Elon Musk Attempts to Challenge NVIDIA in AI Datacenters: Megapod Trademark Filed

. Tesla has filed a trademark application for "Megapod" with the United States Patent and Trademark Office (USPTO), revealing its intention to market modular and pre-configured hardware for datacenters dedicated to artificial intelligence. The filing, marked with serial number 99893717, describes autonomous and integrated computing systems to handle complex workloads. This move comes less than a year after the definitive cancellation of Dojo, the training supercomputer on which the company had initially bet for its proprietary infrastructure.

The documentation describes an unusually detailed integrated solution for a trademark application. The Megapod architecture covers modular hardware systems comprising computing servers, networking equipment, power distribution units, and cooling systems, also integrating the necessary software to monitor and optimize the entire structure. Thus, it is not merely a collection of single chips, but a complete architecture, packaged in a ready-to-use casing for the training and inference of generative models.

A market dominated by NVIDIA and proprietary silicon nodes Tesla's entry into this segment faces fierce and established competition. The current benchmark for liquid-cooled computing modules is NVIDIA's GB200 NVL72 system, which combines 72 Blackwell GPUs and 36 Grace CPUs to act as a single gigantic GPU, while similar solutions based on the same platform are already being produced by Dell and Supermicro. There is also a naming overlap issue: Submer, a company specializing in immersion cooling, already markets a pre-assembled cabinet called "MegaPod" and holds the registered trademark in a related class.

There remain further doubts about the actual feasibility of the project in the semiconductor market. Tesla does not have a sales division for computing servers and its Cortex cluster, operational at Giga Texas, relies on about 67,000 equivalent Nvidia H100 GPUs. The development of proprietary silicon has faced significant setbacks: the Dojo supercomputer was shelved in August 2025 after much of the engineering team left, the AI5 chip has seen nearly two years of delays, and the successor AI6 has been pushed to late 2027 due to difficulties encountered by Samsung on the 2-nanometer production node.

The real strength of the company in the artificial intelligence sector lies in energy supply rather than silicon. Megapack storage systems and the new Megablock are already being marketed as grid stabilizers for AI datacenters; Musk's own xAI has purchased Megapack worth about a billion dollars. The launch of Megapod could therefore translate into an infrastructure package focused on the thermal and electrical management of facilities, an attempt to reposition the company in the eyes of investors after the stock has lost over 20% since the beginning of the year due to margin contraction on electric vehicles.