On 6 October, the S&P 500 closed at 7,818.93, a new all-time high and about 14% higher than at the start of the year. At the same time, however, the US 10-year Treasury yield is hovering near 5.3%, and on 7 October it reached 5.36% intraday, a level not seen since the early 2000s.
The prevailing market narrative presents a major contradiction. On the one hand, equities suggest that growth and earnings justify higher valuations. On the other, the bond market indicates that money is expensive, inflation remains a problem and financing costs do not appear likely to return to the levels of the previous decade.
Behind this contradiction lies a second question: how broad and deep is the equity rally really?
The answer is that the S&P 500 is not simply a rising market. It is a market with historically high concentration, with artificial intelligence at the centre of an investment cycle that requires ever-greater financing.
This does not mean we are facing a new bubble like the one in 2000. It means something more useful for investors: the market may continue to rise while also becoming more vulnerable.
The S&P 500 is no longer “the market”
The S&P 500 is weighted by market capitalisation. The more its largest companies rise, the more they lift the index as a whole.
Today, the three largest stocks — Nvidia, Apple and Microsoft — account for about 21.7% of the index. The ten largest account for about 39%.
The comparison with the past is revealing.
Table 1 — S&P 500 concentration
| Measure | Today | Historical comparison |
|---|
| Three largest stocks | 21.7% | ≈13% in the 1980s |
| Ten largest | ≈39% | ≈26% in 2000 |
| Technology | ≈39% | ≈33% at the dotcom peak |
| Top 10 market capitalisation | ≈$27.4tn | — |
| Shiller CAPE | >40 | 44 in 2000 |
Shiller CAPE: the S&P 500 price divided by average earnings over the past decade (adjusted for inflation). A reading above 40 means investors are paying more than $40 for every dollar of earnings; it has only been higher once, in 2000 (44).
Concentration alone does not mean a market decline is inevitable. It does, however, mean that investors buying “the S&P 500” have much greater exposure to a handful of companies and to their very specific investment cycle (AI) than the index’s name suggests.
This becomes even clearer when we look beneath the surface.
Table 2 — What happened in September 2026
| Index | Change |
|---|
| S&P 500 | −0.5% |
| S&P 500 stocks that closed lower | 75–78% |
| Stocks down ≥5% | ≈60% |
| Stocks down ≥10% | ≈27% |
| S&P 500 Equal Weight | >−4% |
| Russell 2000 | −5.3% |
| Nasdaq-100 | +3.3% |
On 21 September, almost 60% of S&P 500 stocks were already 25% below their highs. The conclusion is now clear: the index is at a record, but a large part of the market is not.
What this means for investors
Risk should not be measured solely by the S&P 500’s return. Investors should look at how much of their portfolio ultimately depends on the same large companies and the same AI cycle.
So is this a market supported only by AI? No. That is an important distinction.
The Magnificent 7 are indeed driving earnings growth. They increased earnings by about 63.2% in the first quarter of 2026 and by 31.1% in the second quarter. But the other 493 S&P 500 companies are also growing their earnings.
Table 3 — Earnings growth
| Quarter | Magnificent 7 | Other 493 | S&P 500 |
|---|
| Q1 2026 | +63.2% | +17.4% | +28.4% |
| Q2 2026 | +31.1% | +22.8% | +24.7% |
| Q3 2026* | +31.8% | +25.5% | — |
| Q4 2026* | +22.8% | +25.3% | — |
*FactSet estimates.
As is clear, earnings growth is genuinely broadening. The notable point is elsewhere: among the biggest contributors to earnings growth in the second quarter, Nvidia and Micron are near the top, alongside Broadcom and major energy companies such as Chevron and Exxon Mobil.
This points to a more precise description of the market: it is not simply a market of ten stocks. It is an AI investment cycle, with energy acting as a second engine. That is a stronger case than saying “it’s all Nvidia”.
There is another important counterbalance: the S&P 500’s forward P/E fell from about 22.4 at the end of 2025 to 19.1 in September 2026. In other words, corporate earnings grew faster than share prices.
So the market did not become more expensive simply because shares rose. The problem is that the fall in the P/E ratio now has to be weighed against the cost of money.
