SemiAnalysis: Six of Nvidia’s Seven Biggest Customers Are Developing Their Own Chips
SemiAnalysis posted on the X platform on November 14, 2026, that six of Nvidia’s seven biggest customers are developing their own chips. The organization listed the seven biggest customers as Google, Microsoft, Meta, Amazon, SpaceX, Anthropic and OpenAI, and said these customers account for the vast majority of deployed GPUs. Except for SpaceX, the other six are making their own chips.
SemiAnalysis further pointed out that if Nvidia did not support CoreWeave and other neoclouds, these industries would not exist, and the hyperscale cloud providers would be in complete control. The organization also said that Jensen Huang is the only executive in the semiconductor industry who can persuade companies in the supply chain to make unprecedented investments and allocate supply; based on its benchmarks, Google’s TPU performance is on par with Nvidia’s, but Google lacks the DNA to deal with the supply chain and customers.
The statement places customer in-house chip development alongside the ability to secure preferential supply-chain allocation. In-house chip development by the seven biggest customers will reduce long-term dependence on Nvidia GPUs, while the existence of neoclouds depends on Nvidia’s support, preventing hyperscalers from fully controlling the open market. After Huang steps down, whether his personal ability to persuade the supply chain to invest can be carried on by a successor will affect Nvidia’s ability to maintain priority capacity. At present, this post is an analytical judgment and does not attach public evidence of specific order losses or capacity shifts.
ABAB AI Insight
Historical structure: Major tech customers launching in-house chip development while still relying on external chips marks a shift from dependence on a single supplier toward a diversified supply structure. Similar cases show that when leading buyers also become designers, the long-term room for purchase growth is limited, and suppliers must seek non-self-designed customers to maintain share.
Capital path: Six of the seven biggest customers developing their own chips will redirect capital from Nvidia GPU purchases toward their own chip design and manufacturing partnerships (such as with Broadcom and TSMC). Neoclouds exist only because they depend on Nvidia’s support, so part of the open market’s compute supply depends on Nvidia’s proactive backing. Huang’s personal ability to persuade the supply chain to invest is a key path to maintaining preferential allocation.
Comparison: Unlike customers that rely entirely on external chips, self-developed customers can benchmark against Nvidia in performance and cost; unlike supply that relies only on contract lock-in, personal negotiation ability determines additional capacity and expansion willingness. TPU benchmark tests show the technology gap is narrowing, but the ability to work with the supply chain still creates differentiation.
Judgment: Pricing power is shifting from a model dominated by a single GPU supplier to one in which customer self-development and preferential supply-chain allocation operate in parallel. If self-developed customers expand internal use and reduce external purchasing, Nvidia’s share will come under pressure; if Huang-style supply-chain persuasion cannot be continued, priority capacity may be won by competitors. The structural change is that the self-development of top customers and personal relationships in the supply chain are both becoming long-term constraints on share.
ABAB News · Cognitive Law
Six of the seven biggest customers developing their own chips compresses long-term purchase space
Neocloud exists only because it depends on Nvidia’s support, and the open market
Personal supply-chain persuasion ability determines priority capacity allocation