What happened
Chinese AI company DeepSeek is reportedly developing its own AI chips. The move is a direct response to escalating US export controls that restrict China's access to high-end processors from firms like Nvidia. According to a report from Ars Technica AI, the decision highlights a growing trend among non-US companies to secure their hardware supply chains by building their own silicon.
How the room's reading it
Infra watchers see this as an inevitable reaction to US sanctions. The prevailing view on developer forums is that while export controls are designed to slow progress, they also create a powerful incentive for domestic innovation. There's considerable debate about the timeline — designing competitive AI accelerators is one thing, but securing manufacturing at scale is an entirely different order of difficulty. The consensus among analysts is that this move, successful or not, points toward a long-term fracturing of the global AI supply chain, potentially creating distinct hardware and software ecosystems over the next decade.
Sailfish's take
We're not surprised by the announcement, but we're focused on the second-order effects for builders everywhere. A world with two parallel AI hardware stacks isn't necessarily a world with more choice or lower costs for those of us outside that new ecosystem. In the short term, it could mean less global competition for high-end GPUs, keeping prices stubbornly high. We think the practical takeaway for builders is to double down on software and model optimisation. Chasing the biggest chip is one strategy — building services that run efficiently on more accessible hardware is a much more durable one.