The cathode bottleneck: Why battery costs still hinge on cobalt and nickel chemistry

Roughly 75% of what you pay for a lithium-ion battery goes to materials. And the single most expensive component inside that materials bill is the cathode, the positively charged electrode that accounts for about half the total cost. It is a problem measured in billions: the lithium-ion battery market hit an estimated $60 billion in 2024 and is projected to triple within a decade as electric vehicle and grid storage demand accelerates.

Arumugam Manthiram has been studying the chemistry inside these cathodes since 1986, when he joined the Cockrell School of Engineering at the University of Texas at Austin. He once worked alongside Nobel laureate John Goodenough, the researcher credited with developing the cathode materials that made modern lithium-ion batteries possible. Now Manthiram leads his own group, and their latest contribution -- a framework published in Nature Energy -- takes aim at a deceptively simple question: how do we make oxide cathodes more efficient?

Three variables, one cathode

The framework identifies three factors that control how an oxide cathode performs. The first is electronic configuration -- the way electrons are arranged in the atoms of the cathode material, which determines what elements can be grouped together effectively. The second is chemical bonding, which influences operating voltage, thermal stability, and safety. The third is chemical reactivity, which affects gas generation during cycling and the long-term stability that determines how many charge-discharge cycles a battery can survive.

Each factor interacts with the others in ways that are not always intuitive. Even iron, one of the most stable elements on the periodic table, can cause problems when paired with lithium in an oxide cathode. Getting the balance right across all three dimensions requires enormous amounts of experimental data -- more than any single research group can generate through bench experiments alone.

Where machine learning fits -- and where it does not

The broader materials science community has already begun training machine learning algorithms to accelerate discovery. Google DeepMind's GNoME project, for instance, predicted 528 new compounds that could potentially serve as lithium-ion conductors. But there is ongoing debate about how novel or practical those predictions actually are, which is precisely Manthiram's point.

"You cannot depend only on machine learning or artificial intelligence," he said. "You also need human intervention. That means whatever comes out of that research, we better understand what it is."

Manthiram's group uses the Texas Materials Institute's facilities to run characterization experiments, generating complex datasets that their ML models then parse. The models make predictions, which the group tests experimentally, feeding results back into the next training cycle. It is iterative by design: data trains model, model suggests material, material gets tested, test data trains the next model.

The supply chain pressure

This work is not purely academic. Most lithium-ion cathodes today rely on nickel, cobalt, and lithium -- minerals whose supply chains are vulnerable to geopolitical disruption, conflict, and environmental regulation. Cobalt in particular has drawn scrutiny for both its cost and the conditions under which it is mined. Reducing the amount of cobalt in cathodes while increasing nickel content is a well-known goal, but higher nickel concentrations bring their own instability problems.

Alternative battery chemistries using more abundant materials like sulfur or sodium are under development at UT Austin and elsewhere, but Manthiram is candid about their readiness. "It's one thing to do something in the lab, and it's another thing to make it, put it in your hand, and use it," he said. These alternatives remain at the prototype stage. Lithium-ion is not going away soon, even if it is eventually supplemented.

A framework, not a finished product

What the Nature Energy paper offers is not a new battery or a new cathode material. It is a conceptual map -- a way of organizing what we know about oxide cathode behavior so that future experiments, whether run by humans or guided by algorithms, are better targeted. Manthiram hopes the framework serves an educational function as well, helping researchers new to the field understand the fundamental chemistry and physics that govern cathode performance before they start feeding data into models.

The practical payoff, if it comes, would be cathodes that are cheaper to produce, safer in operation, and less dependent on conflict-affected mineral supply chains. But that payoff depends on the interplay between academic discovery and industrial scale-up -- a gap Manthiram has watched for four decades.

"We invent the materials; we invent the process in academic labs; and then industry has to scale up and implement it," he said. "I tell my students, we're all learning. That's the attitude I have."

The work was published in Nature Energy. Manthiram is a professor in the Walker Department of Mechanical Engineering at UT Austin's Cockrell School of Engineering.