AI is currently having a decentralizing rather than centralizing effect on the industry.

The claim is that, empirically as of 2025-08-03, AI is having a decentralizing rather than centralizing effect on the industry. True: market data at that time show widespread distribution of capability, e.g. many competing frontier and near-frontier labs, strong open-weight models (DeepSeek, Qwen, Llama, Mistral) close to the top closed models, falling concentration in model usage or market share, and broad international competition, without strong countervailing concentration. Mostly true: the evidence leans clearly toward decentralization with some concentrating trends. Mixed: the evidence is balanced, e.g. many competing models but heavy concentration in compute, chips (Nvidia) and capital among a few hyperscalers. Mostly false: concentration in compute, revenue and frontier capability clearly dominates, with only narrow decentralizing signs. False: the evidence clearly shows strong and growing centralization across the main measures.

Set before the research began.

1. E1 2025 Mid-Year LLM Market Update: Foundation Model Landscape + Economics | Menlo Ventures Menlo Ventures · 31 July 2025

   “The landscape has stayed active, with notable launches from DeepSeek (V3, R1), Bytedance Seed (Doubao), Minimax (Text 1), Alibaba (Qwen 3), Moonshot AI (Kimi K2), and Z AI (GLM 4.5) in the last six months.”

2. E2 2025 Mid-Year LLM Market Update: Foundation Model Landscape + Economics | Menlo Ventures Menlo Ventures · 31 July 2025

   “But despite these benefits and recent improvements, open-source has continued to trail frontier, closed-source models in performance by nine to 12 months.”

3. E3 2025 Mid-Year LLM Market Update: Foundation Model Landscape + Economics | Menlo Ventures Menlo Ventures · 31 July 2025

   “As a result, enterprise dollars are now consolidating around a few high-performing, closed-source models, giving us a new market leader in Anthropic*.”

4. E4 GPU Supply 2026: NVIDIA $57B Quarter, HBM Bottleneck, Hyperscaler $131B Capex - Lambda Finance Lambda Finance · 20 April 2026

   “NVIDIA captures roughly 44 cents of every hyperscaler capex dollar going into AI infrastructure — a concentration unprecedented in the history of enterprise compute.”

E1, a mid-2025 market review, records frequent model launches from many labs, mostly Chinese (DeepSeek, Qwen, Kimi K2, GLM 4.5 and others), which points to broad international competition. The same report also says open-source models still trailed closed frontier models by nine to 12 months (E2), and that enterprise spending was consolidating around a few closed-source models (E3), which points to concentration. E4 describes heavy Nvidia concentration in hyperscaler AI spending, but it is dated 2026, so it is only indirect evidence of conditions in August 2025. Taken together, the evidence shows real decentralizing signs alongside clear concentrating ones, which fits the criteria's description of 'mixed'.