AI Industry's Revenue, Expenditure, and Earnings - A Cyber Dialogue
Listen
问这期播客
会先在本集摘要、章节、转录和笔记里找答案。
TL;DR · AI Summary
The current hot topics in the AI industry include revenue models for AI companies, advertising as a business model, and operator token packages.
Key Takeaways
- AI companies charge through subscription models, which are difficult to sustain
- Advertising can be a business model for AI, but it needs to consider ROI.
- Operators are launching token packages to respond to the development of AI techn
Outline
Jump quickly between sections.
Introduce the main content of this episode, discussing revenue, expenditure, and earnings in the AI industry.
Analyze the charging models of AI companies, including subscription models and advertising business models.
Explore operator token packages and their underlying reasons.
Compare the differences in AI business models between the US and China.
Mindmap
See how the topics connect at a glance.
查看大纲文本（无障碍 / 无 JS 友好）
- AI行业的收钱、花钱与赚钱
- 收钱
- 订阅模式收费
- 广告商业模式
- 花钱
- 运营商 token 套餐
- 赚钱
- 中美商业模式差异
Highlights
Key sentences worth saving and sharing.
Tencent Pony believes that AI companies cannot sustain free services indefinitely because of high training costs.
Advertising can be a business model for AI, but it needs to consider ROI.
There is a difference between the US and China in AI business models, with the US being more pragmatic and China focusing on meaningful yet seemingly useless things.
Chapters
大家都在谈钱
大家都在谈钱
钱的三个角度：收钱、花钱、赚钱
钱的三个角度：收钱、花钱、赚钱
对于豆包收费的感觉
对于豆包收费的感觉
腾讯Pony为什么说“船是漏的”？
腾讯Pony为什么说“船是漏的”？
长期看为什么模型公司不能持续免费提供服务？
长期看为什么模型公司不能持续免费提供服务？
如果你是豆包的负责人，你看着即梦的收入不眼红吗？
如果你是豆包的负责人，你看着即梦的收入不眼红吗？
AI订阅的模式类似健身卡：我赌你不来
AI订阅的模式类似健身卡：我赌你不来
从上市公司财报看，推理的毛利还可以，亏主要是因为训练
从上市公司财报看，推理的毛利还可以，亏主要是因为训练
豆包之后，下一个想尝试收费的是？
豆包之后，下一个想尝试收费的是？
AI产品订阅定价体系的拆分
AI产品订阅定价体系的拆分
大家期待豆包的收费版本到底提供什么东西
大家期待豆包的收费版本到底提供什么东西
未来看，AI也会带来“鸿沟”
未来看，AI也会带来“鸿沟”
Transcript
大家都在谈钱
钱的三个角度收钱、花钱、赚钱
对于豆包收费的感觉
腾讯Pony为什么说“船是漏的”？
长期看为什么模型公司不能持续免费提供服务？
如果你是豆包的负责人，你看着即梦的收入不眼红吗？
AI订阅的模式类似健身卡我赌你不来
从上市公司财报看，推理的毛利还可以，亏主要是因为训练
豆包之后，下一个想尝试收费的是？
AI产品订阅定价体系的拆分
大家期待豆包的收费版本到底提供什么东西
未来看，AI也会带来“鸿沟”
广告呢？是否可以成为AI的商业模式？
运营商为什么最近突然开始推token套餐？
马耳他全民免费token的模式是否会被效仿？
重轻说AI是打篮球，你不能通过看100个视频学会打篮球，你得去拍
最近各种科技大佬被大学生嘘
某种技术宿命论技术是挡不住的
“我们不知不觉的走到了很多事情的人类的极限里”
我们都要去探索去太空建设数据中心了
中国的大模型厂商分两派大厂和所谓的几小龙
海外就是御三家了，国内为什么还有这么多家？
视频模型的案例seedance这么强的前提下，市场的第5678名为什么还能持续拿大钱？
辩证看cursor的价值
订阅之外的API
老黄提出的五层蛋糕的说法
其他层面是供给逻辑，但应用不是
“人类不争气”啊，服务碳基的都不行，服务硅基的都赚钱
模型公司的赚钱时间与标准
市场总是有新的问题用的那些token所带来的ROI是怎么样的？
裁员似乎就意味着有价值，阑夕会被骂的……
还是得看ChatGPT的收入
中国的五层蛋糕和美国有什么异同？
肯定会被打脸的做出预测
美国-功利化，中国-美好而无用的事情上，可能是一个分化
AI硬件的小小展开
Show notes
This live broadcast episode was invited by SinoNews-Cyber Dialogue.
