The big business of spare capacity

In the 1830s, India had plenty of things that New Englanders wanted: luxuries like indigo, silk, spices, saltpeter, and cotton textiles.

But the feeling was unrequited. India had no need for New England’s primary exports of dried fish, beaver furs, and whale oil.

As a result, the many ships carrying Indian goods to Boston typically made the return trip to Calcutta entirely empty. Captains had to fill their holds with rocks just to keep their ships from capsizing in the rough seas on the way back.

New England did, however, have an endless supply of one sought-after commodity that India lacked entirely: ice.

Ice cut from the frozen ponds around Boston — including the one Henry David Thoreau made famous — became one of New England’s top exports in the early 19th century. The breakthrough came when Frederic Tudor developed a packing method that kept large blocks of ice from melting too much en route to tropical climates like Charleston, New Orleans, and the Caribbean.

The journey to Calcutta was much longer, of course. But packed tightly and insulated with layers of sawdust, Tudor hoped that most of his ice would survive the trip.

It was worth a try. Even if half of his product melted away, Tudor estimated he’d still be able to make a profit thanks to his negligible shipping costs: ship captains would carry ice to India for almost nothing because it saved them the trouble of filling their holds with rocks.

Tudor’s ice, they reasoned, would serve as ballast.

It worked! The ships stayed afloat, less than half of the ice melted en route, and the remaining ice was sold at a hefty markup to British officials desperate to chill their gin and tonics.

But even with ice selling for six times what it sold for in New England, the venture was only profitable because transportation was essentially free.

This was perhaps the first example of a business built around existing, unused capacity.

There haven’t been many since.

Airbnb and Uber are the most famous ones, but there are so few others that there’s no established term for the business model of leveraging the economy’s idle resources.

It might be described as excess-capacity utilization, capacity arbitrage, or the sharing economy.

Or we could just call it DePIN.

In real time (as the chatbots like to say)

Using magic internet money to bootstrap real-world infrastructure has always been crypto’s most original and intriguing idea, in my opinion.

But it’s never really caught on. Helium and Hivemapper, for example, were big, hopeful ideas that have struggled to get much traction thus far.

The handful of DePINs offering access to spare GPU capacity have made little impact despite an acute shortage of GPUs.

Geodnet is a rare DePIN success story, but it serves a niche market. 

In total, the results have been meagre enough that the term “DePIN” was starting to feel about as relevant as “NFTs,” “play-to-earn,” or “Web3.”

Recently, though, it’s started to get some buzz again — people increasingly think the Grass protocol could be the DePIN champion we’ve long been waiting for.

Grass — the protocol that turns your unused internet bandwidth into AI infrastructure — might already be DePIN's best proof-of-concept. The Grass network recorded $17 million in revenue in 2025 and is growing fast: it generated another $17 million in just the first half of 2026. 

Pretty good!

And people think it might be about to get much, much better. 

So far, Grass has made its money by routing AI labs’ web-scraping requests through residential IP addresses. It’s a good business because 1) websites often block the labs from scraping the data directly and 2) Grass gets to use our spare internet bandwidth for free (more or less).

Now, however, Grass has leveraged the data it’s gleaned via millions of residential internet connections (mine included) to develop a new service: Live Context Retrieval.

In addition to scraping data for training runs, Grass will allow chatbots like Claude and Grok to search the internet for up-to-date information that’s not in their training data, whenever their users ask for it.

This can get expensive. When you ask Claude, say, what restaurants are open right now, it can’t just go to Google.com like we do, because Google is just for humans.

Instead, LLMs have to pay for API access to a search engine built expressly for machines — a pay-per-view version of Google, so to speak. 

This should be a much bigger opportunity for Grass. “We believe the eventual market for inference-time search and retrieval could be several orders of magnitude larger than the market for pretraining web data,” Multicoin Capital wrote in a recent note announcing their investment in the GRASS token.

For a sense of how much bigger, consider the cost of accessing the Web via API: Anthropic and OpenAI pay $5 for every 1,000 web searches they make.

You can see how that could get expensive for them. ChatGPT alone receives billions of requests per day, and it’s estimated that something between 14% and 35% of these require a search of the Web for live information.  

Multicoin expects that for Grass, capturing “even a fraction of that request volume could support billions of dollars in annual revenue.”

Billions!

Billions would make Grass crypto’s biggest real-world success story by far.

If so, it will be a DePIN story of harnessing our spare capacity to surf the Web.

Just like sending ice to India.

— Byron Gilliam