AI Newsround – SoftBank and Meta Platforms

SoftBank – Storm in a teacup.

• The news that SoftBank has sold its stake in Nvidia has not been well received, as the reality is that SoftBank has exited a relatively safe AI investment and bet it all on a proposition where the outcome is very uncertain.
• Early trading in Tokyo put SoftBank shares down 10% although they have since recovered to around 2.5% down, which in my opinion makes today a meaningless blip.
• SoftBank reported excellent results, but a large proportion of $16.2bn net profit was driven by unrealised gains on investments, which history shows can evaporate in a heartbeat.
• However, the main story was the sale of its entire $5.8bn stake in Nvidia to fund the $30bn it is putting into OpenAI at the current $500bn valuation.
• With a single transaction, SoftBank has greatly increased its risk profile, although if Mr Altman is correct, there is probably more upside in OpenAI than there is in Nvidia.
• However, on a risk-adjusted basis, this is a trade I would not make.
• OpenAI is asking its investors to believe that it will somehow be able to make a positive return on its investments when the financial realities are difficult to fathom.
• If we assume that the average depreciation of this investment is 7 years (most are saying shorter), then the depreciation charge will be $200bn alone every year.
• If OpenAI makes an operating margin before depreciation of 60%, then it will need revenues of $333bn simply to break even and revenues of $600bn to make a real operating profit of 17%.
• These are staggering numbers, which is why the path to these kinds of revenues needs to be meticulously supported by detailed analysis rather than being taken on faith.
• This is why I think that OpenAI is a very risky proposition compared to Nvidia, which utterly dominated its space and does a fantastic job at monetising the opportunity for the benefit of its shareholders.
• Hence, I think the share price of SoftBank should go down more to reflect the huge increase in risk that SoftBank is taking on by moving money from Nvidia to OpenAI.

Meta Platforms – Commercial Pressure.

• One of Meta’s leading lights is moving on in a sign that the emphasis at Meta is on catching up with its rivals as opposed to actually conquering the problem of super-intelligent machines.
• It looks like Yann LeCun, one of the fathers of AI, is moving on and will create a start-up of his own that will continue the quest of seeking other methods to achieve machine superintelligence.
• I can’t say I blame him after having the indignity of reporting to a 28-year-old, even if he is a visionary genius who will turn Meta’s AI efforts around.
• This represents a shift for Meta as Yan LeCun, who is a recent recruit to the more sceptical side of the can-LLMs-become-superintelligent debate, has been looking at methods other than LLMs to take Meta research forward.
• Mr Zuckerberg’s actions since the disappointing release of Llama 4 and the loss of its leadership in the open-source community have been to move towards his rivals rather than doing something different.
• This explains the difference of opinion between him and Yann LeCun and why Prof. LeCun has decided to move on.
• Meta’s pivot towards larger LLMs and more compute explains the hike in capex revealed at the last set of results and is by no means guaranteed to be successful.
• Unlike OpenAI, Meta does not have a direct way of earning a return on the $100bn+ it will invest in capex next year, which is why its capex bill has been met with such negativity compared to everyone else.
• The net result is that Meta is now in a straight fight with everyone else for supremacy in AI, and although it is far better at AI than it was 5 years ago, it now has some ground to make up.
• Against that backdrop, Google appears to be in better condition, and the shares are more attractively valued.
• Hence, I would have Google over Meta if I had to choose.