MIT technology fails spectacularly for grid-scale batteries

Our house is close to the headquarters of NextEra, which has a market cap of around $166 billion and is the parent company of Florida Power & Light. We ran into one of their employees at a restaurant whose job is buying grid-scale batteries to sit next to the company’s grid-scale solar farms. “This will power your data centers,” she explained. (I loved hearing this because it gave me an opening to point out that “No data center is illegal on stolen land.”)

I asked her if the MIT geniuses who revolutionized the grid-scale battery world about 20 years ago, according to the MIT geniuses, had an impact on her job. “No,” she explained. “Those technologies are always 10 years in the future. Utilities buy LFP and NMC batteries. The Chinese are unbeatable.” The same batteries that go into electric cars, such as those made by Head Nazi Elon Musk?!?! Yes, it turns out.

What’s an example of the liquid battery hype? Here’s one from Barack Obama’s Department of Energy, March 2016… “New Chemistries Found for Liquid Batteries: Grid-scale approach to rechargeable power storage gets new arsenal of possible materials”:

Liquid metal batteries, invented by MIT professor Donald Sadoway and his students a decade ago, are a promising candidate for making renewable energy more practical. The batteries, which can store large amounts of energy and thus even out the ups and downs of power production and power use, are in the process of being commercialized by a Cambridge-based startup company, Ambri.

Now, Sadoway and his team have found yet another set of chemical constituents that could make the technology even more practical and affordable, and open up a whole family of potential variations that could make use of local resources.

“This paper brings together innovative engineering advances in cell design and component materials within a strategic framework of ‘cost-based discovery’ that is amenable to the massive scale-up required of grid-scale applications,” says Richard Alkire, a professor of Chemical and Biomolecular Engineering at the University of Illinois, who was not involved in this research.

What does ChatGPT have to say? In a time-honored MIT pattern, the nerds achieved their goal of creating a better widget than whatever was on the market at the time they started talking about how much better their innovation was. And, in time-honored fashion, the comparative dullards improved whatever was on the market by a factor of 10X by the time that the MIT tech finally limped its way into a commercialized version. Excerpts from the full Q&A:

The research actually began around 2005, so “20 years ago” is about right, although that particular DOE release is from March 2016. MIT spun the technology out in 2010 as Ambri, backed by Bill Gates, Khosla Ventures, Total and eventually Reliance.

(Is it possible that Bill Gates will get ahead of divorce plaintiff MacKenzie Scott Bezos in terms of spending tens of $billions on nonprofit activities without leaving any trace of positive impact on humanity? Compare to Stephen Ross. For a tiny fraction of what Bill Gates has spent on becoming a Davos Idol Do-Gooder (over $110 billion in cumulative distributions; maybe that’s $200 billion in post-Biden dollars?), Ross has completely transformed the city of West Palm Beach (nytimes). Ross perhaps will end up having donated about $1 billion, but most of the money he’s “spent” is really an investment that should yield a decent return unless the Climate Doomers are correct about South Florida being washed away. ChatGPT on whether Gates can compete with MacKenzie the divorcée for incinerating cash in an ash-free manner: “$111 billion already spent is an astonishing amount of money, and there isn’t an obvious list of, say, ten Gates-created breakthroughs commensurate with that sum.” The giant artificial brain struggles mightily to find anything the Gates Foundation has done that might have a lasting impact.)

The basic idea was elegant: instead of solid electrodes that crack, swell and chemically deteriorate, use three molten layers that naturally separate by density—a light metal electrode, molten-salt electrolyte, and heavy metal electrode. The commercial version ultimately used calcium and antimony, with a calcium-chloride electrolyte. Because everything remains liquid while operating, the electrochemically active surfaces effectively recreate themselves each cycle. Ambri claimed >20-year life with very little degradation and no lithium-ion-style thermal-runaway fire problem.

First, going from a battery that works to a battery that can be manufactured cheaply by the million is extraordinarily difficult. The high-temperature seal is a perfect illustration. In the lab, having molten calcium/salt/antimony sitting at hundreds of degrees is manageable. For a utility battery expected to operate unattended for 20 years, every seal, feedthrough, weld, heater, electrical connection and insulation system has to survive thousands of thermal and electrical cycles. Ambri eventually solved enough of these problems to certify a ~1-kWh cell, but it consumed years.

Second—and probably even more important—the economic target moved spectacularly while Ambri was developing its technology. In 2010, when Ambri was founded, lithium-ion battery packs cost more than $1,200/kWh. By 2021 they were about $132/kWh; by 2024 about $115/kWh, and in 2025 about $108/kWh on BloombergNEF’s global average measure.

That’s a roughly 91% cost collapse during Ambri’s life.

Was it overhyped?

But there is a lesson in the old press coverage. MIT’s 2015 article said Ambri had built its manufacturing plant and made production sound almost trivial—“just add the electrode metals plus the electrolyte salt to a steel container and heat the can.” It then said the company was dealing with “one final engineering challenge”, the seal.

Nine years after that article, Ambri was bankrupt without a mass-production factory.

The coronapanic enthusiasts at IEEE sung Ambri’s praises in 2023, about 8 months before the company filed for Chapter 11 bankruptcy protection:

Ambri’s grid battery costs $180/kWh to $250/kWh depending on size and duration, the company says. But its projected cost is about $21/kWh by 2030