谷歌算力告急，限制Meta对其人工智能大模型Gemini的使用-36氪

‘Tokenmaxxing’ is dead: How the hardware shortage is blowing up AI budgets

Google capped Gemini access due to compute shortage, ending the "tokenmaxxing" era—Meta burned $900M in 30 days, Uber its annual budget in four months. Enterprises must now link AI spend to ROI; HBM/GPU shortage keeps token prices rising while only 8% measure AI value.