030. Jacobson told Fortune he thinks the market discounted that heavily at the time, and is now more focused on quarterly performance. GM stock’s price-to-earnings multiple is one of the lowest in the S&P 500, he noted, even though it has generated over $10 billion a year in free cash flow for almost five years now. 

“People take a view that’s more historical,” he said. “They look at the GM of the ’90s and the 2000s and the restructuring era versus really looking at what’s going on today.”

The oracle

Sterling Anderson, another former Tesla executive, may be the clearest sign of how far GM is willing to go to remake itself: it handed one of its most important jobs to someone from the tech world. And it handed him a distinctly tech-flavored job title: chief product officer.

Announcing his hire in May 2025, GM President Mark Reuss said the company had “an opportunity to evolve the way we build from the ground up, with tighter integration between software and hardware, shorter development cycles, and an unwavering focus on a seamless customer experience.” CEO Mary Barra said Anderson was joining “at a critical time as our industry continues to reinvent itself.”

Holding a master’s and PhD in robotics from MIT, Anderson developed MIT’s groundbreaking semi-autonomous vehicle safety system, Intelligent Co-Pilot, and later led Tesla’s Autopilot program before co-founding autonomous trucking company Aurora. He joined GM after conversations with Barra, Reuss, Jacobson and the board of directors and told Fortune he made the decision for three main reasons: GM’s trajectory, vision and scale. 

IAN MAULE/AFP via Getty Images

The car is “the most complex robot you and I and most of us will probably ever own,” he said. What surprised him in those conversations, he added, was GM’s “appetite for transformation.” “I saw a much higher neuroplasticity in this leadership team than I’ve seen from many others in this and related industries.”

Anderson said he’s especially pushing a “no sacred cows” mentality. “Engineers thrive when you take the constraints off the problem.” For instance, he said he challenged a team to come up with voice-controlled in-vehicle experiences, and a four-person team had a prototype running in a Cadillac Lyriq about 30 days later. Old GM would have asked, “Who’s the supplier we can go talk to and integrate a module?” he said. Now, his product org gives out an award called “Big Swings,” for “effort that swings through the fences, whether successful or not.” 

Anderson described GM’s overhaul of product development as “effectively the third phase in the evolution of product development for humanity,” built on “probabilistic, large data-requiring methods.” In plain terms, that means fewer handoffs between departments, shorter timelines and more decisions made by simulation and data instead of trial and error. 

“It is definitely a philosophical bent,” Rashed Haq, GM’s VP of Autonomous Vehicles, said of Anderson. Haq, who came from a robotics background before joining GM in 2020 to work on its self-driving unit Cruise, added that many leaders in the physical AI space have to “think very deeply” about these issues. “I think everybody in [autonomous vehicles] thinks of [the car] as a robot. When we had been hiring, we either hired people from a deep, [machine learning] background or robotics background,” he added. He estimated that half of his team has PhDs in robotics, and the other have PhDs in deep learning or machine learning.

Some aspects of the new GM are taking on a downright Big Tech-ish tone, evoking the radical candor from the halls of Amazon’s “disagree and commit” or Bridgewater’s “radical truth and radical transparency.” 

About a year-and-a-half ago, Kelty said, GM began encouraging employees to open a disagreement with a set phrase: “I want to grant myself permission to speak fearlessly.” Whatever follows afterward is understood to be fair game. It’s part of how the LMR debate broke open, Oury said. Recently, on a call with a senior vice president listening in, Kelty said an engineer from a neighboring group used it on him: “Kurt, I want to have permission to speak fearlessly. What you said is absolutely wrong. It is not the way we should be going.” Kelty grinned. “It’s great to have that.”

The B-plus A student

Jason Fischer, a 28-year GM veteran who leads the AI push inside engineering, described the old instinct as a hunt for perfection. “The old GM would have been like, ‘Wait till this is the most perfect,’” he said. He said the end result is the same high standard as always, but the company is taking pains not to “stifle innovation.”

Fischer turned to a high-school metaphor to refer to himself as a perpetual learner. “I tell people I’m a B-plus student,” he said, in the sense of always having more to learn. He joked, “If you ask me an answer, you’ve got an 89% chance that I’m going to give you the right answer.” It’s essential for using AI, Fischer told me during Fortune’s site visit, as he stood in front of a TV where a digital car slid off a course of cones until engineers tuned its calibration in simulation, something he said would otherwise take months with a physical car and a human driver. 

“When I use AI, it gives me the wrong answer most likely because I gave it the wrong engineering prompt,” he said, arguing that this B-plus approach is worth it because of the scale of the efficiencies. “AI actually enables engineers to have some time back to do what we should be paying them to do, and that’s innovate.”

courtesy of Jason Fischer

Few people have a job like Fischer’s, said Haq, who added that he swaps notes with him at least once a week. Fischer is “creating how we should work, the future of how we should work, and how to infuse AI into our entire process.” Haq does this for autonomy, but Fischer does it across all of GM engineering. 

Fischer’s education in failing fast started with GM’s self-driving acquisition, Cruise, where Fischer was assigned in 2017. Nearly 20 years into his time with the company, Fischer started spending two weeks of every month in San Francisco, embedded in a startup environment with a team of roughly 40 people that later grew to thousands. When senior GM product executive Doug Parks heard Fischer was breaking some rules, he called him into his office. “I don’t want you to break the rules anymore,” Parks told him, as Fischer described the conversation. “I want you to completely destroy every rule that we have at General Motors and I want you to document it.”

GM folded the work into its own teams in 2024, buying out Cruise’s outside investors. Haq said GM skipped robotaxis but kept autonomy engineers, along with a lot of the code and “a lot of the methodologies,” which are now going into the next-generation Super Cruise, GM’s driver-assistance system set to launch in 2028. The original robotaxi thesis assumed that “car sales would go down, rideshare would be dominant,” Haq said, and that simply “never played out.” He noted that rideshare still accounts for less than 1% of vehicle miles traveled.

Fischer, for his part, said he was innovation-pilled by his time at Cruise, citing the electrification of the GMC Hummer, which needed to come together in record time. GM is hoping to move toward a two-year vehicle development cycle, he said, with weekly “tall pole” reviews of whatever is holding up each job function. He said he was confident: “I’m convinced that we’re going to be able to do things just as fast as China.”