Meta Platforms – Sunshine pt. III

Meta makes hay while the sun shines.

• Meta has launched an offering to sell AI compute and services to enterprises, and in the short term, this is likely to be highly profitable.
• However, as prices for AI compute normalise (as they surely must), this proposition will become much less interesting, and I suspect will eventually be fairly insignificant.
• Meta has launched Meta Enterprise Platform, which will offer AI compute, the Muse agent, Meta Business Agent, Muse API, and Muse Code to businesses and developers.
• This is nothing more than an expression of market forces, as the prices currently being paid in the spot market for compute are so high that 5-year returns are north of 50% per year and, in some cases, above 80%.
• This means that anyone with capacity has a very strong incentive to park whatever it is they are doing and sell their capacity to the highest bidder for as long as the current situation lasts.
• This is exactly what xAI has done, what most of the cryptocurrency industry is doing and what Meta has now announced that it will also undertake.
• Meta is dressing this up as a business that will also sell all parts of the AI stack, but I suspect that most of the current interest that it has is simply for the compute capacity itself, as this is where almost all the demand is currently coming from.
• Hence, I think that the first deal that it announces will be with an unnamed frontier company (i.e. Anthropic) where it rents something in the order of 500MW for which it gets paid around $1.5bn per month.
• Demand from enterprises for its AI tools may follow over time, but for the moment, I think that this business will simply be megawatts and nothing more.
• Meta needs to strike this deal as soon as it can, as every quarter that goes by without the capacity that it has being rented is a lost opportunity, and this opportunity is not going to last forever.
• In any commodity business, the cure for high prices is high prices, which is why everyone is selling everything they have and building more as quickly as they possibly can.
• The numbers are excellent, as 5GW rented at $37.7bn per GW is worth $100 per share in incremental value ($245.4bn) on a 5-year basis.
• If this were to become a permanent part of Meta’s financial performance and last into perpetuity, then 5GW is worth far more per share, but this is not a realistic proposition.
• Instead, prices are going to fall back as more supply comes online to meet demand, and RFM Research estimates that this will settle at around $20bn – $25bn per GW.
• At these sorts of levels, the 5-year business is worth $70 per share and a permanent business is worth around $300 per share.
• However, given how good Meta Platforms is at monetising user activities, I suspect that the returns on offer from selling to enterprises will fall below the return of keeping it in-house, and so I don’t think this business is going to become an engine of long-term growth.
• It will make some money on the side selling Meta’s AI models, agents and tools, but this is going to be small compared to what is available now.
• This is why I see Meta as a short-term opportunity, as I don’t think that the long-term incremental value expected in this business will materialise.
• I have a position in Meta Platforms where I am looking for $770 per share, which it very briefly touched last week before falling back.
• Mr Zuckerberg has launched his proposition, and now he needs to announce a deal.