MSFT, AWS, GOOGL, META, OpenAI and More...
MSFT: Microsoft needs to build out its proprietary LLMs and multi-modal models if it wishes to compete with OpenAI in the GenAI space. Microsoft seems to be taking the small language model approach as it relates to this effort.
My view is that META’s open source Llama models will be a formidable competitor to all smaller language models given its relative transparency and cost advantages over proprietary models. META is driving more LLM innovation than MSFT as is Google.
OpenAI and Anthropic are cooler places for data scientists and engineers to work than Microsoft as OpenAI and Anthropic are driving LLM innovation. Microsoft on the other hand has played more of a support role to date.
OpenAI: We wrote about Strawberry yesterday. While the LLM king continues to drive model innovation, it must strike a balance between LLM innovation with prospective investments in proprietary chips (in competition with NVIDIA) and its cloud services business (to better compete with AWS, Azure and GCP).
AWS: AMZN’s cloud services unit is in a similar boat to Azure/Microsoft. AWS has partnered with Anthropic for LLMs. AWS is also building proprietary models. AWS will face a similar challenge in that the best talent wants to be with Anthropic, OpenAI, Mistral and Groq, not necessarily the incumbent firms.
GOOGL: Google is a bit unique in that it has a proprietary cloud services business (Google Cloud Platform / Google Cloud) and proprietary LLMs built in part with its proprietary TPU chips (Google also uses NVIDIA chips and Anthropic LLMs). We ought to see advanced agents (Level 2) come out of Google in 2025 based on what Google’s AI employees have said publicly.
META: I am a fan of Meta’s open source LLM approach. I see the Gen AI space bifurcating between high-end LLMs such as OpenAI’s GPT family and smaller models with inexpensive chips that will be more than sufficient for most real-world use cases (think of Meta’s Llama in combination with Groq’s inference capability). We recently wrote about Meta’s open source LLM approach HERE.
Anthropic: Anthropic seems to be running neck-and-neck with OpenAI, although we will see what kind of distance Strawberry may put between OpenAI and everyone else’s LLM portfolio later this week. Anthropic appears far more stable than OpenAI within the executive management tier.
Gen AI will drive meaningful ROI when AI models reach Level 3 status (below).
