you2idea@video:~$ watch eC85AquemOU [18:33]
// transcript — 242 segments
0:02 OpenAI made three major announcements last week, and if you only caught one or
0:06 two of them, you may have a completely wrong idea about what's going on. Chad
0:12 GPT now has ads. They are rolling out an $8 a month plan, and they finally
0:18 revealed some actual revenue data. All three of these happened within the same
0:22 week, and it's not a coincidence, and I'm going to explain why. So, first
0:27 Sarah Frier, the CFO, chief financial officer of Open AAI, released
0:33 information about ChatCBT's revenue. And for anybody thinking that OpenAI wasn't
0:38 doing well, you were wrong. First, both weekly active users and monthly active
0:43 users are continuously reaching all-time highs. And they specifically say it's
0:48 because of the flywheel that they have built across compute, Frontier Research
0:54 products, and monetization. And that's the key word. We're going to get to that
0:57 in a moment. Investment in compute powers leading edge research and step
1:01 change gains in model capability. Stronger models unlock better products
1:06 and broader adoption. Adoption drives revenue. And revenue funds the next wave
1:11 of compute and innovation. More compute equals better models. Better models
1:15 equals more use cases and more usage. More usage equals more revenue. And more
1:20 revenue means better models. And here's the key. Their revenue growth is
1:25 directly tied by how much compute capacity they have. So they are
1:30 literally constrained by how many GPUs they have. Compute grew 3x
1:36 year-over-year or 9.5x from 2023 to 2025, from2 gawatt in 2023,6
1:43 in 2024 and their revenue followed the same trajectory. So 2 billion to 6
1:49 billion to 20 plus billion. In fact, just a couple months ago, Sam Alman
1:53 talked about having 20 billion in revenue and people looked at him like,
1:57 "I I don't think you do because they had previously shared numbers that were a
2:01 little bit lower than that, but it turns out he was basically talking about their
2:05 current revenue. So, here's what that looks like. This is compute and of
2:10 course this is revenue on the right. And as you can see, they are completely
2:15 correlated. That is an insane place to be as a company. The only constraint you
2:20 have on revenue growth is how many GPUs you can buy and implement. And this is
2:25 also why they struck a $10 billion deal with chip company Cerebrris. Cerebras
2:30 makes specialized AI chips, some of the fastest inference on the planet, and
2:35 they're hoping that that's just going to unlock that much more inventory. So, one
2:40 thing they did not reveal in this post is their profitability. they are almost
2:44 certainly losing money on most tokens that they're generating. But again, keep
2:48 that in mind as I tell you about these other announcements. So, Dr. Ginger
2:53 Balls did a quick analysis about their potential profitability. And let's take
2:58 a look. So, OpenAI has made about $1.20 per kilowatt hour of compute each year
3:02 for the last 3 years. Zero cost reduction in 3 years. Models are getting
3:07 cheaper to run, but simultaneously they're building bigger and better
3:10 models which are more expensive to run. And so that's probably why we're seeing
3:15 the cost stay the same. It takes 1 kilowatt to run an H100. They make $1.20
3:22 per GPU hour reserved instances. So if you go to the market and rent a GPU, $2
3:27 to $7 per hour per GPU. He even mentions it's probably being run at a loss for
3:31 everybody. So in the best best case, OpenAI is slated to lose 20 billion this
3:35 year. signal reminds us that they projected $und00 billion in revenue in 3
3:40 years, which implies a 2x year-over-year from here, which since they've been
3:46 doing 3x is not crazy to think that they might actually achieve this, especially
3:50 when you factor in all of the other products and services that they're
3:54 rolling out, increased adoption of AI more generally, increased enterprise
3:58 usage, and again, a lot of people think OpenAI mainly focuses on consumer, but
4:02 they are doing quite well in enterprise. I'll get to that in a moment. And of
4:05 course, an entirely new consumer hardware device to compete directly with
4:11 the phone. And they haven't even started factoring in ads. >> I kind of think of ads as like a last
4:16 resort for us for a business model. >> So, let me put OpenAI's revenue into
4:20 context. Of course, a lot of people are comparing them to Anthropic. This is an
4:25 article from Techrunch from just a few months ago. At the beginning of 2025,
4:28 less than 2 years after launch, Anthropic's run rate revenue had grown
4:31 to approximately a billion dollars. By August 2025, just eight months later,
4:37 our revenue run rate is over 5 billion, making Anthropic one of the fastest
4:40 growing technology companies in history. And Anthropic is primarily making their
4:45 money from API calls. They almost exclusively are making their money from
4:49 API calls because that's what their models are known for. They're known for
4:52 being the best coding models on the planet. And so plugging into cursor,
4:57 plugging into cloud code, plugging into factory, all of these different products
5:01 and services are consuming their API calls. And that is very different from
5:04 what OpenAI is doing where they are primarily subscription-based. And that
5:07 probably speaks to why they're rolling out an $8 a month plan. Again, I'll get
5:12 to that in a moment. And here's a source for how that revenue mix breaks down.
