I Dug Into Michael Burry’s Nvidia Post So You Don’t Have To. Here’s Why I Think He’s Just Fishing.

When Michael Burry decided to short Nvidia (NVDA), he wasn't betting against the company's technology. It was a bet against the AI financing structure and macroeconomic conditions. We know today that he is losing money on this bet as Nvidia inches toward its all-time highs. But Burry isn't sitting idle with his short position. He recently took to social media platform X, posting about a startup that could challenge Nvidia's dominance.

Here's what Burry's post is about. A small startup is making a big claim against Nvidia. Founded by a group of Harvard dropouts, Etched is building a chip called Sohu. What the chip does is run transformer models, the architecture behind most of today's artificial intelligence. By designing the chip to do only that, Etched says it can run AI inference far faster and cheaper than Nvidia's general-purpose GPUs. It recently raised money at a $21 billion valuation. But to see why challengers like Etched are emerging, it helps to understand how AI actually runs.

Soros Fund Management Opened a New Position in Nebius During Q2. What This Means for NBIS Stock.

Nvidia Earnings, Jackson Hole and Other Key Things to Watch this Week

Ahead of Nvidia Earnings, Here's What Barchart Data Says Comes Next for NVDA Stock

Our exclusive Barchart Brief newsletter is your FREE midday guide to what's moving stocks, sectors, and investor sentiment - delivered right when you need the info most. Subscribe today!

AI runs in two phases. Training teaches a model, and inference is when the model answers queries. Inference is where the volume is heading as AI gets used more. It is also the area where specialized chips like Etched's — or custom silicon from Amazon (AMZN), Alphabet (GOOGL), and Broadcom (AVGO) — can compete the hardest. Nvidia dominates AI computing, but inference is an area where specialized chips can compete.

This is why the noise around challengers keeps growing. On Nvidia's last earnings call, CEO Jensen Huang pushed back firmly. He argued that building AI infrastructure is genuinely hard and that most custom-chip projects never reach real production. So far, Nvidia's dominance has held, with the company continuing to lead the AI chip market. The company will report earnings next on Aug. 26, and investors will want to know whether its inference business is still growing fast enough to stay ahead of rivals. Startups like Etched make the headlines. The earnings call is where investors will find out if they're a real threat or just noise.