Hidden behind the incredibly boring term Board Governance lurks future Shakespearean or Machiavellian dramas of amazing proportions. We rarely think about it that way, but for anyone who followed the tech press, or for that matter, general business press last weekend, it was impossible NOT to learn that OpenAI in the span of four days had four different CEO’s, or I should say three, as Sam Altman was reinstated.

It was all about Board Governance. Or lack thereof.

You probably also learned that OpenAI has a peculiar structure whereby a non-profit, four-member board, presides over a for profit company where investors have return on investment capped at 100x.

Having invested in close to 120 companies since 2005 we have at NGP seen our fair share of dysfunction at the level of the board, but nothing like what transpired last weekend. Without a doubt the members of the board that presided over the decision to oust CEO and co-founder Sam Altman are intelligent people, from afar however, it did appear that they misjudged the situation. Smart people also make dumb mistakes.

We don’t (yet) know what really triggered the events, (‘irreconcilable differences’ was the direct quote from the board), various quoted reasons go something like:

• Disagreement over the strategy of OpenAI and questions as to what speed new versions of ChatGPT should be opened to the public.
• Disagreement over how close OpenAI is to AGI like performance.
• Sam Altman wanting to acquire or start a chip company.
• Sam Altman planning a smart device company with Softbank and Jony Ive
• An imminent $1B secondary share sale at a reported $86 billion valuation

Could it be that Sam Altman showed his own lack of judgement? Absolutely. Having previously ousted Elon Musk from the company he is no stranger to conflict or taking on rich/famous people with large egos.

What about the money? I may be a cynic, or tainted by my occupation, but I would follow the money to get to the bottom of the story.

• About $1B worth of shares were to be sold by employees. Surely for many of these employees that could be a life changing event, however, it may have rubbed the ‘non-profit’ motive the wrong way.
• Running ChatGPT is expensive. Investment bank Bernstein’s analyst Stacy Rasgon has calculated that each query costs roughly 4 cents. It has been quoted that running ChatGPT cost around $700,000 daily, meaning about $255M per annum. In 2022 OpenAI has been quoted as having spent $545M. According to Reuters, should ChatGPT queries grow to a tenth the scale of Google search, it would require roughly $48.1 billion worth of GPUs initially and about $16 billion worth of chips a year to keep operational.
• Microsoft has reportedly invested $13 billion in the company, however, a) the investment is in tranches and b) payment is also ‘in kind’ in the form of Azure compute, chips, as well as tech resources.
• OpenAI is quoted as being on a path to generate $1.3 billion in revenue in 2023.

Maybe I digress on the money side as there seem to be little constraint, at least up until last weekend, on the company’s ability to raise financing. Maybe the board suddenly got scared of the amount of money required to keep the company afloat. What about the role of Microsoft?

Microsoft

Even Microsoft may run out of patience. When your turnover in the last 12 months (till Q3–2023) was $212 Billion, and you have $144 Billion cash on hand what has been invested so far in OpenAI is pocket change. However, sorting out OpenAI to secure a large part of your future and stay ahead in the AI arms race is probably the most important job for Satya Nadella right now. He was quick out of the box giving an interview (Sunday) with famous tech journalist Kara Swisher. In the interview he was patiently talking through the importance of AI and the importance to Microsoft products. He was busy emphasising in his words that Microsoft could happily continue without OpenAI given the agreements in place between the two companies.

At the time of investing in OpenAi, Microsoft said: ‘Microsoft is integrating the technology into its Bing search engine, sales and marketing software, GitHub coding tools, Microsoft 365 productivity bundle and Azure cloud.’ I bet that even Mr Nadella might have had difficulties sleeping over the weekend given the dependence on OpenAI. Let’s just say that the relationship is so material that it thrust Microsoft to the forefront of not only applying AI, but suddenly Microsoft/OpenAI became a threat to Google’s core search business. It was suddenly both perfectly viable and even obvious to everyone. How much is that worth? Making Google look like a deer caught in the headlight.

According to press reports Microsoft reportedly owns 49% of OpenAI, while other investors and employees control 49%, with 2% owned by OpenAI’s nonprofit parent. This has never been denied nor confirmed by Microsoft.

Competition

Alphabet, Meta, Antropic, Hugging Face and many others could hardly believe their luck witnessing the fall-out over last weekend. On one hand it provided massive PR for ChatGPT, on the other hand if you are a developer, a large corporate user, or integrating AI in your core workflow, you had pause for thought as to how much money and effort you would spend in building dependency on OpenAI. LLM’s are one way of using AI and ML tools, however, there are many other ways to deploy and use AI and ML. We are witnessing and explosion of companies and tools seeking to make it faster, simpler, easier, and more transparent. Although OpenAI is the ‘media darling’ and Microsoft has stolen the thunder, it would not be the first time that an early mover lost the lead regarding important technology shifts.

Also, despite the name, we should not forget that there is nothing ‘open’ in OpenAI. Everything happens in a hidden black box kept within the walls of the company. And if the AGI debate takes most of the headlines, the real important debate should be between open versus closed models.

Board Governance

And now to the real issue of my post. A crisis is often a time where board resolve, and organisation is tested. You can not plan for everything as a company nor as a board, but a certain paranoia is required to be effective as a board member. Management is often heads down dealing with the onslaught of daily challenges, and boards, when they are well functioning, should help management ‘look around corners’ and put in place plans of action for the known but rarely occurring risks such as a cyber-attack, employee harassment, misconduct, litigation, accidents, whistleblowing and more. There is a long list of items and depending on what business you are in they may have higher or lower likelihood of materialising. Good board governance includes a ‘risk catalogue’ that should be reviewed at regular intervals. Such a risk catalogue should also include which stakeholders are affected and what actions should be taken vis-à-vis affected stakeholder groups.

Clearly the former board of OpenAI fumbled whatever conflict they were trying to resolve, and from the outside it was all down to stakeholder management. They failed managing the relationship with at least five core stakeholder groups: management, employees, investors, partners, and media.

• Whatever the cause of the rift was, it is astounding that the board would communicate to a 49% shareholder that the CEO would be let go only minutes before the public announcement went out and apparently without any consultation.
• The same 49% shareholder is also the most important partner of OpenAI.
• Naturally you cannot ask a wide range of employees whether you should let go of a CEO, but the apparent misjudgement of the wide support Sam Altman enjoyed among employees was astonishing.
• The board members seemed each on their own being busy on X, following up and attempting to defend their actions in various posts rather than having one single spokesperson.

New technology may at first appear magical and wonderful, but over time deliver, as economists would put it; negative externalities, or it could be called unintended negative consequences. It is impossible to know apriori the medium to long term impact of our inventions. OpenAI occupies a disproportionate share of the AI debate at present, making the mismanagement by the old board even worse. A company with a dominant public profile, at the centre of the most important new technology for decades! What confidence did the fallout instil in the public? What will it do to politicians and regulators?