SemiAnalysis Slashes 2028 AI Accelerator Spend Forecast by 28%, Nvidia Takes Biggest Hit
SemiAnalysis has cut its 2028 global AI accelerator spending forecast from $1.054 trillion to $758 billion, a 28% reduction. The revision also slashes Nvidia’s projected 2030 compute product revenue by 66%, while Broadcom’s cumulative custom chip revenue falls only 11%.
Industry Analysis and Implications
The downgrade signals a significant recalibration of expectations for AI infrastructure growth, with profound implications for the entire tech ecosystem, including crypto and decentralized compute networks.
Nvidia’s outsized exposure to the forecast cut reflects its dominance in general-purpose GPUs, which have been the primary beneficiary of the AI boom. However, the sharp reduction suggests analysts now expect a more moderated pace of data center buildout, potentially due to power constraints, shifting model architectures, or slower enterprise adoption.
Broadcom’s smaller hit highlights the growing importance of custom ASICs (application-specific integrated circuits) for hyperscalers. As major cloud providers like Google, Amazon, and Meta design their own chips to reduce costs and optimize for specific workloads, Broadcom stands to benefit from this trend. The relative resilience of custom silicon also implies that AI compute demand is becoming more diversified and specialized.
For the crypto sector, this revision could have mixed effects. Decentralized compute networks that aggregate GPU resources may face lower demand if overall AI spending slows. However, the shift toward specialized hardware could create opportunities for networks that support a variety of chip architectures. Additionally, if Nvidia’s revenue growth decelerates, it might reduce the pricing power of GPU providers, potentially making decentralized compute more competitive.
Forward-Looking Perspective
Investors should watch for further forecast revisions from other research firms and monitor capex guidance from hyperscalers. The AI infrastructure buildout remains substantial, but the era of exponential growth may be giving way to a more sustainable, albeit slower, expansion. Crypto projects in the AI space must adapt to this new reality by focusing on efficiency and real utility rather than speculative excess.

