The OpenAI saga definitely isn’t over. As reported first by The Information, then by Reuters, the company’s in-the-works enlarged board will not have any seats for main sponsor Microsoft, nor for other investors such as Khosla Ventures, Sequoia Capital, or Thrive Capital.
When CEO Sam Altman was reinstalled last week after five days of tumultuous exile, The Verge reported that Microsoft—which owns 49% of the nonprofit’s profit-seeking subsidiary—and Altman both wanted seats on the enlarged board, which will apparently comprise nine people. Wanting isn’t the same as getting, though.
“Microsoft will have something to say about it, given the amount of money that they have put behind them,” Great Hill Capital chair Thomas Hayes told Reuters. No kidding, given the deeply unwelcome surprise that was the previous board’s firing of Altman—which, incidentally, still remains unexplained. Altman’s reinstatement largely returned OpenAI to stability, but Microsoft was surely hoping its $10 billion would give it protection from future surprises. “We will wait until the board officially says something,” added a spokesperson for Microsoft itself.
The Information reckons keeping investors off the board will signal the prioritization of safety over revenue, which would be in line with the thinking behind OpenAI’s founding purpose and weird corporate structure.
On one level, that would not be a bad thing—if you believe there’s a chance that artificial general intelligence will a) happen anytime soon and b) have the potential to threaten humanity, then OpenAI’s original incentives are surely worth holding on to. However, it would signal a continuation of the tension and uncertainty that led to the company’s crisis this month, which could annoy Microsoft but also, due to OpenAI’s ongoing need for Microsoft moolah, leave OpenAI’s customers wondering if they’re backing the right horse.
The company’s rivals aren’t missing out on the opportunity this uncertainty affords them. “You need not look any further than the events of the past 10 days to understand how there will not be one model to rule them all,” Amazon Web Services CEO Adam Selipsky told Wired ahead of yesterday’s announcement of Q, AWS’s business-oriented chatbot, which is powered by a variety of large language models from Cohere, Anthropic, and Amazon itself. (If OpenAI really is working on a new AI model code-named Q*, which reportedly may have played a role in Altman’s ouster, it should probably now choose a different name for the commercial launch.)
Interestingly, AWS uses some secret sauce to decide which underlying LLM should answer each query, unless the customer has explicitly chosen one. Selipsky said leaving the decision up to AWS is “more cost-efficient, but also flat-out more effective.”
Speaking of AWS and the announcements at its re:Invent conference yesterday, the company had quite a lot of big chip news. It unveiled new generations of its in-house processors for training neural networks (Trainium2) and running more traditional workloads (Graviton4), and extended its Nvidia partnership so AWS customers can tap into 32-strong bundles of team green’s shiny new GH200 Grace Hopper Superchips—AWS and Nvidia are even working together on designing the “world’s fastest GPU-powered AI supercomputer,” which will comprise a whopping 16,384 of those superchips, amounting to performance in the order of 65 exaflops.
Perhaps most intriguingly, AWS utility computing chief Peter DeSantis announced in his keynote that the Amazon unit has come up with a new quantum chip that’s very good at suppressing certain kinds of errors that have been holding back the quantum-computing field. Don’t expect the dawn of the quantum era just yet, though—we can all keep focusing on the chaotic AI revolution for now.
More news below. Also, check out the inaugural edition of Fortune's new CEO Weekly Europe newsletter, which launched earlier today.
David Meyer
Want to send thoughts or suggestions to Data Sheet? Drop a line here.
