How Crypto Is Helping AI Solve Its Biggest Bottleneck
Those who control the chips, control the future…
By: Zack Abrams
June 2, 2023
Connect your wallet if you have a Coinage membership pass. Or, mint a Subscriber pass for 0.0175 ETH to unlock all our stories.MINT SUBSCRIBER MEMBERSHIP PASS
This year, ChatGPT triggered an “iPhone moment” as we all woke up to how artificial intelligence might change life as we know it. Suddenly, for free, users could type into a chatbox and unlock a world of possibilities.
But behind that chatbox is a lot of computing power; analysts estimate that ChatGPT costs on the order of $700,000 per day to run. That’s on top of the tens of millions of dollars OpenAI spent training their models in the first place.
The advanced chips that power the likes of ChatGPT and its ilk are made by just a few companies, including AMD, Intel and, most notably, NVIDIA, which joined Apple, Microsoft, Google, and Amazon as the latest company to crack a trillion-dollar market cap this past week.
But the companies cranking out chips to power the AI arms race can barely keep up with demand, meaning smaller startups are having to beg cloud companies for more server space. Elon Musk even told the Wall Street Journal that GPUs, the type of chips that run everything from AI to crypto mining to video games, are harder to get in San Francisco than drugs (though he joked that’s not a high bar).
As more companies pivot to AI and Apple prepares to launch its flagship mixed-reality headset, it’s clear that the struggle for computing power will define this decade. But where will the chips come from?
Better question: why don’t we just get everyone to pitch in the chips they have currently lying around?
That’s the idea behind decentralized cloud computing platforms like Render Network and Akash Network. Just as Airbnb brought spare bedrooms to the real estate market, decentralized compute projects want to incentivize users to run their graphics cards full time — not mining cryptocurrency, but doing useful work like training AI models and rendering advanced graphics — and paying them for their contribution.
“The current cloud is unable to provide the supply and they have to restrict and choose who they serve because that's the only way they can function as a business,” said Greg Osuri, the founder of Akash.
That’s an important point. When Elon Musk needed more computing power for his AI startup, he called up Oracle and they were happy to oblige, much to the consternation of smaller founders who were clamoring for the server space.
If developments in AI are going to define the future, therefore, it’s important that everyone has access to those systems, not just today’s tech billionaires. And with just three companies, Amazon Web Services (AWS), Microsoft Azure, and Google Cloud, controlling two-thirds of the cloud computing market, cloud computing provides a ripe opportunity for decentralization.
As of today, these crypto efforts are small: Render Network’s market cap currently hovers around a billion dollars, while AWS did $20 billion dollars of revenue just last quarter. But the crypto proposition makes intuitive sense: crypto tokens allow for fast and easy payments to users who contribute computing power all over the world.
“At a certain point scaling GPU rendering power is too difficult…it’s impossible to leverage all the GPUs in the world efficiently without some sort of decentralized system that allows that to be tracked and managed,” said Jules Urbach, CEO and Cofounder of Render, in a video introducing the project.
Both Akash’s token and Render’s token, which is used as payment on their respective networks, have seen incredible rallies in 2023. AKT is up nearly 225%, while RNDR is up more than 500%.
But is it all just hype? This week on Coinage: we’re getting cloudy with a chance of crypto.
For all that and more, check out our video above. To support our community-owned outlet, own it with us, and unlock exclusive benefits, mint one of our NFTs today!