Would You Buy Generic AI?
Автор сравнивает текущий момент в индустрии ИИ с появлением дженериков в фарме: китайская модель DeepSeek V3 сопоставима с GPT-5.2 по большинству бенчмарков, но стоит на 90% дешевле. OpenAI и Anthropic заработали в 2025 году $22 млрд, тогда как китайские ИИ-лаборатории — $1,8 млрд, разрыв 12:1 объясняется ценой, а не объёмом использования. Падение цен в Китае разгоняют три силы: дистилляция знаний из Claude и других моделей, субсидии гиперскейлеров (Alibaba Cloud снизила цены на LLM до 97%, Baidu, ByteDance и Tencent потратили $1,1 млрд за один китайский Новый год 2026) и ценовая планка DeepSeek, обучившей V3 за $6 млн против $100+ млн у GPT-4 и достигшей $220 млн ARR со штатом 122 человека. В отличие от фармы с её 20-летней патентной защитой, в ИИ окно для дженериков открывается за недели. Главный вопрос — как защитить актив стоимостью в сотни миллионов, если его можно скопировать за месяц.
Kirkland ibuprofen is the same molecule as Advil. Same dosage, same FDA requirements, same therapeutic effect.1 It costs 80% less.
AI has its generic drug moment. DeepSeek V3 matches GPT-5.2 on most benchmarks.2 It costs 90% less. OpenAI & Anthropic generated $22 billion in 2025.3 Chinese AI labs generated $1.8 billion.4 The ratio : 12:1.
Pricing explains the gap. Chinese AI API prices collapsed 90% in 2024.5 US frontier models average $3.38 per million input tokens. Chinese models average $0.48.
OpenAI processes roughly 8.6 trillion tokens per day.6 Chinese labs likely match or exceed this volume. The 12:1 revenue gap isn’t usage. It’s price.
Three forces drive Chinese prices down.
First, distillation commoditizes capability. Anthropic accused DeepSeek, Minimax & Moonshot AI of conducting “industrial-scale campaigns” to extract knowledge from Claude.7 OpenAI made similar accusations to Congress.8
Second, hyperscalers subsidize AI to win cloud customers. Alibaba Cloud cut LLM pricing by up to 97%.9 Baidu, ByteDance & Tencent spent $1.1B on AI subsidies during Chinese New Year 2026 alone.10
Third, DeepSeek set the floor. They trained V3 for $6 million versus OpenAI’s $100 million+ for GPT-4,11 price at $0.14 per million input tokens & hit $220 million ARR with 122 employees.
In the US, Chinese models also price at a discount. Together AI charges $1.25 per million input tokens for DeepSeek V3.12 DeepInfra offers $0.21 per million.13 DeepSeek’s own API charges $0.14 - 12x less than GPT-5.2.14
Pharma companies spend billions developing a molecule, then enjoy 20 years of patent protection to recoup R&D costs before generics flood the market. AI follows the same pattern - massive R&D costs upfront, then commoditization. But the timeline is compressed.
In pharma, the generic window opens after two decades. In AI, it opens in weeks. DeepSeek V3 costs $0.14 per million tokens. GPT-5.2 costs $1.75. Same capability. Different label. The 90% discount isn’t coming. It’s here.
The question : how to protect an asset that takes hundreds of millions to develop when it can be copied in a month?
Sources
OpenAI Revenue - The Information, Anthropic Revenue - Bloomberg ↩︎
Chinese AI lab revenues (trailing 12 months as of mid-2025) : 4Paradigm $834M - KrASIA, SenseTime $610M - Yahoo Finance, DeepSeek $220M - Business of Apps, [Minimax $79M, Zhipu $70M - industry estimates] ↩︎