Cathie Wood is not being subtle about where she thinks the AI hardware market is heading. Her firm, Ark Invest, purchased 82,779 shares of Cerebras Systems, the AI chip company that builds processors the size of dinner plates.

This latest buy follows a 105,616-share purchase on May 14, bringing Ark’s total accumulation to roughly 188,000 shares in the span of about a week.

What Ark is actually buying

The purchases were executed primarily through two of Ark’s flagship funds: the ARK Innovation ETF (ARKK) and the ARK Next Generation Internet ETF (ARKW).

Cerebras is not your typical semiconductor company. While most AI chip firms build processors measured in square centimeters, Cerebras builds wafer-scale engine chips, meaning each processor occupies an entire silicon wafer. The company’s CS-3 AI supercomputer is designed for large-scale AI training and inference workloads, the kind of heavy computational lifting that powers large language models and complex machine-learning systems.

This is not just an addition to Ark’s portfolio. It is a substitution. Ark has been trimming its holdings in AMD while simultaneously ramping up its Cerebras position, signaling that Wood is betting that the next generation of AI infrastructure companies will outperform the incumbents.

Why Cerebras matters in the AI chip race

Cerebras occupies a different lane from Nvidia and AMD. Rather than competing on the GPU battlefield, the company has wagered that wafer-scale computing, where a single chip can handle workloads that would normally require clusters of GPUs, represents a fundamentally better architecture for certain AI tasks.

The CS-3 supercomputer, Cerebras’ flagship product, is purpose-built for large-scale AI training and inference workloads. Cerebras has positioned itself as an alternative to Nvidia’s GPU offerings, competing alongside emerging alternatives such as Google’s TPU.

What this means for investors

The rotation out of AMD and into Cerebras suggests Wood believes the AI hardware market is entering a phase where architectural innovation matters more than incremental GPU improvements.

For crypto-adjacent investors, the Cerebras bet carries additional relevance. The company’s hardware has potential applications in decentralized computing and on-chain AI, areas identified in Ark’s broader commitment to AI infrastructure relevant to on-chain AI and decentralized computing.

Ark’s aggressive accumulation of roughly 188,000 shares in under two weeks also introduces a practical consideration: when a high-profile fund manager buys this aggressively into a recently listed stock, it can create short-term price dynamics that may not reflect the company’s underlying fundamentals.