Is OpenAI the Next Netscape?

This weekend was one to remember for those actively invested in the future of AI. In this special edition of B2BaCEO, I’ll summarize what unfolded (as of Monday, November 20 at 5 pm PT). I’ll offer perspective on the future of OpenAI and its diminished status as the public face of generative AI. I’ll also build on my recent editorial about the opportunities for startups, together with what this all means for other model providers like Google and Anthropic, the open-source community, and more. 

This remains an extremely fluid situation. This morning, OpenAI’s Chief Scientist Ilya Susketver sent a Tweet that opens the door to the clock being reversed. But for now, here’s what I’m thinking.

What’s Next?

1. Implications for OpenAI
   The weekend’s events have caused significant unrest across the tech industry—and especially within OpenAI. Reports indicate that most of the company’s 700+ employees have signed a petition threatening to jump ship for Microsoft or another competitor unless Altman is reinstated as CEO. This talent shift could dramatically slow OpenAI’s pace of innovation, reducing its standing from one of the world’s most important companies to just another player in the AI revolution. 
   
   The executive turmoil may also impact many of the product launches announced at their inaugural developer conference earlier this month, including the GPT Store. All the buzz around OpenAI triggering a “startup mass extinction event” could be significantly interrupted as the company works to transition their new outsider CEO and recover from the fallout of the past 72 hours. 
   
   
2. Microsoft’s newfound path to dominance
   Though initially blindsided by the news, Nadella quickly jumped in and made efforts to reinstate Altman and Brockman. Once it was clear that was a lost cause, Nadella changed course and instead recruited both (and now potentially Murati as well) to join Microsoft to lead a new advanced AI research team. Because Microsoft has an exclusive perpetual license to all OpenAI IP, which includes source code and model weights, they have the ability to exploit that IP—and, in fact, make it bigger and better.
   
   As I mentioned in my editorial on Friday, Microsoft recently announced that it is integrating Copilot, its AI assistant, across its product suite. Analysts predict that Copilot could generate $14B in annual revenue. This new AI research team will supercharge Copilot’s development, putting Microsoft in a dominant position and lending momentum to other incumbents.
   
   
3. Implications for other model providers
   As OpenAI works to recover from this weekend and regain both internal and external confidence, what was once a formidable barrier to entry will now lift for folks like Google’s Gemini and Anthropic. While Gemini is thought to have capabilities on par with (and potentially even exceeding) GPT-4, its launch had been pushed from this month to Q1 2024—due, in part, to ChatGPT’s strength as a rival. 
   
   Even before Altman’s firing, all was not perfect at OpenAI. Some of its largest enterprise customers, including Morgan Stanley, were increasingly relying on Azure’s OpenAI service, which provides access to OpenAI’s models with the privacy, governance, and security safeguards of Azure. Over the past 72 hours, other rivals including Anthropic, Cohere, and AWS have been quick to capitalize on the situation and acquire OpenAI customers both large and small.
   
   As alternatives to OpenAI rise in importance, the situation starts to recall the “cloud wars” of years past. Customers know to avoid the same mistake of committing to a single sticky vendor, given the risk of lock-in and high switching costs down the line. 
   
   The open-source community is another likely winner, as customers continue to realize the risk of dependence on closed providers. Expect to see increased adoption of companies like Databricks, equipped with improved open-source model training and inference offerings as a result of the MosaicML acquisition, as well as Meta’s next-gen open-source Llama 2 in the coming months.
   
   This same risk will make room for AI middleware startups like LlamaIndex and LangChain to thrive. Larger customers with the right resources and talent can use these tools to create customizable chatbots with models that can be easily swapped in and out. In parallel, companies like Arize with its third-party model evaluation and monitoring tools will become increasingly more important. (For more on the generative AI stack, see our FMOps post from this March.)