5:16 This is from the Observer. Consumer subscriptions 55 to 60%, enterprise
5:21 solutions 25 to 30 and API and developer platform 15 to 20%. So the vast majority
5:27 of their revenue strictly comes from monthly subscriptions. And also the
5:31 enterprise solutions count as subscriptions. They're just not consumer
5:35 subscriptions. And next I just want to briefly show you this report published
5:40 by RAMP, a financial services company which has really good visibility into
5:45 how AI adoption is fairing. And so if we go down here and just specifically look
5:50 at model adoption rate, we can see still OpenAI is the absolute leader. And so
5:57 now enter the $8 a month plan. Just a couple days ago, OpenAI announced that
6:02 they are rolling out their $8 a month plan worldwide. And the reason I say
6:06 worldwide is because it was already available in India. And so why did they
6:10 make an $8 a month plan? Well, they needed a way to capture some of the
6:15 value of a market that traditionally doesn't have the purchasing power of a
6:20 United States or other countries that can afford the $20 a month plan. Of
6:25 course, they came up with $8 a month. So, in August 2025, we introduced Chai
6:29 GPT Go, that's what the plan's name is, in India as a lowcost subscription
6:34 designed to expand access to Chai GPT's most popular features and help more
6:37 people use advanced AI in their daily life. And now they're rolling it out
6:42 worldwide. But why? That's the critical question. First, let me tell you what
6:45 you actually get with the Go plan. You get 10 times more messages, file
6:49 uploads, image creation as compared to the free tier. But here's the key. It's
6:55 all with GPT 5.2 instant. So although you get access to the latest model, you
6:59 get the instant version, which is typically a smaller, cheaper, and faster
7:04 model to serve. And so what they're really trying to do is capture more of
7:08 the market, but push them towards their smaller, easier to serve models. And
7:13 what they're hoping is that the smaller and easier to serve models will continue
7:17 to get cheaper, and the profit margins on those models will increase. But
7:22 profit probably isn't important right now. In fact, at $8 a month, they're
7:27 probably losing money on each of their users. Now, I'm sure there's a large
7:31 swath of users who aren't coming close to spending $8 per month in inference
7:36 even though they're paying $8 per month. And so, I don't have information about
7:40 what that mix looks like, but most likely they are losing money. So, why
7:44 would they do that? This is intentional. It is called a loss leader. They are
7:48 trying to capture the market because once you start using a product like
7:52 chat, GPT, like any artificial intelligence, you become more and more
7:56 locked in. And that is because the more you use the AI, the more personalized it
8:01 gets, the more it remembers about you and the more locked in you are, the
8:06 harder it is to switch off of chatbt. So of course, offer it as a loss leader.
8:10 Get people to come use it. Get them locked in. And we're only talking about
8:15 consumers. Once consumers adopt a product, they are even more likely
8:19 within the context of their company within enterprise to adopt it there as
8:24 well. Imagine you are super familiar with chatpt and then at your company you
8:29 obviously say well of course I want to use chatbt that's what I know best
8:32 that's what I'm familiar with why would I use another AI product and so there's
8:39 a confluence of things happening one they are expecting the models to get
8:42 cheaper two they're expecting their ability to serve the models to get more
8:48 efficient again that's why they made this deal with cerebras cerebras at
8:53 scale is incredibly good at serving inference really efficiently. And as I
8:58 said a few videos ago, inference is where the money is at. And this kind of
9:02 proves it. As we see demand for inference increase, we're also seeing
9:06 their revenue increase. And this is that grand shift to inference that I've been
9:09 talking about. And there's an additional flywheel that happens. As they get more
9:14 users, they're collecting more data. All of that data is then used to train
9:20 future models. So all of this usage is actually really valuable to them. It's
9:24 exponential in nature. So this is why they're willing to lose so much money
9:28 now. This is probably the story they're telling investors as they're asking for
9:33 more billions and billions of dollars. And then think about when the hardware
9:38 device comes. Everybody's already locked into Chad GPT. They're looking at the
9:43 iPhone, which now is going to be powered by Gemini from Google. They're looking
9:47 at this new device from OpenAI thinking, "Wow, Chachi BT is really personalized
9:53 to me. I'm already locked in. I'm already paying a monthly subscription to
9:58 OpenAI. Maybe I should check out whatever OpenAI is going to be
10:01 releasing." And as they capture more and more of the consumer market, they will
10:05 also capture more of the enterprise market. And so, this is something that
10:09 Anthropic is really missing. As long as Anthropic continues to build the best
10:13 coding models in the world, they're going to be just fine on the enterprise
10:18 side. But for non-coding knowledge work, which hasn't yet completely exploded,
10:22 like the coding use case, there's still so much potential there. And Chai GPT
10:28 seems to be the go-to model for those use cases. And so, what is bundling?
10:32 Bundling is selling products and services together for a single price as
10:36 a bundle. Microsoft has been doing this for years. is it is an extremely
10:40 aggressive and sometimes even monopolistic strategy and one of the
10:45 best examples of this is Microsoft Teams versus Slack. Slack came out years
10:50 before Microsoft Teams. Slack is a superior product in pretty much every
10:57 way, but as you see here, because of the distribution that Microsoft has with
11:02 Microsoft Office and Outlook and Windows and everything else that they're already
11:07 selling, they were able to completely dominate growth. And Slack is still
11:12 growing, but it's nowhere near what Microsoft Teams is. And again, Teams is
11:17 an inferior product, but Microsoft would say, "Hey, you're already buying Office.
11:21 We're just going to throw in Teams for you." And so, it made it very difficult
11:26 for the accounting department, for the purchasing department in a company to
11:31 justify having a separate subscription, let alone paying for one where Teams is
11:35 essentially free at the beginning. And so, that's probably what OpenAI is
11:39 thinking they're going to do. capture as much of the market and then as they
11:42 release subsequent products and services, they're going to bundle it
11:46 together in enterprise. Well, you're already using Chad GPT. Why not use our
11:51 device? Why not use our coding API? We'll give it to you really cheap since
11:55 you're already paying us a subscription. Forget Anthropic. Forget Google. And
11:59 that is a very aggressive and successful strategy. And now let's talk about the
12:02 third pillar of the three announcements that were made last week and that is
12:08 ads. Ads are coming to Chachi PT specifically. They're coming to the free
12:12 tier and they're coming to the go tier. So even though you're going to be paying
12:16 $8 a month, you might still get advertising in ChachiPT. Now, this is
12:23 kind of crazy because even just 2 months ago, they were saying they weren't going
12:28 to be doing ads. Every major tech company goes through this. There's
12:32 examples from Google. In the early days, Google said, "We will not have
12:36 advertising." But of course they went on to have advertising and it became the
12:41 best business in the history of all business. Then Meta making billions and
12:45 billions of dollars from their advertising. They have free products.
12:50 Facebook, Instagram, WhatsApp, these are all free products that they monetize
12:55 purely from advertising. Of course, there's a couple subscriptions here and
13:00 there, but like 95% of the revenue that they make is from advertising. And so
13:05 here's some of the announcement. What matters most? Responses in Chai GBT will
13:08 not be influenced by ads. Of course, they say that ads are always separate
13:12 and clearly labeled. Good. Your conversations are private from
13:15 advertisers. Plus, pro business and enterprise tiers will not have ads. And
13:20 so, here's an example of what an advertisement might look like. So, the
13:23 user is basically asking for a recipe. They give a recipe and then below it, an
13:30 ad for hot sauce. Fine. Why is this actually connected to the other two
13:34 stories that I mentioned? Well, I think if they had rolled out ads independently
13:40 of those other two stories, specifically about their revenue details, a lot of
13:44 analysts may have looked at and said, "Uh-oh, they're in trouble." Because
13:47 they are scrambling to make more revenue. But instead, in parallel, they
13:52 said, "No, we're making a cheaper plan. Plus, by the way, we're making a ton of
13:55 revenue. We have 20 billion this year, and by the way, we're going to be
13:59 introducing advertising now." So, it's a really smart communications play now
14:04 because they have proven that they have the revenue because they have proven
14:08 that they have the growth. Nobody's going to look at them and say, "Oh,
14:11 they're introducing advertising because they are basically going to be out of
14:16 money in 18 months." Meta made $58 per user in 2025 purely from ads. $58 per
14:24 user, which is quite a bit for an absolutely free product. Now imagine
14:29 chat GPT having advertising. Would they monetize users at a higher rate or a
14:35 lower rate? Well, in my opinion, Chat GPT is actually closer to the final
14:39 intent of a user. And whichever product or service is closer to the final intent
14:44 of a user, that is where the most money per user will be made. Imagine OpenAI
14:49 hits a billion free users in 2026. It's at 800 million weekly right now. So,
14:53 this is conservative. If they monetize at just 9% of Meta's ARPO, average
