There are a ton of companies producing their own large language models (LLM) to compete with Chat GPT. Major companies like Meta, Google, Alibaba and even Baidu all have their own versions of basically the same thing. Some of OpenAI's former executives even left and started their own company called Anthropic where they have released an LLM called Claude that is one of the more promising LLMs out there. It seems like the vast majority of them are targeting software engineering as their selling point. There are even some promising to replace software engineers entirely and faking the results at conferences cough cough Devin AI from Cognition Labs cough cough. People are constantly talking about ChatGPT, Claude, and Devin. However, Google's Gemini is starting to be a better option.
The Opportunity
These LLMs really are not even close to being good enough to replace software engineers right now. By the way I should probably mention that I am a software engineer as my full time job. I should also mention that I wish to the sweet lord above that AI would replace me. Although these cannot replace software engineers right now, what these can do is knock out some boilerplate things for you pretty quick. These are things that basically every software engineer knows how to do but it's just boring and tedious when your time would be better spent on the actual meat of the project rather than styling a drop down menu.
This is where the opportunity is though. It would be insanely difficult for AI to replace software engineers, but it can definitely help them be more efficient. And that is how the companies should be marketing their products. There are a ton of software companies that would be willing to spend thousands of dollars on subscriptions to an AI that would make their software engineers more efficient, and really big companies could pay hundreds of thousands to millions per year. This seems like a lot, but it is cheap in comparison to hiring a few software engineers. For example, Netflix has 2,000 software engineers, many of which make 300k-1M+ per year. If there was an LLM that could make their software engineers 15%-20% more efficient and they were charging, let's say 100 per month per employee for a subscription, I think they would certainly do it. That only costs them a total of $2.4 Million, but the productivity boost would potentially be like having 100 more engineers which would cost them a ridiculous amount of money. And this is just Netflix. Amazon has 36,000 software engineers and Google has 27,000, and there are software engineers at almost every company imaginable. So, you can see the potential that this would have for the company that makes the dominant model.
Who Has The Best?
Many of the ones mentioned above are all very similar in performance. However, Gemini by Google is potentially one of the best out there, in particular Gemini 2.0. The response time of their model almost seems like you are watching it at 2X speed. It is insanely fast compared to its competitors, but it also performs just as good if not better. But speed is the real game changer here because of the efficiency we are talking about. Software engineers are not asking AI to code something complex, they want the time wasting monotonous tasks done, and this does them the fastest in my opinion (other than maybe Groq). And when compared to ChatGPT it is not even close; it feels like it is 5X faster.
If it is this fast and impressive it must be pretty expensive though. Nope! I believe it is the cheapest out of all of the LLMs out there. They are even cheaper than Chat GPT 4o mini. It is 10 cents for 1 million tokens in and 40 cents for 1 million out. It is hard to explain how cheap this is. Claude, which is one of the best, if not the best at performing coding tasks costs $3 per million tokens in and $15 per million tokens out, yet they have almost the exact same coding ability. Who do you think a company would go with? Unnoticeable difference in ability, faster, significantly cheaper. I think Google might be the best package overall right now. And it performs BETTER in many of the other tasks other than pure coding as well.
Their Advantage
Google has an enormous advantage over most other companies in the AI race due to the amount of data they have and get every single day. It is almost impossible to get as much data as them. They own the two biggest search engines, Google and YouTube. The amount of data they can collect every day for basically no additional cost is so hard to beat that it is hard to imagine how they were not at the forefront of this AI race since the beginning.
Google handles over 2,500,000,000 Gigabytes of data every day. That is unfathomable amounts of data at their disposal. And not only that, but they also already have an enormous staff of software engineers as well as the CPUs and other resources needed to make their models more efficient. It is really hard for me to imagine them not being one of the leaders of this movement. But is this being priced into their stock?
AI Priced Into Google?
AI stocks have been commanding extremely large multiples for years now, many of them are barely profitable or not profitable at all. But Google is a bit different, and it always has been for some reason. They have never really traded at some ridiculous multiple in recent history.
They are currently trading at 22.64X earnings, 20.75X forward earnings, and only 18.17X 2026 expected earnings. They have an average multiple of around 22-24X earnings and are now well positioned to be a front runner in the AI race. So, one would assume they are pretty close to being fairly valued right now if their guidance is correct, but let's see if that is the case.
Here are the present values for Google based on their guidance. Assuming their guidance is accurate, and the assumptions are correct, they are basically at fair value right now. If AI goes better than expected for them then they are probably undervalued, and after seeing how impressive their new model is, that might be the case. If AI does not go as expected for them then they are quite a bit over valued. A lot hinges on their ability to continue to drive growth and command a decent multiple. For their base case I gave them a multiple of only 17X which is the lowest they have traded at in recent history besides the crash in 2022 where they got to 16.5X at their bottom. It is better to be conservative for estimates like this, and a market downturn making them cheaper would be appreciated.
Conclusion
I don't think their potential AI dominance is being priced into the company. Partially because they were somehow late to the game despite having everything available at their fingertips. And partially because these LLMs aren't really profitable businesses yet and I don't think people on Wall Street really have any technical expertise, so they don't fully understand what these companies are doing. But there is a lot of potential for Google there, especially since they are able to own the entire end to end process from data to hardware. If they are one of the best and by far the cheapest then it might just be a waiting game. What I mean by that is if they are widely adopted into businesses to the point where it would be a real pain for these businesses to switch to something else, then they can slowly raise prices over time. This is similar to what streaming platforms do, where they build up a huge user base, get them accustomed to having the product and then they slowly raise prices which generates a ton of cash for the company. That is the potential of these AI models for Google. It doesn't mean that it WILL happen. It is a very competitive space and there are many good alternatives out there. However, it seems like almost every other company that has one is pricing in the potential that it is the next best thing except for Google. This is NOT a recommendation to buy GOOGL (I will never recommend buying anything), I am just bringing to attention the possible underappreciation of their AI capabilities by Wall Street.