Nvidia Doubles Revenue as AI Chip Boom Accelerates

Nvidia reported yesterday, August 23, results that comfortably beat analyst forecasts, confirming that the generative AI frenzy is translating into real business, not just promises. The company posted $13.51 billion in revenue for its fiscal second quarter, which ended July 30 — up 101% from the same period a year earlier and up 88% from the previous quarter.

The Engine: Data Centers

The data-center division, which includes the GPUs that train and run models like GPT-4 or Llama, brought in $10.32 billion, up 171% year over year. It's the first time this unit has so clearly outpaced Nvidia's historic gaming-chip business, which contributed $2.49 billion, up 22% from a year ago.

Adjusted earnings per share came in at $2.70, well above the roughly $2.09 analysts had expected on consensus. But what really moved the market wasn't the quarter that just closed — it was the forecast. Nvidia projected revenue of about $16 billion for the current quarter, up 170% year over year and far above the $12.5 billion analysts had been modeling before the announcement.

"A New Computing Era"

Jensen Huang, Nvidia's CEO, summed up the moment in the company's earnings release: "A new computing era has begun. Companies worldwide are transitioning from general-purpose to accelerated computing and generative AI."

That's not just PR spin. Nvidia controls the vast majority of the market for GPUs used to train AI models, and its H100 chip has become the most coveted — and scarcest — component in tech in 2023. Companies including Microsoft, Google, Amazon, Meta and a legion of generative AI startups have been lining up for units, in many cases facing months-long waits.

An Already Extraordinary Year on Wall Street

These results come after Nvidia's stock had already climbed more than 200% so far in 2023, a rally that had priced in much of the enthusiasm around generative AI since ChatGPT's launch in late 2022. Precisely because of that, much of the market was already expecting a strong quarter: this time, the surprise wasn't so much that Nvidia beat expectations, but by how much — and above all, in a revenue forecast that suggests demand shows no signs of cooling.

Why It Matters Beyond Nvidia

These figures serve as a thermometer for the entire sector. If large language models and generative AI were just a passing fad, it wouldn't be translating into tens of billions of dollars in chip orders placed months in advance. Nvidia isn't selling a promise: it's selling the hardware without which training a large model — or serving its responses to millions of users — is simply impossible today.

That also exposes a bottleneck for the rest of the industry. The scarcity of advanced GPUs is driving up costs and slowing down the plans of any company that wants to train its own models, while reinforcing Nvidia's position as an almost irreplaceable link in the AI value chain — something rivals like AMD and Intel, and customers like Google and Amazon with their own custom chips, have long been trying to erode without notable success so far.

What remains to be seen is whether chip supply — also constrained by the manufacturing capacity of partners like TSMC — can catch up with demand that, judging by these numbers, keeps surging.