- Mag7’s combined valuation as a percentage of the S&P 500’s total price tag peaked in autumn 2025
- Investors have therefore broadened their search for AI winners
- Range of share price performance within the Mag7 is broad, too, as investors sift for potential winners (and losers)
- Attention seems to have shifted toward near-term beneficiaries of AI spending, and away from those making the biggest investments
- NVIDIA’s quarterly results on 26 August are the next test for the AI narrative
“SpaceX may have gone up like a rocket and come down like a stick, and Korean silicon chip stocks gyrated wildly over the summer, but a number of equity sectors have continued to bask in the glow cast by artificial intelligence, including power generation equipment specialists, real estate plays and even airlines,” says AJ Bell investment director Russ Mould.
“However, investors still seem a little less sure about the so-called Magnificent Seven themselves, judging by how the septet are very gently underperforming what is, admittedly, a rip-roaring S&P 500 index, to suggest right now that markets seem more confident in those companies who are benefiting from the AI investment boom rather than those who are doing the spending.
“Record earnings from Siemens Energy, booming cargo shipments out of Asia for Cathay Pacific and a bid for British land regenerator and developer Harworth, thanks to its data centre sites, and surging share prices for all three show that investors are looking at every angle they can find when it comes to AI, and not just at the biggest, best-known names.
“That said, the aggregate market capitalisation of the Magnificent Seven – Alphabet, Amazon, Apple, Meta Platforms, Microsoft, NVIDIA, and Tesla – is $23.6 trillion, a fraction below the $24 trillion all-time high of May.
“This figure represents 35.7% of the S&P 500 benchmark index’s $66.3 trillion price tag, but it is down from the 37.9% peak seen in October 2025. In plain English, the value of the S&P 500’s other 493 members has gone up faster since last autumn, so the Mag7 have underperformed.
“Some investors may not be too distressed, as they cannot pay the bills with relative performance, only absolute returns, and the latter will still keep many happy and in the manner to which they are accustomed. The Mag7’s total market cap is up by 23% over the past year, or some $4.5 trillion, and $2.8 trillion of that has been added since late June alone.
Source: LSEG Refinitiv data
“But even within the Mag7 there has been a wide range of performance. Alphabet is easily the best performer over the last year, while Tesla, Microsoft and Meta Platforms are all down. Slashing cuts to earnings forecasts are hurting Tesla, while Microsoft and Meta’s ever-increasing capital spending budgets continue to nag away at investors as they ponder when, or even whether, the companies will start to generate an adequate return on the spending.
“Microsoft at least has its cloud business, which is positioned to be a winner from the ongoing build-out of the data centres and large language models that power AI.
“This is one of Alphabet’s trump cards, too, although the owner and operator of Gmail, Google maps, Google search, Google Drive, YouTube, Chrome and Android has the additional advantage of huge customer bases that provide data which in turn helps to train the large language models and customise the AI offer to best effect.
Source: LSEG Refinitiv data
“Meta is seen as having the weakest cloud offering, and thus the least direct route to benefiting from AI spending now, while its spending drive means share buybacks are no more, the net cash pile is shrivelling, and free cash flow is negative. Meta may feel it has no option other than to spend, given the potentially existential threat posed to its business model by Claude, ChatGPT and Gemini, let alone rival Chinese AI offerings such as DeepSeek and Kimi K3, but shareholders seem nervous as to when the returns may start to flow.
“Perhaps memories of the near-$90 billion splurge on the Metaverse earlier this decade remain fresh in the mind, given how that initiative was scrapped and a sharp cost-cutting programme helped to get Meta’s share price back on track in 2022. Even so, the spending now way outstrips anything Meta laid out back then.
Source: Company accounts, Marketscreener, analysts’ consensus forecasts for Alphabet, Amazon, Meta Platforms, Microsoft, and Oracle
“Meta’s shares were not the only ones of the Mag7 to flounder then, as interest rates started to go up, and bulls of the stocks may be tempted to look to how valuations now compare with those from back then.
Source: Company accounts, Marketscreener, analysts’ consensus forecasts, LSEG Refinitiv data
“Microsoft emerges as the one that is most interesting on that basis but it still faces the same challenges as Meta, Amazon and Alphabet in that their aggressive spending means free cash flow is drying up, fixed assets are piling up on their balance sheets so fast they are now showing the sort of incremental capital intensity that would make a steelmaker blush and they are starting to compete with each other more actively. All three of those chip away at key planks of the historic investment case for each, namely scalable and asset-light business models, copious free cash flow, and dominant competitive positions in their chosen market.
“Ten years ago, the Mag7’s quarterly capex-to-sales ratio was just 7.2%, while tangible fixed assets represented 19.5% of sales. After the latest results from Alphabet, Amazon, Apple, Meta, Microsoft and Tesla, those figures now stand at 26.6% and 59.1%.
Source: Company accounts for Alphabet, Amazon, Apple, Meta, Microsoft, NVIDIA, and Tesla. NVIDIA is yet to report Q2 2026 earnings.
“All of this raises the stakes for the final set of quarterly results from the septet, when NVIDIA reports on 26 August.
“The graphics processing unit (GPU) specialist is one of the more immediate beneficiaries of hyperscaler spending, and NVIDIA is doing so well that it can afford to help finance some of the investment planned by its customers, and its customers’ customers. This does not sit easily with all investors, who remember how vendor financing eventually ensnared Lucent, Nortel, Cisco and others when the technology, media and telecoms bubble burst in 2000, but this model worked for GE and GE Finance for a very long time (until it didn’t).”
Source: Company accounts. Financial year to January.
