AI chip crunch: Startups vie for Nvidia’s vital component

SAN FRANCISCO: The artificial intelligence (AI) revolution is fully underway, but soaring demand for its most crucial component has startups scratching their heads on how they can deliver on AI’s promise.

Generative AI’s lifeblood is a book-sized semiconductor known as the graphics processing unit (GPU) – built by one company, Nvidia.

Nvidia chief executive officer and founder Jensen Huang made a wild bet years ago that the world would soon clamor for a powerful chip usually used for making video games, but that could build AI as well.

No company working with the generative AI models that fuel today’s frenzy can get off the ground without Nvidia’s singular product: the latest model is the H100 and its accompanying software.

That painful reality is one that Amazon, Intel, AMD and others are scrambling to fix with their own alternatives, but those attempts could take years.

And with the biggest tech companies throwing all their financial might into generative AI, the smaller fish must go on the hunt to secure Nvidia’s holy grail.

“Around the world, it is becoming very hard to get thousands of GPUs because all these big companies are putting in billions of dollars, stockpiling GPUs,” said Fangbo Tao, co-founder of Mindverse.AI, a Singapore-based startup.

“There’s not a lot of GPUs around,” he said.

Tao spoke to AFP at the TechCrunch Disrupt conference in San Francisco, where AI startups jostled to make their pitches to Silicon Valley’s venture capitalists (VC).

ChatGPT took the world by storm just as Silicon Valley was caught in a nasty hangover from the pandemic when investors threw money at startups, convinced that life had gone irreversibly online.

That turned out to be far-fetched, and the US tech scene entered a downturn with rounds of layoffs and VC money dried up.

Thanks to AI, some of the old mojo is back, and anyone with those two letters on their resume will likely see a red carpet rolled out on the legendary Sand Hill Road, home to Silicon Valley’s most storied investors.

But as the startups walk away with their VC cash, the money in their pockets will be quickly forked out to Nvidia for GPUs either directly or through providers to bring their AI dreams to execution.

“We call on a lot of the big cloud providers (Microsoft, AWS and Google), and they all tell us even they are having trouble getting supplies,” said Laurent Daudet, chief executive officer of AI startup LightOn.

The problem is most acute for companies involved in training generative AI models, which requires that power hungry GPUs work at peak capacity to process troves of data ingested from the Internet. — AFP