Me, CEO Gao Fei of Ziding Technology, and my old friend Lan Xi discussed the current situation in the AI industry and related money matters.
The video version of this episode of Cyber Dialogue is already online. [Here:] (https://weibo.com/5893229038/R1aFXtghh)
Every time there is a program with Lan Xi, there will inevitably be a summary called "Lanxin Yiyi".
This episode is also particularly grateful for the support from MWC26 Shanghai;
MWC26 Shanghai focuses on "Innovative Thinking through Collective Intelligence",
centering on strategic directions such as intelligent infrastructure, interconnected coexistence, mobile AI, and enterprise digital transformation,
creating a comprehensive intellectual feast covering technology, policy, business, and ecology.
Over three days,
every speech directly hits the core of the industry,
each discussion concerns future trends,
and every segment showcases excellent windows for showcasing thought leadership!
Exhibition website: www.mwcshanghai.cn
Scan the QR code below to get your ticket
Or enter 【FVPTSGL3QW】 on the official website to get your entry pass
【Interview Guests】
- Lan Xi: Founder of Ludong Network, Renowned Tech Media Personality
- Gao Fei: CEO of Ziding Technology
(If you want to join the fan group, please add WeChat first: rosicky311)
Now let's start the main content...
【Timeline】
03:11 Everyone is talking about money
04:08 Three angles of money: earning money, spending money, making money
- Part One: Earning Money
05:08 Feeling about charging for Douba
08:16 Why does Pony Tencent say "the boat has holes"?
09:46 In the long term, why can't model companies continue to provide services for free?
11:42 If you were the boss of Douba, wouldn't you be envious of Jinmeng's income?
12:30 The subscription model for AI products is similar to fitness cards: I bet you won't come
12:56 From the financial reports of listed companies, the gross profit margin of inference can still be achieved, but losses mainly come from training
13:52 After Douba, who do you think will be the next one to try charging?
15:01 The breakdown of AI product subscription pricing systems
17:58 What do people expect the paid version of Douba to provide
18:32 Looking ahead, AI may also bring an "achievement gap"
19:54 Can advertising become a business model for AI?
20:40 Why are telecom operators suddenly pushing token packages recently?
22:21 Will the Maltese universal free token model be imitated?
23:19 Light and heavy talk: AI is like basketball; you can't learn to play basketball by watching 100 videos; you need to practice
25:17 Recently, many tech giants have been booed by college students
25:37 A certain technological determinism: technology cannot be stopped
- Part Two: Spending Money
27:34 "We didn't realize we had reached the limits of human capabilities in many things"
28:49 We all need to explore and build data centers in space
29:23 China's large language model vendors divide into two factions: big players and so-called several small dragons
30:24 Overseas are the three giants, but why are there so many more in China?
31:57 Video model case: despite Seedance's strength, why can the market's 5678th place still receive substantial funding?
34:26 Dialectical view of cursor value
35:00 APIs beyond subscriptions
- Part Three: Making Money
39:40 Old Huang proposed the five-layer cake theory
41:56 Other layers are supply logic, but applications are not
42:56 "Humans are not good at it" ah, service providers for carbon-based materials don't work, but those for silicon-based materials make money
43:47 When will model companies start making money and what are the standards?
44:45 There are always new problems in the market: what is the ROI of the tokens being used?
45:47 Layoffs seem to mean value, and Lan Xi would be criticized...
46:06 It still depends on ChatGPT's revenue
47:15 How are China's five-layer cakes different from those in the US?
50:46 Predictions that will definitely be proven wrong
51:27 There may be a division between the US - utilitarianism, and China - beautiful but useless things
52:57 A brief look at AI hardware
[Podcast Introduction]
Dragon Slayer is a podcast focusing on technology, internet, and risk investment.
Anchor Zhuang Minghao is someone who operates at the edge of the internet and risk investment industries.
His accumulated insights over many years seem to have become something of a "dragon slaying technique"...
Perhaps when a new "dragon" appears again, this podcast will stop updating.
[IMA Knowledge Base] Dragon Slayer VIP Report Library **ima.qq.com**