Heavy investment in AI infrastructure is a test
This may be the most important point in the analysis. The four major hyperscalers — Alphabet, Amazon, Microsoft and Meta — are directing about $720bn–$745bn to capital expenditure this year, depending on the definition used and individual estimates.
Including Oracle and SpaceX, S&P Global’s estimate reaches about $870bn for 2026. The same analysis puts the figure at about $1.3tn for 2027.
Table 4 — The AI investment cycle
| Company | Capex 2026
Capital expenditure | Capex/sales | FCF 2027E
Free cash flow |
|---|
| Alphabet | $195–205bn | ≈46% | −$82.7bn |
| Amazon | ≈$220bn | ≈25% | −$60.1bn |
| Microsoft | ≈$175bn | ≈47% | +$33.6bn |
| Meta | $130–145bn | ≈54% | −$3.5bn |
| Oracle | ≈$56bn | ≈86% | −$41.6bn |
| S&P Global total* | ≈$870bn | — | — |
*The total estimate includes six companies, including SpaceX.
The question, therefore, is not whether AI is generating real demand. It is. Nvidia is the clearest proof. In the second quarter of its 2027 financial year, it generated revenue of $96.2bn, up 106% year on year, while data centre revenue rose 117%.
The real question is a different one: when will these investments translate into free cash flow for the companies making them?
That is where the nature of the risk changes.
Debt financing for capex: this is where the risk lies
For years, large technology companies were able to fund investment from their cash flows.
Now the picture is changing. S&P Global estimates suggest that, among the six companies in its sample, only Microsoft is expected to have positive free cash flow in 2027.
The shortfall has to be financed. And a significant part is already being funded through the bond markets or private debt (see our recent related article).
Table 5 — AI debt
| Measure | Amount |
|---|
| Hyperscaler bond issuance in 2025 | ≈$109bn |
| Through 10 August 2026 | ≈$220bn |
| Goldman Sachs estimate for 2027 | ≈$420bn |
| AI issuer spreads | ≈115 basis points |
| Investment-grade spreads | ≈78 bps |
| Oracle five-year CDS | ≈200 bps |
| Oracle 2025 | |
This is not a credit crisis. Leading hyperscalers have strong balance sheets, enormous revenues and access to markets. But the direction of travel matters. As capex rises and free cash flow falls, the continuation of the cycle depends increasingly on the ability to borrow.
Investors should therefore no longer track only AI earnings. They must also monitor its financing costs.
The bond market is back
This may be the most underappreciated feature of today’s market. With the forward P/E at about 19.1, the S&P 500’s earnings yield — the inverse of the P/E — is about 5.2%. The 10-year Treasury already yields about 5.3%.
These figures are not directly comparable: 5.2% is an accounting earnings yield, not cash received by a shareholder. The comparison is nevertheless indicative. For much of the previous decade, equities had a significant advantage over bonds.
Today, that advantage has almost disappeared.
Table 6 — The relationship between equities and bonds
| Measure | Approx. |
|---|
| S&P 500 forward P/E | 19.1x |
| S&P 500 earnings yield | ≈5.2% |
| 10-year Treasury | ≈5.3% |
| Difference | approximately zero |
| Shiller CAPE | >40 |
This does not mean “sell equities”. It means investors must now demand higher quality from their portfolios. When bonds once again offer a meaningful return with low risk, a share cannot justify any price simply because it has a good story. The opportunity cost is back.
The greatest risk is not a 10% correction
Corrections are normal. Since 1990, the S&P 500 has fallen by about 14% on average at some point during the year, even in years that ultimately ended with a positive return. A fall of 10% or 15% is therefore not, in itself, a reason to panic.
The real risk is a combination of a narrow market + high valuations + expensive money + massive capex + rising debt. If all of these remain under control, the bull market may continue.
But if, at the same time:
- hyperscalers’ free cash flow remains negative,
- their bond spreads widen,
- the 10-year yield rises above 5.5%,
- and the market continues to depend on an ever-smaller number of stocks,
then the correction could become much more severe. The risk is not the decline itself. It is the market losing its ability to absorb it.
