30 March 2023 |

When your baby is a battery


On Tuesday, I had the chance to sit down with Andrew Hsieh, CEO and co-founder of Liminal, to discuss their approach to scaling battery manufacturing. 

Liminal’s role in battery manufacturing is, to start, a complementary one. They’re not in the business of building factories. They’re not in the business of designing new cathodes or separators. They don’t explore new battery chemistries.

However, their technology could have downstream impacts on all components of the battery manufacturing processes. Good time for it, too, as America attempts to birth a new battery manufacturing industry.

While birthrates in the U.S. are declining, battery manufacturing is booming. I promise I’m going somewhere with this. The technical thread that unites the twain is the ultrasound.

When your baby is a battery

I’d be surprised if ultrasounds call to mind anything other than babies for you. The same was true for me until two days ago. But ultrasounds aren’t just for babies. Liminal uses ultrasound technology to peer inside batteries on manufacturing lines on factory floors. Their ‘EchoStat’ platform offers an ultrasound inspection solution engineered specifically for EV-grade batteries.

Liminal’s EchoStat ultrasound inspection technology

This technology is designed to help battery manufacturers on two fronts. First, they offer manufacturers a major assist in quality control by identifying batteries that are built to spec, ones that may not quite be but can still pass overall controls, and ones that aren’t up to snuff at all. 

Quality control isn’t just a box to check for EV manufacturers. Some of the biggest consumer concerns with respect to EVs pertain to performance (e.g., range and whether the on-board battery intelligence system is accurately calibrated to said range) and safety. Lithium-ion batteries do, after all, occasionally catch fire catastrophically.

Second, in doing the quality control work over time, Liminal should also be able to help battery manufacturers zoom out to spot trends and then zoom back in to improve their processes and efficiency. If you do enough QC and capture the data, there’s a lot of pattern recognition to do. 

Here’s where I take a step back to look at the climate tech ‘data layer,’ i.e., the various forms of data available to firms that underpin their work.

Earth observation data, whether from drones or satellites, is a good example of one form of data that has gotten better, cheaper, and more accessible over the past decade. Many more active satellites orbit Earth as we speak (~7,000) than even a few years ago (+2,000 or so since 2021). And the quality and cost of satellite imagery they can provide has improved tremendously, too.

Satellite imagery is perhaps the most macro-level type of data that’s better and more available in 2023 than it was in 2013 or 2003. Other examples abound, from LiDAR to supply-chain level emissions accounting. 

At a more micro-level, however, you can also, well, use ultrasounds to examine the insides of batteries. As Andrew noted in our conversation, the opportunity here is the same as with other foundational types of data:

We’re opening up new data via ultrasound, identifying new signals.

We landed on this point by discussing all the AI hype all around us (which spans sectors and industries). As Andrew and I discussed this, we came back to the fact that one of the most powerful levers for any AI or ML application is what data is available to it to train, learn, and generate new insight. 

Liminal is fundamentally in the business of creating data that wasn’t previously available. That’s the v1 of their offering. Down the line, there will, of course, be lots of ways they and their partners might use that data. But for now, as battery manufacturing matures (like a newborn babe), data and insight on the health of battery cells and the entire battery system are valuable in and of themselves.

Nurturing the battery boom

Billions of dollars are flowing into onshoring battery supply chains in North America and Europe. As the transition to EVs and other battery-dependent energy systems accelerates, neither the public nor the private sector wants to be wholly dependent on China. 

Still, as I discussed with Andrew, little has changed over the past decade regarding the core pain points for battery manufacturers. While companies are trying to produce batteries at a greater scale and more quickly, and while battery chemistries shift, many of the same foundational quality control issues crop up in manufacturing.

Example include torn electrode tabs (one cause of GM/LG’s Chevy Bolt recall in 2021 and Ford’s one-month shut down for F-150 Lightning production), and folded or wrinkled separators (another cause of the GM/LG’s Bolt recall).

