Chat-GCP

Key Capital Private, Investment Note - Issue #34

Chat-GDP

Modern Portfolio Theory (MPT) is the framework which underpins portfolio construction at basically every major asset manager. It was developed by Harry Markowitz and, without getting into the mathematics, boils down to saying that diversification – not putting all your eggs in one basket – not only reduces risk, it enhances returns. This is a unique combination of characteristics; normally, to reduce risk, one must accept lower returns. “Diversification is the only free lunch in investing” was his famous (in investing circles) quote. This was an excellent observation by Dr Markowitz, so excellent, it was recognised with a Nobel Prize.

The fundamental assumption underpinning MPT is that different securities behave differently – a pretty modest assumption – and history backs this up, stocks mostly behave differently than bonds, technology stocks behave differently than utility stocks. Importantly, the relationships between different securities are not fixed; they tend to ebb depending on the overall macro picture. During crises, everything tends to move in the same way. COVID is a good example of this. During the first two weeks of March 2020, tech stocks, which would become major Covid winners, sold off alongside Covid losers. But over a longer timeframe, the fundamentally different characteristics of different businesses give investors their free lunch.

An AI Stock Market

The past two years have delivered one of the most powerful technological booms in decades; AI has been injected into seemingly every facet of society; this is especially true for financial markets. The ‘Magnificent Seven’ big tech companies have eaten the US equity market – these seven companies now make up 30% of the S&P 500 and have generally moved as one AI trade. According to JP Morgan, AI stocks have made up 75% of S&P 500 returns, 80% of earnings growth and 90% of capital expenditures since ChatGPT's release.

That capex has risen to such a level that it is distorting economy-wide statistics. AI spending is now 1.2% of US GDP and, per Reuters, was two-thirds of GDP growth for the first half of 2025. You have to go back to the 1800s US railroad boom for a precedent. Relative to GDP, current AI spending remains below the peak reached in the railroad buildout (1.2% vs 6% of GDP), but railroad builders weren’t planning on rebuilding their tracks all over again in a few years, while AI chips are obsolete within three years. Adjust for that mismatch and the AI buildout becomes genuinely unprecedented.

The problem for portfolios is that this Cambrian explosion of capex is being spent in the rest of the economy, hitching other sectors to the AI wagon.

Data centres are the main driver of this AI related spend. In the US, construction spending on data centres has already surpassed that of office buildings, connecting the usually diversifying real estate sector to the AI economy. Data centres also need power. A lot of power. Last year 70% of the increased electricity demand in the US came from data centres – add the utility sector to the AI trade basket. They also need electrical equipment from industrial companies, raw materials like copper from mining companies and even fibre optic cables from telecom companies. All now living by the sword of continued AI spending.

Also, an Increasingly AI Bond Market

Bonds, the traditional ballast of portfolios, have not avoided the AI explosion either; the capital needs to be raised from somewhere. A Meta-adjacent company in October issued the largest-ever US corporate bond with $27bn to help fund one(!) of Meta’s data centres. Bond funds piled into it; the offering had $125bn of bids. Alphabet had a $17.5bn bond offering hit the market in November, with an order book of $90bn. Other big tech companies are partnering with large asset managers to issue all sorts of increasingly complex debt securities. The NYT recently reported that Blackstone is packaging ten data centre leases into a $3.5bn commercial mortgage-backed security (CMBS), modest compared to some of the numbers thrown around in AI circles, but it is unprecedented for the CMBS market, where issuance for data-centre-backed deals was just $3 billion for all of 2024.

Private Markets

Private markets offer little relief. Private credit is also helping fund the AI buildout. Morgan Stanley suggests that $800 billion in private credit will be needed over the next two years to fund data centres. Bloomberg Intelligence estimate private equity funding for private AI companies like OpenAI and Anthropic at $150bn in 2025 alone – a 6x increase on 2024. Private equity-backed deals have also accounted for up to 90% of data centre transaction value since 2022, according to Synergy.

Summary

The 60/40 stock-bond portfolio emerged as the practical implementation of Harry Markowitz’s work. It has been wildly successful and forms the basis of most portfolios today as an easy way for investors to have their free lunch. But AI is tying markets together in surprising ways; it has increased concentration at the top of indices, linked previously unrelated sectors and strategies, and encouraged investors to respond to the same signals. If the AI boom cools, a lot of things may cool at the same time.

For investors, ensuring that portfolios are actually diversified will require a more thoughtful approach than the 60/40 going forward, so that they are not just making one giant trade on the continued expansion of the data centre industrial complex.

Which, to be fair, has been a great trade. Until it isn’t.

Postscript

As we finished this note, NYU professor Aswath Damodaran, described by the FT as “probably the world’s pre-eminent academic expert on equity valuation”, sat down with Scott Galloway to discuss markets and left us with the following: “We live in a world where everything seems to be correlated ... correlation has risen to the point where the classic rule of [diversification across sectors and geographies] doesn’t apply ."

Key Capital & Schroders Event Video:

Key Capital and Schroders recently hosted an event on the Key Capital Balanced Multi Strategy (BMS) Fund.

• BMS Overview: Ian Kilcullen, Managing Director of Key Capital Private, gives an overview of the BMS Fund and how it differs from traditional multi asset funds. - 14 minutes.

• Panel Discussion: A discussion with Rishi Sivakumar, BMS Fund Manager at Schroders, and David Rees, Head of Global Economics at Schroders, on the diversification illusion at the core of most multi-asset funds and how the BMS Fund seeks to address this for Irish private clients. - 14 minutes.