Breaking the NVIDIA monopoly: Tenstorrent matters is the open alternative we need

Hello everyone,

We talk a lot here about the dangers of closed ecosystems. Right now, the AI hardware market is dominated by the NVIDIA CUDA monopoly, which overprices hardware and is simply more focused on renting clouds than supplying hardware to people’s homes. It is a massive barrier to entry.

Cost is the biggest hurdle. It currently takes between $3,000 and $5,000 to get a single GPU capable of serious AI processing. Most of us are stuck scavenging for used or degraded consumer cards just to keep our projects running. Even when you have the hardware, you are often forced into proprietary bottlenecks like NVLink, which makes scaling up exponentially expensive.

If we want an AI future based on DeusOp principles like individual autonomy and transparency, we need an alternative.

That is why I think Tenstorrent is worth watching.

They are building a different kind of future by focusing on RISC-V and open source software. Instead of locking developers into a proprietary ecosystem, they are designing hardware that is transparent and modifiable. Crucially, they solve the scaling problem by integrating networking directly into their chips. This makes it much easier to connect multiple processors together without depending on the expensive, closed-off interconnects we see elsewhere.

It is a direct challenge to the centralized overreach we see in tech today, and it aligns perfectly with our goal of keeping technology open to serve human potential.

If you want to help build this open infrastructure, they have engineering and software roles open worldwide, including in Santa Clara, Austin, Toronto, Tokyo, and Belgrade.

Beyond professional roles, if you want to get involved in community-driven work, you can also support our own development efforts here: