Transcription
[music] In November of 2025, Google launched its latest AI chatbot called Gemini 3, which was much better than its previous chatbot, Gemini 2. According to many users, it's also much more capable than OpenAI's Chat GPT. They concurrently launched a new image generation model called Nano Banana Pro, which can generate hyperrealistic images and infographics with readable text. It is far superior to chat GPT's image generation capabilities.
This has been a disaster for OpenAI. Upon the release of Gemini 3 and Nana Banana Pro, Gemini's app downloads have surged. Google's gain is OpenAI's loss. According to some estimates, ChatGpt's usage numbers have recently started declining. Open AAI incurs billions of dollars of losses per quarter.
In the past, Sam Alman was able to convince venture capitalists that OpenAI is a technological leader and they need billions of dollars of investment to stay ahead. Despite its massive investments, OpenAI is no longer the leader. With Google's Gemini now appearing to be better than Cad GBT, Sam Alman declared a code red internally. He's telling his employees they need to make Chad GBT better, otherwise they'll lose market share to Google.
This couldn't come at a worse time for OpenAI. In addition to losing billions of dollars per quarter in their core business, they have $1.4 trillion of data center spending commitments, which they have no way to pay for. OpenAI is reportedly in discussions with sovereign wealth funds and other large investors to raise $100 billion in early 2026. This would be the largest funding round for any company ever. But why would anybody invest so much money into OpenAI when they're no longer the technological leader?
Since its release in late 2022, Chad GPT has grown rapidly. Open AAI is not a publicly traded company, so they're not required to officially disclose their financial results or key performance indicators. But Sam Alman has sporadically disclosed some information about OpenAI in an attempt to garner media attention. In October of 2025, he claimed that Chat GBT has 800 million weekly active users, a 60% increase from the 500 million users they claim to have in April. But there is reason to believe that growth has started to slow down and may in fact have turned negative.
The third party data provider Apptopia provides estimates for Cad GPT's daily active users. The number of daily active users was increasing strongly throughout 2024 and most of 2025, but in September it flatlined and even decreased slightly. Apptopia also estimates the average amount of time that each user spent on the app each day. Back in July, the average Chat GBT user spent about 27 minutes per day on the app. By October, this had declined to 21 minutes.
The decrease in time spent doesn't necessarily mean that consumers want to use AI less. It's more likely the result of increasing competition. The vast majority of consumers only use the free version of these AI apps. Maybe you used to only use ChatGpt, but now you also use Gemini sometimes, so you use ChatGpt less.
OpenAI's biggest competitive threat by far is Google. The data provider Similar Web provides market share estimates for generative AI companies. As of October 2024, OpenAI had about 85% market share. By October of 2025, this had declined to about 75%. Google has about 15% market share. In July, Google disclosed that Gemini had 450 million monthly active users. By October, this had increased to 650 million, and that's only as of October. In November, they launched Gemini 3, which was a big improvement over Gemini 2. Their monthly active users and market share likely increased significantly since then.
When Google launched Gemini 3, they published its results on various benchmark tests. According to the benchmarks, it is significantly better than Claude Sonnet 4.5 and ChatGBT 5.1. These were the two best models made by Anthropic and OpenAI at the time. Most benchmarks rely on narrow artificial tasks that can be gained through fine-tuning and prompt engineering. Companies choose which benchmarks to highlight and when to refresh them, making the results as much a marketing exercise as a technical one. So, the best way to compare various AI models is to just tinker around with them and see how useful the results are.
As an experiment, I gave both Gemini 3 and ChatGBT 5.2 the same prompt: "Tell me about OpenAI recently declaring code red as I lose market share to Google's Gemini 3. What implications could this have for OpenAI and their ability to raise venture capital or complete their expected upcoming IPO?" They both gave me very similar results. Both of them explained that OpenAI is worried about losing market share to Gemini. This has caused OpenAI to prioritize improving Chat GBT at the expense of some other initiatives. And while OpenAI still has major venture capital backers, the loss of their technological leadership could impair their ability to raise capital in the future.
I've tinkered with both Gemini 3 and ChatGBT 5.2 for a few hours each, and to me, they seem approximately equivalent. But one area where Google is indisputably ahead of Open AI is image generation. Google's new Nano Banana image generator can generate hyperrealistic images that look much better than anything Open AI can make. OpenAI does have the ability to generate images and even videos, most notably with their video generation app, Sora. Sora videos can look shockingly real at first glance, but upon careful inspection, it's pretty obvious that they're fake. The image quality of Nano Banana is unquestionably better than anything we've seen from ChatGpt or Sora.
Shortly after the release of Gemini 3 and Nano Banana Pro, Sam Alman sent a memo to OpenAI employees. He declared a code red emergency. He told his employees they need to make ChatGBT better, otherwise they'll lose market share to Google. OpenAI will postpone work on other initiatives, including advertising, AI agents, and a personal assistant called Pulse. Instead, they should focus all their resources to improving Chad GBT, including its image generation capabilities.
On December 11th, OpenAI launched its latest model, Chad GBT 5.2. This was just a few weeks after Google launched Gemini 3, and less than 2 weeks after Sam Alman declared code red. The timing is probably not coincidental. Alman saw that Chad GBT was losing market share, so he rushed to push out whatever new update he could. Sam Alman claims that Chad GBT 5.2 is the most advanced AI model in the world, far surpassing Plaude and Gemini on various benchmarks. We'll reiterate again that these benchmarks aren't very useful because there are ways to manipulate them or specifically fine-tune the models to perform well on narrow benchmarks.
The actual consumer reception has not been great. The most common criticism is that it feels significantly more restrictive than prior versions. Many users describe it as an overfiltered or nanny model that injects safety disclaimers, warnings, or moral framing into otherwise normal requests, which they feel breaks conversational flow and makes the assistant tiring to use.
On December 16th, OpenAI launched a new image generation model called Chat GPT image 1.5. This allows you to make hyperrealistic images. You can also upload a picture of a real person and have it incorporate the real image into a fake background or make other changes to it. While image 1.5 is a significant upgrade over OpenAI's previous image generation tools, it still lags far behind Google's Nanobanana.
In a recent article on CNET, reporter Caitlyn Chadawi compared ChatGpt image 1.5 to Nano Banana. The images on the left were made with Nano Banana, while the images on the right were made with ChatGpt image 1.5. The first image is a cat sitting on a Monopoly board. The Chat GBT image is obviously fake. The cat's fur is too smooth to be real. If you look closely at the Monopoly board and Monopoly money, they have incoherent squiggles where words are supposed to be. The Gemini image on the left looks far more realistic. Had you told me it was a real photo, I'd probably believe you. In the second image, Gemini also looks far more realistic than Chad GBT. In the Chad GPT image, the falling snowflakes are far too big and fluffy to be real. Chat GPT continues to struggle with the minute details of images, especially when they contain text or intricate shapes. It can make images that look real at a distance, but gets the small details wrong.
On the same day that GPT image 1.5 was launched, Sam Alman tried to showcase his capabilities on his Twitter account. He posted a graphic of a sexy Christmas calendar showcasing himself as a jacked firefighter. Creepiness aside, notice that it shows December 1st as a Thursday. In 2025, December 1st was actually a Monday. That's the problem with ChatGpt. It just generates slop. It can't fact check itself.
Open AAI, Google, and all the other AI companies are constantly improving their models. So, it's not particularly important to compare the capabilities of their products at any given point in time. It's more important to look at the bigger picture and we can see that Google has a few key advantages over Open AI.
Open AI burns billions of dollars per quarter and is completely reliant on venture capital funding to sustain its losses. Today, OpenAI has raised approximately $60 billion of venture capital. Google is also spending tens of billions of dollars on AI research and compute. Gemini is almost certainly losing money, just like Chat GBT. But Google can afford these losses. In the first 9 months of 2025, Google's parent company, Alphabet, generated $86 billion of operating cash flow. Just in the first nine months of this year, Google's generated more cash flow from its core businesses than all the venture capital OpenAI has raised since inception. This gives Google a huge advantage. They can afford to invest as much money as they want into Gemini indefinitely. Market sentiment and the health of the venture capital industry are irrelevant to Google because they generate more than enough money internally.
Secondly, Google has massive distributional advantages. They incorporate Gemini into Google search. People get used to Gemini as they interact with it on Google search. Some of them will go on to download the Gemini app. Google also makes Gemini integrate seamlessly with their other products like Google Docs and Google Sheets which are already used by hundreds of millions of people. Finally, Google pays huge amounts of money to smartphone makers including Samsung to have Gemini pre-installed.
Open AAI had a first mover advantage allowing Jad GBT to gain dominant market share. But in the long term, it should be expected that Google with its massive advantages will become the dominant player in generative AI.
Open AAI currently loses money on chat GPT. Their computational costs are so high that they lose money even on the paid chat GPT plus and pro subscriptions which cost $20 and $200 per month respectively. And they obviously lose money on free users who generate zero revenue. According to the information, only 5% of ChatGBT users pay for plus or pro subscriptions. So, OpenAI generates zero revenue from 95% of its users.
To address this imbalance, OpenAI has long planned to introduce advertising inside Chat GBT as a way to monetize free users and has reportedly had engineers working on building an ad platform. However, when Sam Alman declared a companywide code red, those resources were reportedly pulled away from advertising and redirected towards improving Chad GBT's core capabilities in response to competitive pressure from Gemini. The result is a strategic trade-off. Even if OpenAI succeeds in closing the technical gap with its competitors, it is pushed further into the future any credible plan to monetize the vast majority of its users.
This begs questions about how OpenAI plans to generate enough revenue to cover its $1.4 trillion of data center spending commitments. Alman claims the company has an annual revenue run rate of $20 billion. This isn't nearly enough to cover their ongoing compute costs, let alone their massive data center expansion plans.
On October 30th, Sam Alman appeared on a podcast where he attempted to explain OpenAI's plans to increase its revenue. "We do plan for revenue to grow steeply. Revenue is growing steeply. We are taking a forward bet that it's going to continue to grow grow and that not only will chache keep growing but we will be able to become one of the important AI clouds that our consumer device business will be a significant and important thing that AI that can automate science will create huge value."
So, Alman says he expects open AAI to increase its revenue with three new business lines. Firstly, OpenAI will create an AI cloud. It sounds like he wants OpenAI to build its own data centers and sell computing power to other AI companies. I have no idea how he intends to do this. Open AAI does not own or operate its own data centers. They purchase computing power from cloud service providers such as Microsoft, Oracle, and Coreeave. Open AAI's $1.4 trillion of current spending commitments is not to build its own data centers. The new data centers are being built by the likes of Oracle and Coreeave. OpenAI just promised to purchase compute from them in the future. I honestly have no idea what Alman means when he says that OpenAI will become one of the important AI clouds. Maybe they're going to purchase compute from hyperscalers and resell it to other AI companies as a middleman.
Secondly, OpenAI plans to develop and sell its own consumer hardware device. In May of 2025, OpenAI paid $6.4 billion to acquire a company started by former Apple chief designer Johnny IV. OpenAI wants to use Iive's expertise to build a new AI consumer device. It's supposed to be some Amazon Echo type thing that can fit in the palm of your hand. It will have cameras, microphones, and speakers on it. You can talk to it and ask it questions.
AI devices have been tried before. For example, in 2023, a startup called Humane AI unveiled an AI pin that clips onto your shirt. The co-founders of Humane previously worked as designers at Apple just like Johnny IV. Alman was a big believer in Humane. He personally invested into one of Humane's private funding rounds. Humane started selling its AI pin in early 2024. It received extremely poor customer reviews and generated minuscule sales. The company shut down in early 2025. The Humane AI pin was basically just a more expensive and less reliable version of the Amazon Echo. We actually made a video about Humane last year, which is linked in the description below. Given that you can already use Chat GPT or any other AI model on your phone or computer, I see no reason why anybody would want a dedicated AI device. The humane AI pain was an idiotic idea that was doomed to fail from the beginning, but Sam Alman thought it was a good idea and invested his own money into it. The development of OpenAI's consumer device is reportedly not going very well. They'll probably eventually make something similar to the Humane AI pin and it will be a commercial failure.
Finally, Alman says OpenAI will create tremendous economic value by automating science. Alman has long made grandiose statements about the potential for AI to make scientific discoveries. For example, in a recent blog post, he said that if OpenAI had 10 gigawatts of computing power, they could figure out how to cure cancer. To date, neither Chat GPT nor any other AI model has made any significant scientific discoveries. I'm sure that many scientists use AI tools, including ChatGpt, to do some of their grunt work, like crunching data. Maybe this can speed up their scientific discoveries, but I don't see how OpenAI can generate huge revenue from this. A scientist pays the same subscription cost for Chad GBT as anyone else. As an analogy, scientists use laptop computers. They are far more productive with a computer than they would be without one, but they still pay the same normal retail price to buy the computer. If a scientist makes a great scientific discovery, this doesn't benefit the laptop manufacturer. Even if Chat GBT becomes useful to scientists, I fail to see how this will bring in a large amount of revenue to OpenAI.
Open AAI claims that Chad GBT 5.2 is marginally better at doing some science tests as compared to Gemini 3. But just because you can memorize a bunch of answers to science tests doesn't mean you can make new discoveries. The intelligence of large language models is constrained by their training data. It may not even be possible for them to make novel discoveries. When Sam Alman talks about automating science and curing cancer, this is an artifice to fool the gullible. He makes grandiose predictions to justify his massive spending commitments, but there's no proof that any of this is even possible.
Alman claims he can eventually cure cancer with AI, but in the meantime, he's resorting to some far less glamorous initiatives to bring in revenue. In October, Alman said they will soon roll out an age verification feature for ChatGpt. Adult users will then be allowed to use ChatGpt to make erotica. Maybe we'll see an erotic cartoon series about Sam Alman, the firefighter.
Altman also has a plan to start monetizing the Sora video generation app, which has to date been a cash incinerator. In early December, they announced a partnership with Disney. Disney agreed to license 200 of its characters, including Marvel and Star Wars characters, to be used on Sora. Open AI incurs astronomical computational costs to operate Sora and to date is believed to generate minuscule revenue. Starting in early 2026, Sora users will be able to pay real money for the privilege of generating AI slop videos with Disney characters. For example, you can make a fake video of yourself as a Jedi. Disney will get a cut of this revenue. I would be shocked if a significant number of people would be willing to pay real money for this. I expect the Disney Open AAI partnership to generate minuscule revenue.
Open AAI needs to increase its annual revenue by more than 10-fold if it is to pay for its massive data center spending commitments. Now that they're losing market share to Google, this outcome seems increasingly remote. We're already starting to see fear spread across the market.
In September, OpenAI signed a $300 billion cloud computing deal with Oracle. Starting in 2027, OpenAI will need to pay Oracle $60 billion per year for cloud computing services. Oracle needs to spend hundreds of billions of dollars to build these data centers before OpenAI starts paying them. Oracle doesn't have enough cash to fund this construction on its own, so they're borrowing money from various banks and other financial institutions. One of the lenders is a private equity firm called Blue Owl Capital. Blue Owl already agreed to fund one of Oracle's data centers in Texas. They were in talks to provide an additional $10 billion to fund another data center in Michigan, which will also be leased to OpenAI. On December 17th, it was reported that Blue Owl pulled out of the deal. They were probably concerned about OpenAI's ability to make good on its spending commitments. If OpenAI can't make good on its payments to Oracle, it's unclear how Oracle will be able to pay its massive data center loans. Credit default swaps on Oracle's bonds have recently surged to over 100 basis points. Investors are now betting that there's a 1% chance that Oracle will default on its debt within the next 5 years. While 1% doesn't sound like a lot, that's pretty high for a company as massive as Oracle.
Open AAI itself continues to incur operating losses. They've raised $60 billion of venture capital funding today. It seems like they've already burned through most of this. The company is currently trying to raise $100 billion at an eyepopping $830 billion valuation. They want to do this funding round in the first quarter of 2026. $100 billion is an absurd amount of money to raise in a private funding round. It would be the largest funding round in history. I don't know what investors even have this kind of money. And why would any investor pay such an absurd valuation for a company that has $20 billion of revenue, incurs massive operating losses, and is losing market share to Google?
Over the coming months, we'll be watching very carefully what happens with this $100 billion funding round. If this funding round fails, it will mark the beginning of the end for Open AAI. They will default on their spending commitments and the entire house of cards will come crumbling down.
All right, guys, that wraps it up for this video. What do you think about OpenAI's code red? Let us know in the comments section below. As always, thank you so much for watching and we'll see you in the next one.