This spring, a peculiar silence has descended upon the trading floors and earnings calls of the largest tech companies in the United States. They are still making headlines, printing money, and announcing capital plans that sound like a mid-sized nation’s budget, so it’s not exactly silence. However, the swagger has diminished. For nearly three years, the Magnificent Seven pulled the entire S&P 500 up the mountain like a group of well-trained sled dogs. However, they are no longer pulling in the same direction. A few are running. A few are hobbling. And the market, keeping a close eye on things, is beginning to take notice.
The numbers appear to support investors’ belief that the AI story is still intact. Together, the four hyperscalers—Microsoft, Amazon, Meta, and Alphabet—increased their 2026 capital expenditure guidance to about $725 billion from previous projections that were closer to $670 billion. Citing increased component prices and unknown data center expenses, Meta added $10 billion at both ends of its range. Alphabet increased its estimate from $91 billion the previous year to $180 to $190 billion. $190 billion was confirmed by Microsoft. This scale is really difficult to comprehend. Ten years ago, the S&P 500‘s overall capital expenditures were significantly lower than those numbers.
| Indicator | Detail |
|---|---|
| Combined market value of the Magnificent Seven | Around $22 trillion as of late April 2026 |
| Year-to-date market cap erased (per IBD analysis, April 2026) | Roughly $2.1 trillion |
| Hyperscaler AI capex projection (2026) | Approximately $725 billion across the big four |
| Alphabet’s revised 2026 capex range | $180B–$190B, up from $91B in 2025 |
| Microsoft’s expected 2026 capex | Around $190B, with $25B tied to higher component pricing |
| Meta’s revised 2026 capex range | $125B–$145B |
| NVIDIA’s market cap (late April 2026) | About $4.32 trillion |
| Meta’s forward P/E ratio | 26.89x — the cheapest of the seven |
| Apple’s iPhone share of Q1 2026 revenue | 59.3%, with revenue of $85.3 billion |
| S&P 500 cumulative gain (2023–2025) driven largely by Mag 7 | Approximately 78% |
The problem with spending so much money, though, is that eventually someone has to consider the return on investment. Wall Street analysts now predict that in 2026, Alphabet’s free cash flow will decline by about 70% annually. It’s not a typo. In essence, the company is placing a wager that the assets it is purchasing today—such as Nvidia GPUs, custom silicon, server racks, and water-cooled data centers in locations like Council Bluffs and Eemshaven—will continue to be valuable long enough to produce profits that are worth the investment. There’s a good chance they will. Additionally, considering how quickly the underlying AI hardware is evolving, it’s possible that a significant amount of this capital expenditure will become outdated in three years. It’s still unknown. The uncomfortable part is that.
The most intriguing development in 2026 is probably the way the story splits at this point. In comparison to its own history, Meta currently trades at a forward P/E of 26.89, which is less expensive than the S&P 500’s forward multiple of 21.9. It would have sounded absurd in 2024, but some analysts are referring to it as the bargain of the bunch. In contrast, NVIDIA has a market capitalization of more than $4 trillion, a P/E ratio of almost 36, and a P/S ratio of more than 20. These multiples require the company to continuously double its business in order to justify them. A year ago, no one could have predicted Alphabet’s quiet AI-led turnaround. In contrast, nearly 60% of Apple’s revenue still comes from the iPhone, nearly 20 years after the original model was introduced at Macworld. That’s an incredible run, and if any AI-native gadget ever gains the same level of cultural traction, it’s difficult not to sense a slow-burning vulnerability.
Nifty Fifty existed in the 1980s. The dot-com darlings emerged in the late 1990s. The scale is new, but concentration is not. With companies like Broadcom and TSMC pushing into trillion-dollar territory, the WSJ recently questioned whether the market can handle a “Magnificent 10,” only partially rhetorically. The entire developed-market equity market outside of the United States is already rivaled by their combined value. In the past, that kind of imbalance has not ended amicably. Whether the AI capex cycle eventually benefits shareholders or just makes chip suppliers and utility companies that sell them power fatter is still up for debate.
As this develops, it seems like the Magnificent Seven label is beginning to outlive its usefulness. These are no longer seven comparable wagers on comparable futures. The only reason they share a basket despite being seven distinct stories with seven distinct risk profiles is because they all once ascended together. They don’t in 2026. Perhaps it’s the most honest thing the market has done in a long time.
