Nvidia’s Earnings Are Strong, But The Road Ahead Looks Bumpy

So $30bn is now 'bad'....?

So, a few folks pinged me today asking about Nvidia’s latest earnings report, and let’s just say—it’s a tale of two headlines. On the surface, Nvidia’s numbers are eye-catching: $30 billion in revenue, more than doubling from a year ago[^1]. Impressive? Absolutely. But, as always, the devil is in the details. While the topline growth looks robust, the underlying performance may be giving us a sneak peek into a looming medium-term slowdown in demand.

For those who've been following my commentary over the last two years, you’ve heard me argue that the demand for Nvidia’s kind of chips—especially those powering the development of foundational AI models—is fundamentally transient. Sure, there’s still plenty of money to be made right now, but let’s not kid ourselves: the writing is on the wall. Even the notion of ‘foundational model’ is open for debate (even if you can define it).

Take a look around. Our own company has been running multi-billion parameter language models on nothing more than off-the-shelf laptops. This isn’t some hypothetical future—it’s happening now (and has been for quite a while!). The idea that you need sprawling data centers and massive cloud infrastructure to support most of today’s AI applications is increasingly a relic of the past. This shift isn’t just about the tech nerds like us tinkering in our labs. Even big players like Apple are moving towards on-device processing, a strategy we’ve championed since last year at Frontier Foundry. Our first LLM was called “Limni” - “Lake” in greek in case you didn’t know.

What does this all mean? It means that companies spending eight or nine figures on hardware may soon find themselves questioning those decisions. Software is eating the world, as they say, and nowhere is this truer than in AI. Companies will increasingly pivot from those heavy, capex-heavy hardware investments towards more agile and cost-effective software solutions, the kind that deliver almost the same power without the need for the massive physical footprint or energy consumption.

How many times do we need to learn this lesson? Software always beats hardware. We’re just watching it play out on a new stage.

==========

[^1]: Nvidia's revenue and earnings data sourced from their Q2 2024 financial report, showing a remarkable increase in data center revenue and a total earnings beat, but with slowing growth compared to previous quarters.

[^2]: [Investopedia Live Coverage](https://www.investopedia.com) detailing Nvidia's shift in market dynamics and investor concerns about long-term demand trends.

[^3]: Commentary on Apple's shift to on-device AI processing, a movement away from heavy reliance on centralized cloud infrastructures