OpenAI's Near Meltdown: The Winners and Losers So Far
It will certainly make the competition more interesting in 2024
OpenAI has not fully recovered from its self-inflicted wounds in the November management fracas. While the fallout could have been worse for the ChatGPT and GPT-4 maker, the company has not emerged without damage, while others received a boost. Here is a quick breakdown of the winners and losers coming out of OpenAI’s meltdown.
Losers from the OpenAI Meltdown
OpenAI
The company was on a path to becoming a tech giant approaching a $100 billion valuation with a dominant market share in enterprise large language models (LLM) and the most popular LLM-fueled chat app. Even though the company brought back key management, implemented a re-constituted board with seasoned managers, and looks like its desired valuation may still be in the cards, OpenAI’s prospects were subtly undermined.
OpenAI rivals immediately received calls about accessing their LLMs. Many also received new injections of outside financing. Numerous enterprises that were on course to move ahead exclusively with OpenAI are now rethinking that strategy and evaluating alternatives. The GPTs store was supposed to launch in late November, but that was moved to 2024. Momentum slowed. The company still leads in market share and mindshare, but the aura of invincibility was tarnished.
The Microsoft backstop through Azure is a key reason why many customers will stick with the platform. However, Microsoft was already starting to build its own models to save costs and recently released Phi. The company is sure to redouble that effort now that it is clear that OpenAI’s unusual management structure can be a liability. Satya Nadella, Microsoft’s CEO, also mentioned Mistral and Llama at the Ignite conference, and the company is working with Databricks on MPT. Microsoft is still committed to OpenAI, but it may not be as much as it was.
Sam Altman
It may seem like Sam Altman’s status was reinforced through the OpenAI’s team solidarity and his many offers to build a new competing business if he didn’t return to the company he co-founded. However, the breakdown in board relations happened on his watch, and there is now going to be more scrutiny about his management practices and conflicts of interest by a more seasoned board. His latitude to act as he likes will be constrained, and OpenAI looks less invincible than it did in early November.
Ilya Sutskever
It is unclear how OpenAI co-founder Ilya Sutskever can continue working with OpenAI. Many people on the team resent his role in nearly blowing up the company. In addition, it seems clear his reservations about Altman’s conflicts of interest or the lack of concern about Q*’s potential impact were not resolved. Reports say he has retained a lawyer and that generally leads to a separation agreement. Elon Musk has said there is a spot for him at X.ai, which could be a good home for him to pursue his interest in AGI. However, it will be a significant step backward in pursuing his goals.
Adam D’Angelo
Adam D’Angelo may seem to have emerged with a similar position post OpenAI’s recalibration. He still sits on the board. However, many people with OpenAI and elsewhere placed much of the blame on him as he was the board member with the most business experience. In addition, he appeared to have a clear conflict of interest with Quora’s Po, looking like a direct competitor to ChatGPT. The former CTO’s industry connections helped him retain his role on OpenAI’s board, but he appeared to burn a lot of bridges along the way.
Software Developers Reliant on OpenAI
Many companies have built solutions on the back of GPT-3.5 or GPT-4. The uncertainty about the company’s governance and whether it will refuse to provide access to future models puts them all at risk. While the company may continue to benefit from fierce competition that is improving LLM performance combined with price declines, they all know they must have a backup foundation model provider. That will complicate their business and product architecture while also diverting resources away from customer growth.
Winners from the OpenAI Meltdown
Anthropic
OpenAI’s nearest LLM rival received a hundred calls from enterprises wanting access to its API during the height of the management shake-up. It subsequently released Claude 2.1, showing closer performance to GPT-4 and including an industry-leading 200k LLM context window. More recently, it has been rumored to be raising $750 million in fresh capital at an $18.4 billion valuation, up over 5x from the beginning of 2023. Anthropic’s position as the leading OpenAI alternative was reaffirmed by the strong enterprise interest and a steeply rising forecast for 2024 revenue.
Amazon
Amazon looked woefully behind its tech giant peers in generative AI offerings six months ago. Even in the second half of this year, OpenAI’s market dominance was leading AWS loyalists to try Azure for the first time. However, the chink in OpenAI’s armor and Amazon’s deal with Anthropic provided an easy way for many enterprises to try out the OpenAI alternative. Amazon Bedrock looks like a stable alternative to OpenAI and Azure that is customer-focused and not subject to the whims of a few engineers or AI safety advocates.
Google was falling further behind OpenAI with the delays in the release of its Gemini LLM. That was partially rectified in December with the Pro Model emerging and a tease of the Ultra Model. While Google would have done much better if Gemini were available in early Q4, enterprise reservations about OpenAI will give it a chance for a fresh look by technology teams.
Mistral
Mistral has been an interesting open-source LLM provider with a claim to cost-efficient performance quality and notable origins in the EU. Shortly after OpenAI’s issues, the company raised a new $415 million funding round at a $2 billion valuation. It also released a new open-source LLM that showed strong promise as a GPT-3.5 alternative. Mistral and Meta (due to its open-source Llama 2) received renewed interest from enterprises considering open-source options that might avoid future conflicts arising from management decisions of proprietary model providers.
Other LLM Providers
While no LLM providers received the level of attention as Anthropic in the midst of OpenAI’s near meltdown, they all received renewed interest. AI21 Labs was able to add $53 million to a recently announced funding round. Cohere received new enterprise interest, and Inflection AI launched a new model that closed the performance gap with GPT-3.5.
Microsoft Azure
Any enterprise that wants to stick with OpenAI’s foundation models will forever think twice about relying solely on OpenAI’s APIs. Azure has its own instance of GPT-3.5 and GPT-4 that will enable it to continue to operate the foundation models regardless of what happens with OpenAI. Many companies had already started using Azure OpenAI Services and others were looking at it. You should expect nearly all of them to migrate to Azure in 2024 to reduce risk.
Satya Nadella
Microsoft CEO Satya Nadella was “grace under pressure” during the OpenAI debacle. He tried to negotiate a resolution with the board. After those initial efforts fell through, Nadella immediately offered Altman and everyone else from OpenAI a new job at Microsoft where they could continue their work together in an independent decision. He even had an office set-up in San Franciso within 48 hours, ready to accommodate a 700-person team. Nadella looked like the most sensible person in the industry and displayed the most strategic and tactical acumen in a difficult and fast-moving environment. In the end, Microsoft ended up with a board observer seat at OpenAI, and Nadella didn’t have to establish an entirely new company.
What to Expect
Management shake-ups and conflicts are not exactly uncommon. Twice during Paypal’s storied start-up years, the entire management team threatened to quit unless the CEO was replaced. Boards also replace CEOs for a variety of reasons. A key difference with OpenAI was its central dominant position in an emerging $1.3 trillion market.
OpenAI remains the top company for LLMs, and ChatGPT has no near rival in terms of adoption. The company has had a month to reset and should come out with some significant updates in the first quarter of 2024, including the GPTs marketplace and maybe a rumored GPT-4.5. The company is still the most likely industry actor to set the tone for generative AI competition.
However, the unforced error by its board did not achieve the board members’ core objectives and has created an opening for competitors to close the gap in technology and customer adoption. The first quarter is a critical time for every industry competitor. Will OpenAI re-assert its dominance? Will Anthropic or Google capitalize on the opening offered by new concerns related to OpenAI? Will open-source options become more attractive?
The next year, and particularly Q1, is bound to be even more interesting in generative AI land than anticipated.