Kevin Gibbs wants to franchise
the dirt under your feet.

The third act of a quiet builder

Kevin Gibbs is the kind of engineer whose work you have touched today without knowing his name. If you typed two letters into a search box this morning and the third one appeared on its own, that little ghost belongs partly to him. If you ever deployed a Python app to App Engine and watched it absorb traffic like a sponge, that is his too. If you used Quip at a startup between roughly 2013 and 2020, you were inside a building he and Bret Taylor put up.

And yet, if you Googled him, you would have to scroll. He is not a stage founder. He is the type who keeps showing up where the infrastructure is missing and quietly pours concrete. Right now the missing infrastructure happens to be at the bottom of the supply chain for carbon removal, and the concrete he is pouring is charred cocoa shell.

His current company is Terraton. It is based in San Francisco, founded with Greg D'Alesandre in 2024, and as of August 2025 it is sitting on $11.5 million of seed money. The lead investors are Chris Sacca's Lowercarbon Capital and Gigascale Capital. The angel list reads like a reunion of his last two lives: Jeff Dean, the legendary Googler who once championed Google Suggest internally, and Bret Taylor, his Quip co-founder, now on the board of OpenAI.

The pitch is unembarrassed about its borrowed metaphor. Gibbs and his co-founder want to be the McDonald's of biochar. Not McDonald's the food. McDonald's the franchise system. The thing that takes a stove, a sign and a teenager, drops it into 40,000 locations, and somehow makes the fries taste the same in Lagos and in Lincoln, Nebraska.

Biochar, if you have not met it yet, is what happens when you bake plant matter in the absence of oxygen. You get a brittle, sooty material that looks like the leftovers of a campfire. Drop it into soil and it does two unfashionable things at once. It locks carbon away for hundreds of years. And it makes whatever grows in that soil grow a little better. The technology is old. The market is new. The bottleneck, Gibbs says, is not demand.

"When we talk to the big buyers like Microsoft, Google, Airbus - those sort of companies - they want to buy more, and they can't find more places to buy it from," he told TechCrunch in August 2025.

So the company is going where the agricultural waste already is. Terraton's first facility is with Three Mountains Cocoa in Ghana, where the husks of cocoa pods normally rot at the edge of the farm. The second is in Kenya, attached to a nut processor called EcoFix. Together those two sites are expected to remove around 20,000 metric tons of CO2 every year. That is a number that sounds large until you stand it next to one of the hyperscaler data centers being commissioned to feed an AI model. As Gibbs noted, "That's a lot, but that's not a lot if you've got an AI data center."

Stop one: the shuttle bus

To understand why a serial software founder is now running a hardware company in the soil business, it helps to rewind to a shuttle on the 101 in 2004.

Gibbs had graduated from Stanford with both a BS and an MS in Computer Science in 2002, both with distinction. He spent some time at IBM and then joined Google a couple of months before the company went public. He was assigned to work on Google's data center infrastructure, which was a generationally good place to be standing but also involved a long commute from San Francisco down to Mountain View.

The shuttle had patchy connectivity. He did not want to nap. He started writing small experiments. The first was a tool that would autocomplete URLs as you typed them. Cute. Useful for the engineer who had built it. Unremarkable. Then a colleague asked him the question that changes products forever: "That's cool. What if you did it for search?"

The answer turned out to be: Google Suggest. He pushed it as an internal demo. Jeff Dean and Rob Pike took an interest. Marissa Mayer renamed it, because his working title - "Google Complete" - was not going to fly. It launched on Google Labs in December 2004 and slowly, over four years, ate the internet. By 2008 it was the default behavior of Google's search box. It is, by now, one of the most-touched user interfaces ever built.

Gibbs himself describes losing interest in the project not long after it shipped. He had already moved on to a problem he found more interesting: the question of how anyone, anywhere, could deploy a web app to Google's infrastructure without first having to befriend a sysadmin. That project became Google App Engine. It launched in 2008 and is, depending on how you count, the first true platform-as-a-service offering from a major cloud provider.

Stop two: the doc that grew

In 2012, Gibbs left Google to start Quip with Bret Taylor, the former Facebook CTO and Google Maps co-creator. The product was a collaborative document and spreadsheet app that lived comfortably on a phone, which in 2012 was a less obvious idea than it sounds now. It was acquired by Salesforce in August 2016. The headline number was around $750 million. Gibbs stayed on, running the unit as CEO, and the team grew the business to north of $100 million in annual recurring revenue inside Salesforce. (Quip was finally end-of-lifed by Salesforce in early 2026, a long second life by enterprise software standards.)

Each chapter ended with the same instinct. Build the platform underneath. Move on when the work becomes legible. Start again at the lower, harder layer.

Stop three: the soil

Terraton is that instinct applied to climate. The hard part of biochar is not the chemistry. It is the box - the standardized, deployable, financeable, monitor-able package that turns a clever idea into a unit that an agricultural operator in West Africa can plug into the back of their facility and start running. Software, hardware, financing, market access, and verification all bundled together. That bundle is what Terraton sells. The agribusinesses operate the sites. Terraton handles the rest, and verified credits flow back out to corporate buyers who have run out of things to buy.

It is, in a real sense, App Engine all over again. The substrate is just a little dirtier.