The potential acquisition of the startup Hugging Face by Nvidia for approximately $12.9 billion (207 billion rupees) appears, at first glance, to be an unusual move. The target company, founded 10 years ago, has an annual revenue of about $150 million and was formally valued at $4.5 billion during a funding round in 2023. Such an almost threefold increase in valuation for a company that is only approaching profitability is not typical for a trillion-dollar chip manufacturer.
However, this deal, first reported by the technology publication The Information and which has not yet been signed or publicly confirmed by either company, makes sense if one views Nvidia not merely as a hardware business, but as a company fighting to maintain control over the entire artificial intelligence stack. The likely goal of this move is not just diversification, but a defensive strategy.
Nvidia enjoyed an almost monopoly over much of the current AI boom, with its market share in AI chips estimated at 70% or higher, thanks to its Graphics Processing Units (GPUs) and the Cuda software layer, which kept developers within its hardware ecosystem for nearly two decades. However, this position is under constant attack from the very companies that spend the most on Nvidia's silicon.
Among these attacking players are: Google, which is implementing its seventh-generation Ironwood tensor processors and has entered into an agreement with Anthropic to use up to a million of its TPUs in a multi-billion dollar deal by October 2025; Amazon, which has deployed over a million of its proprietary Trainium chips; OpenAI, which is developing its own silicon jointly with Broadcom and has separately formed an alliance with AMD; Microsoft, which possesses Maia accelerators; and Meta with its MTIA lineup.
This trend is evident: advanced model laboratories and hyperscalers, which are Nvidia's largest and most strategically important customers, are the same players who are creating their own chips to avoid dependence on Nvidia. It is expected that custom accelerators will account for nearly 28% of the market in 2026, growing significantly faster than commercial GPU sales. A 'chip moat' alone is no longer enough.