Meta Platforms – Sunshine

Make hay while the sun shines.  

• Meta is likely to report more of the same at its Q2 26 results on the 29^th of July. Still, there is growing evidence that it will unveil a new business of selling compute to 3^rd parties given the very large returns that are currently on offer.
• This has already begun with the launch of a paid API for Muse Spark 1.1, which promises almost frontier performance but at 25% of the price.
• Q1 26 saw excellent financial performance, but the lack of clarity in terms of how a return would be earned on $135bn of capex in 2026 sent the shares down 10%, where they have largely lingered for the last three months.  
• This capex makes up the vast majority of Meta’s cash flow, which has caused concern given that Meta to date has had no mechanism by which to monetise the investment.
• Promises of higher growth in its other businesses and better profitability have largely fallen on deaf ears, which is why the shares have performed so badly.
• At the same time, the market for compute has gone ballistic with companies like Anthropic, Google and ReflectionAI being willing to pay $37bn – $56bn per GW to access the capacity that xAI decided to make available.
• RFM Research has concluded that while these deals are in force, xAI will be making a 60%+ return on investment even after spending $40bn – $60bn per GW to build the capacity in the first place.
• This will not last forever as capacity is being built at breakneck pace, but for the next 12-months at least, supply looks certain to remain very tight.
• These returns are far higher than Meta could reasonably expect to deliver to its shareholders through its Family of Apps business, and so it, too, appears to be looking at addressing this market.
• This is not a sign of overcapacity, as Mr Zuckerberg said in an interview yesterday that he was using all of the compute that he had built, but I see it as a purely rational financial decision.
• Selling API access is a start, and here Meta is clearly going for the mass market where developers don’t need the best model available for their agents, which I suspect is going to be the vast majority of volume in the market.
• This will not earn Meta the big returns as the market is demanding that it simply rent the compute capacity to build and run their own models.
• Meta has hinted at this, and I am hopeful that something along the lines of what xAI has announced will be revealed at or before the Q2 26 results.
• The recent gyrations of the share price clearly indicate that this is what the market wants to hear, and I hope that financial rationality trumps world AI domination in Mr Zuckerberg’s mind.
• This, of course, is the risk with Meta Platforms, as anyone who comes along for the ride remains completely dependent on what Mr Zuckerberg decides to do, as he has unilateral control of this company.
• In recent years, Mr Zuckerberg has become more sensitive to what the market demands as the “year of efficiency”, and the recent job cuts clearly indicate.
• Hence, I think that there is a better than 50% chance that something along the lines of xAI will be announced at or before the next set of results.
• This announcement would serve as both a catalyst and an addition to fundamental value, as renting out capacity over 5 years should add around $70 per share to the value of the shares, which I currently estimate at $706.
• I have opened a long position in Meta and am looking for $700-$760 over a 9-12 month period.