How will the LLM market converge?
OpenAI vs Anthropic vs Llama vs Gemini vs....
When ChatGPT first launched back in late 2022, OpenAI had 100% market share of the LLM market. They seemingly came out of nowhere and were first to the space, defining what the product looked like for us consumers.
Since then, every major tech company has either developed their own models (Meta, Google) or made strategic investments in AI companies (Amazon, Microsoft), providing strong alternatives to ChatGPT. According to Menlo Ventures', Anthropic and Google’s Gemini has seemingly captured some market share off OpenAI this year, as Sam Altman’s company remains the market leader but now with only 34% market share, down from 50% last year.
Competitors are playing catch up at a rapid pace. Google launched Gemini 2.0 Flash Thinking this week, their own version of a reasoning model similar to OpenAI’s o1. Amazon, besides investing in Anthropic, is also working on its own set of AI models and AI compute cluster. These companies have both the talent and budget to innovate and execute quickly, as they see staying ahead of AI development as a key defensive strategy in maintaining their market share of the wider software market.
I wonder if OpenAI will follow the paths of other first-movers in the software space such as Myspace, Netscape and Yahoo, who all failed to maintain their early market share.
It also seems like the price of these models will eventually converge to near-zero, as it has been in recent times for most software services. Meta’s strategic move to open-source Llama, as well as the open-source nature of AI research and development in general, makes it hard for any one model to justify a pricing premium unless they can perform exceptionally better than others. Will OpenAI be able to continue maintaining it’s monthly subscription fee of $20 in the face of dwindling market share?
What’s next for AI?
Shorter post this week, I suggest reading the full writeup from Menlo here on the state of AI in 2024. One of my 2025’s to-dos is to ramp up the use of AI tools in my daily life and be more familiar with this space. Some other tidbits from Menlo’s report:
AI spending surged more than 6x to $13.8 billion this year, as enterprises shift from experimentation to execution in incorporating AI.
Code copilots lead the charge as the most valuable gen AI use-cases, followed by support chatbots and enterprise search + retrieval.
Enterprises have adopted a multi-model approach instead of relying on a single provider, routing to different models depending on the use case or results.
Agentic automation will drive the next wave of AI transformation, going beyond content generation and knowledge retrieval to tackle multi-step tasks.
Wishing you a Merry Christmas in advance!
-
This Week:
🤔: What are the checks and balances on the power of Elon Musk? - by Noahpinion, in the face of Elon’s involvement in U.S’s spending bill shutdown drama this week.
🇺🇲: Fed cuts rates by 25 basis points but indicates fewer reductions ahead as inflation remains above target
🚗: Waymo announces plans to begin its first international testing in Japan in 2025, to gain experience of driving conditions that it has not previously been exposed to. The Alphabet-owned self-driving taxi company is already providing 150k paid rides per week to customers in the U.S. in four cities.
🇯🇵: Why Tokyo’s proposed 4-day work week may not solve its population woes - by FT