What does all this mean for Europe and Greece?
The pressure is not confined to the US. Long-term European interest rates have also risen, as inflation and energy constrain central banks’ room for manoeuvre.
For Greek investors, however, there is another story. Greece is gradually returning to the core of developed markets. FTSE Russell, STOXX and S&P Dow Jones have moved to upgrade the country, while MSCI has decided to reclassify Greece from Emerging to Developed Market in May 2027.
This is a significant development. It is not, however, an automatic buy signal. MSCI itself estimates that the Greek stock market will account for about 0.40% of MSCI Europe, while the Greek index remains heavily concentrated in banks.
Table 7 — Greece: major institutional shift, small weighting
| Factor | What it means |
|---|
| MSCI Developed Market | Upgrade in May 2027 |
| Indicative MSCI Europe weighting | ≈0.40% |
| Main sector | Banks/financials |
| Institutional benefit | Greater access to developed-market investors |
| Investment risk | Concentration and sensitivity to interest rates |
Greece, then, is not Europe’s version of AI. Its investment case is different: banks, profitability, dividends, economic growth, infrastructure and the cost of money. The upgrade is a reason to keep an eye on the market. It is not a reason to pay any price.
So what should investors do now?
This is the most important conclusion of the analysis. There is no need to predict when the rally will end. Investors need to make sure that if it continues, they participate — and if it reverses, they are not wiped out.
First: Measure your real exposure
It is not enough to look at how much you hold in the S&P 500. You should add your indirect exposure to Nvidia, Microsoft, Amazon, Alphabet, Meta and the broader AI cycle through the Nasdaq, technology ETFs and individual stocks. Diversification should be measured by ultimate economic risk, not by the number of securities.
Second: Rebalance, don’t time the market
When an asset class grows because it has risen, the answer is not to predict the peak but to bring the portfolio back within the limits set from the outset. The rule: review and rebalance quarterly if an asset class moves more than five percentage points above its target (for example, equities at 65% rather than 60%). Discipline, not prediction.
Third: Liquidity and quality income
With the 10-year yield above 5%, cash is no longer “dead capital”: it yields almost as much as equities. And when the stock market corrects (by an average of 14% during the year), it allows you to buy at lower prices rather than being forced to sell.
Fourth: Quality over story
For every stock that has risen because of its links to AI, investors should ask four questions:
- Are revenues actually growing?
- Is free cash flow increasing?
- How much debt is needed to sustain growth?
- Does today’s price leave room for error?
If the answer to the fourth question is “almost none”, the problem is not the company. It is the price at which the stock is trading.
Three scenarios — and the appropriate response
We do not need to predict which one will happen. We need to be prepared for all three.
Table 8 — A guide for investors
| Scenario | What would confirm it | Prudent response |
|---|
| The rally continues | Strong Q3 results, high 2027 capex, improving FCF, 10-year yield holding below 5.5%, broader market participation | Maintain exposure, rebalance, don’t chase the rally |
| Normal correction | Pressure from interest rates or results, without an earnings collapse | No hasty selling; gradual purchases of quality stocks |
| Deeper correction | Capex cuts, wider spreads, 10-year yield >5.5%, further narrowing | Protect liquidity, reduce excessive concentrations, buy only companies with strong balance sheets |
That is the real advantage of discipline: you do not need to know the future to have a plan.
Four indicators every investor should watch
The vast amount of daily market information can be distilled into four key indicators.
Table 9 — The investment dashboard
| What to watch | Why it matters | Positive sign | Warning sign |
|---|
| Hyperscaler FCF | Shows whether capex is paying off | An upward trend | Persistently negative FCF |
| Bond spreads/CDS | Measures financing costs | Stable spreads | Sharp widening |
| 10-year Treasury | Determines the cost of capital | <5% and stable | >5.5% and rising |
| Market breadth | Shows how broad the rally is | More stocks are participating | A handful of stocks are holding up the index |
If all four indicators are improving, the market has reason to keep rising. If they deteriorate at the same time, risk increases even if the S&P 500 continues to hit new highs.
What would make the picture genuinely better?
Analysis should not focus only on risks. There are strong arguments for a continued rise.