Further, Andrew rattled off a host of other challenges manufacturers face but discuss less publicly:

  • Metallic dust particles get trapped between electrode layers during cell assembly, which can cause short circuits
  • Improper saturation of the electrolyte into the electrodes can lead to many different performance problems
  • Misalignment of positive and negative electrode layers can also cause short circuits, especially over longer time horizons
  • Incomplete/improper “activation” of the battery cells during formation can impact performance reliability and safety

These are all issues that Liminal can help battery manufacturers identify and subsequently sort out. And while Liminal’s initial focus is on the factory floor, where they will work with partners to install inspection equipment to generate new primary data, in time they can expand their analytics capabilities beyond quality control. At first, this will look like building on-tool analytics. Longer-term, here’s how Andrew outlined the vision:

As you generate more primary data up and down the manufacturing ecosystem and feed it into the software layer that sits above the factory, you can build factory-level analytics for factory GMs and executives, offering a more holistic view of their operation.

And yes, if you or people you know are frothing at the mouth to bring the words AI back into this conversation, you could deploy all types of fancy AI solutions to glean additional insight from all this primary data. Ideally this will allow Liminal to help companies improve the overall efficiency and health of their manufacturing. 

Which is critical. There’s a new announcement seemingly weekly of both domestic and international firms committing to building out more battery manufacturing capacity in the U.S. in the wake of the IRA.

But these are announcements. There’s still a wide gulf between a headline and operational production capacity. And then a further gulf between production capacity and efficient operations and finished batteries that are fit to ship. And then a further gulf between all of the above and manufacturers actually making any money (even giant factories have thin profit margins, see below).

This graphic was sourced from Andrew Wang’s stellar newsletter, Intercalation Station

The second and third gulf, i.e., the ones between production capacity and production and said production and actually making money, are valleys Liminal hopes to help manufacturers cross. 

Of course, Liminal has its own work to do to move from announcements to deploying in factories. After a Series A2 fundraising announcement a few months back (a round in which Northvolt, the European battery manufacturing giant, participated) the team is marching towards their first factory-integrated deployment in September of this year (and hiring a lot to make that happen).

Whether and to what extent Andrew’s grand vision of harmonizing battery manufacturing operations will pan out depends largely on identifying the right partners for them to go to market with. I imagine fruitful partnerships will require a degree of flexibility, not just investment. Building a new battery manufacturing facility is one thing. Setting it up to also capture and ingest a whole new suite of data involves a whole ‘nother learning curve. 

The net-net

There are a few highlights I’ll now try to pull out here. 

The first is one we have covered already: Getting to look inside battery cells via ultrasounds is a new, high-leverage data point for battery manufacturers (and probably for hardware developers, too).

The second is another favorite of mine, namely that there is so much incredible technology in this world that we can repurpose for different use cases. It’s not always easy to see the light in that regard; by nature of the exercise, it requires thinking outside the box (or womb). 

Ultrasounds for batteries, not babies, is one example. Space gels repurposed for residential window insulation is another. If I were an entrepreneur, I’d be racking my brain, thinking in this vein, excited about potential financial gain. 

Thirdly and finally, I’d say that some of the most exciting climate tech stories still live in the enablement layer. As much as climate VCs on the bird app squabble and squawk about whether investing in software only disqualifies you from being a ‘real’ climate VC (whatever that means), and as much as deploying existing technologies like wind, solar, nuclear, geothermal and energy storage (including, again, batteries) is the lynchpin of the energy transition, invariably, there are wonderful businesses and technologies that accelerate all that deployment. 

To be sure, Liminal is very much a combined hardware-software play; this isn’t a case of getting bogged down in the software vs. hardware argument. But, even though I began this article by describing their ambition and business as complementary to manufacturers, presenting their role purely as such was probably be a misstep. 

My revised view (from ~1,300 words above) is that Liminal and other firms like it could be a significant accelerant for battery deployment. Which means more EVs on the road, fewer gas miles driven, and fewer carbon emissions in the atmosphere. All of which doesn’t really constitute a liminal role at all.