14:59 revenue per user, that's 5 billion in incremental revenue. 18% gets you 10
15:05 billion. Full parody is 57 billion per year from ads alone. So if Chhat GPT is
15:11 able to monetize as well as Meta has been on free products, they would make
15:16 $57 billion per year. That's why they're doing it. There is a ton of revenue to
15:20 be had there. So in parallel, while they're trying to capture as much of the
15:23 market as possible, they can actually increase monetization per user. And so
15:28 that $8 a month starts to look like 10 or 12 or 15 or even $20 a month because
15:34 they're doing this hybrid paid subscription plus advertising. Again,
15:38 part of the flywheel. You're giving OpenAI more data. They're building a
15:44 personal model of you and thus the advertising will be more personalized
15:50 and more relevant to you and your needs. And that's what it's all about. It's not
15:54 only about being as close to intent as possible, but the way you get as close
15:58 to intent as possible is by having the best information about the user so you
16:03 can serve the most relevant ad. Now, personally, I hate ads. I am fine paying
16:09 $20 a month or $200 a month if I never see ads. But remember, a lot of people
16:14 can't afford that. A lot of people throughout the world can't afford the $8
16:18 a month. So, they're happy using a free plan with some advertising in it. And
16:22 you know what? That's fine. That's their choice. That's OpenAI's choice. They're
16:27 getting a free amazing product. They're basically getting alien intelligence in
16:32 the sky absolutely free. Go ahead, serve some ads to people. Open AAI said
16:36 they're making a ton of money, 20 billion this year. That is 3x
16:40 year-over-year every year for the last 3 years. And they're continuing to grow.
16:44 They continue to get all-time highs of weekly average users and monthly average
16:49 users. At the same time, they introduced an $8 a month plan to capture more of
16:54 the market. They're already making a ton of revenue. Why not make some more? Why
16:58 not capture people who can't afford the $20 a month plan? Then finally, they're
17:03 serving ads. So at the same time that they're making so much money from
17:06 subscriptions directly, they can now layer advertising revenue on top of
17:10 that. Nobody's questioning whether they're making money or not, nobody
17:15 looks at advertising and thinks, "Oh boy, they are scrambling to make revenue
17:19 and maybe they're going to run out of money." No one believes that anymore.
17:24 That is why the release of these three announcements in the same week was
17:29 really incredibly smart. 4D Chess. But there's one company we haven't really
17:34 talked about, and that's Google. Google is probably OpenAI's biggest
17:38 competition. As I've said before, they have the hardware with Android. They
17:43 have the users, billions of them, using Google search already. They have
17:47 frontier models with Gemini. And one thing that they have that OpenAI does
17:52 not, diversified revenue. They essentially have unlimited funds to
17:57 throw at their AI products and services. And so when we think about bundling,
18:02 they already have Google Workspace and they can offer Gemini as part of that
18:06 basically free. So really that's probably OpenAI's biggest competition
18:11 today. So that's the full breakdown. Let me know what you think in the comments.
18:13 If you enjoyed this video, please consider giving a like and subscribe and
18:17 I'll see you in the next one. And a quick shout out to Theo who made a video
18:21 about this topic and inspired me to make this video. And by the way, if you want
18:25 to stay up to date on all of the AI news coming out like these three stories, be
18:29 sure to subscribe to our newsletter at Ford Future. I link in the description
$

OpenAI just made a genius move

[AI agents and automation][community building and audience growth][marketing and growth hacking][revenue model and pricing strategy]
// description

OpenAI coming out with a bunch of huge news lately, and it all means something if you take it together. Sarah Friar https://x.com/i/trending/2013013390281568746 Ads in ChatGPT https://x.com/OpenAI/status/2012223373489614951 Go https://openai.com/index/introducing-chatgpt-go/ Revenue VS Anthropic https://x.com/signulll/status/2013055746015899670?s=20 Ramp Report https://ramp.com/velocity/ai-index-january-2026 Analysis https://x.com/signulll/status/2013066125970473128?s=20 Join My Newsletter for

now: 0:00
// tags
[AI agents and automation][community building and audience growth][marketing and growth hacking][revenue model and pricing strategy]