Nvidia is generating real revenue and huge cash flows. The other 493 companies are also growing their earnings. The forward P/E has fallen despite the index’s rise. And the leading hyperscalers have strong balance sheets.
These are not the hallmarks of a classic bubble without fundamentals. For the picture to become much more convincing, however, we would need to see three things:
- First, hyperscalers’ free cash flow should start to recover as their huge capex turns into revenue.
- Second, bond spreads should remain under control even if debt issuance increases.
- Third, the 10-year Treasury yield should fall back below 5%, without inflation returning.
There is also a fourth sign that would matter even more: the majority of stocks should start participating in the rally again.
If the S&P 500 rises alongside the equal-weight index, smaller companies and a broader range of stocks, the record will have real depth.
Conclusion: The record is real. But is it broad?
The S&P 500 is not the bubble of 2000. Earnings and demand for computing power are real, and the large groups have customers and strong balance sheets.
But the risk is high for four reasons:
- Concentration is at a record high: the ten largest stocks account for about 39% of the index.
- AI requires hundreds of billions of dollars a year, free cash flow is under pressure and financing relies increasingly on debt.
- The 10-year Treasury now yields as much as the S&P 500’s earnings yield.
- Three out of four index stocks closed lower in September.
This is not a signal to exit the market. It is a signal to be disciplined. Four rules:
- Measure how much of your portfolio depends on the same cycle.
- Rebalance anything that has moved beyond your limits.
- Keep liquidity so you can buy when others are forced to sell.
- Prefer companies that generate cash flow over those that continually need fresh capital to grow.
The same logic applies in Greece. The country’s return to developed markets is an institutional achievement, not a substitute for valuation. The Athens Stock Exchange remains a bet on banks and interest rates.
The biggest mistake today is to have only one scenario, whether optimistic or pessimistic. The record is real. The question is whether it can hold up when the market starts demanding proof of earnings, cash flows and the cost of money.
You do not need to predict when the party will end. You need to still be in the game even when it does.
Sources: 1. FactSet, Earnings Insight, S&P 500 earnings and forward P/E data, 2026, 2. FactSet Insight, Are “Magnificent 7” Companies Still Top Contributors to S&P 500 Earnings Growth for Q2?, 20/7/2026, 3. FactSet Insight, Industry Analysts Project 20% Increase in S&P 500 Price Over the Next 12 Months, 28/9/2026, 4. S&P Global, estimates for hyperscaler capex and free cash flow, 2026, 5. CreditSights, Tech: Raising Hyperscaler Capex 2026 Estimates, 6. UBP, Financing the AI Build-out, 7. NVIDIA, second-quarter fiscal 2027 results, August 2026, 8. U.S. Treasury, US government bond yield data, Federal Reserve, FOMC decisions and minutes, September–October 2026, 10. Reuters, data on the S&P 500, Treasury yields and hyperscaler financing, October 2026, 11. Associated Press, US market data, 7/10/2026, 12. CNBC, Broken market: 75% of S&P 500 stocks had a lousy September, 30/9/2026, 13. Nasdaq, September 2026 Review and Outlook, 14. U.S. Bank, Is a market correction coming?, September 2026, 15. MSCI, Greece’s Journey Back to Developed Market Status, 22/4/2026, 16. MSCI, MSCI to Reclassify the MSCI Greece Indexes from Emerging Market to Developed Market Status, 31/3/2026, 17. Euronext Athens / FTSE Russell, announcements on the upgrade of the Greek market to Developed Market status, 18. Goldman Sachs, CreditSights and Reuters, data on hyperscaler bond issuance and spreads.
Note: Market, bond yield and index data are dynamic and refer to the dates cited. Estimates for future earnings, capex, FCF and index levels are not forecasts. This article is for informational purposes only and does not constitute investment advice.
* Nicholas Havoutis has many years’ experience leading strategic finance teams, having held executive roles at JPMorgan (New York), Chase Manhattan Bank (London) and Eurobank (Athens). He also has a significant presence in the media. Today, as head of SoZone Limited, he advises businesses and investors on international growth, operational optimisation and mergers and acquisitions strategies.
