Episode Details

Warsh was exceptionally hawkish at the Jackson Hole symposium last week suggesting that the “Fed still has work to do”. Still, we don’t believe that the FOMC will raise in Sept. There are supporting economic factors including slowing GDP & slowing employment but more likely, so close to the mid-terms, it’s unlikely that  the Fed will want to appear to be choosing sides. If a raise were to occur, the timing is would be year-end or Q1 '27. Whilst most economists assume the next move is a hike, the currency markets don’t appear to be supporting this given the DXY remains 3-4% below its June highs.

Employment has not grown over the last 5-6 quarters in the US and the latest GDP figures for Q2 suggest a significant decline in momentum falling from 2.1% in Q1 to 1.5% in Q2. The only supporting indicator for a rise in rates remains inflation. Short-term, the PCE indicator most favoured by the Fed (out this week) will not read well. Inflation is likely to persist above 3%, well off the target rate of 2%, thanks largely to energy market disruption. However, moving forward, energy prices post the mid-terms are likely to recede as Iran will have achieved its objective of providing external interference and productivity gains from the AI capex spend will begin to show through in keeping prices stable whilst reducing costs for producers and manufacturers.

Nvidia blew past analysts forecasts in their announcement last week with a 70% revenue growth guidance vs 45% forecast. As an example of the pace of growth, their digital agents number 400k pushing towards 4mn working 24/7 over the next year. Compare this to their human employees who number 40k. Management suggest an average datacentre is breaking even in less than a year and sees a $2trn industry backlog and growing. Customer base expanding beyond just hyperscalers to governments and large corporates keen to build their own AI capabilities. The stock is trading on a forward PE of 21x with 25% earnings growth consensus forecast. As the stock in the centre of all things AI, this looks like a very reasonable valuation.

The Meta strategy for its build-out could well turn out to be the best risk adjusted mechanism of all and bring the company closer to Free cashflow generation before its peers. 

Meta decided to go down the route of leasing rather than owning and thus becoming an asset heavy business when it comes to data centre AI capacity. The mechanism is simple and neat. Meta invests in 20% of an SPV with the balance owned by major financial institutions including Blackrock and Apollo. The SPV then raises debt in the market from investors to pay for the DC land and assets. 

Meta is the leaseholder and signs rolling 4 year contracts with the SPV. The Meta liability becomes its 20% equity holding and its contractual obligations. The financial institutions and bond holders take on the financial exposure with no debt showing on the Meta balance sheet. The bond holders have no recourse to Meta but rather the SPV. So this is not an off balance sheet risk for Meta. 

The bond investors enjoy a long-term income on a long-lease from a blue chip covenant but have to accept that at the end of the lease life, they will end up with a highly specialised asset which may be difficult to monetise. 

In case of demand stalling for AI buildout capacity, Meta would still need to meet its contractual obligations but does not face other financial penalties bar its 20% stake in the SPV. 

Meta turned up late for the DataCentre party but may end up being an early beneficiary. At 18x forward PE, a substantial discount to the S&P500 Index and a consensus revenue forecast of 33% for next year, Meta looks increasingly like a value play. 

We’ll cover asset allocation further next week but as a starter, our winners over the summer have been Cybersecurity, Energy, Healthcare, Biotech & Crypto. Whilst the losers have come in with Space, Chip