The Tech Concentration Crisis of 2026: How to De-Risk Your Portfolio Without Missing the AI Boom
The S&P 500 is now roughly a third tech and a third Magnificent Seven — the most concentrated it's been since the dot-com peak. Four tech-resistant strategies to dilute that risk without bailing on AI.
Amazon, Alphabet, Microsoft, and Meta collectively guided roughly $725 billion in 2026 capex, up about 77% from around $410 billion in 2025 — and a growing share is funded with debt rather than operating cash flow.
Get new posts in your inbox
No spam, unsubscribe anytime.
About the Author
Sujit KarkiFinance Researcher & Market Analyst
Independent finance researcher and market analyst with expertise in macroeconomics, equity markets, and personal finance. I help regular investors make better-informed decisions through rigorous, data-driven analysis.
The S&P 500 is the most concentrated it's been since 2000: tech is ~a third of the index, the Magnificent Seven were ~32.5% in mid-July, and the top 10 holdings ~37-38%
The concentration measure (HHI) is near 185-195 in 2026 versus ~123 at the dot-com peak — you're running a tech bet whether you meant to or not
The IMF's July 2026 outlook shows a real growth split: AI-hardware exporters beat forecasts by 4.4 points while everyone else missed by 0.3
You don't have to bail on AI. Four tools — equal-weight, exclusionary, value/dividend, and mid-cap/international ETFs — each structurally own less mega-cap tech
My approach: a barbell. Keep ~60% in a market-cap core, split the other ~40% across equal-weight, mid-cap, international, and value
The Paradox That Kicked Off This Post
Here's the thing that's been bugging me all week. AI spending is exploding — TSMC just bumped its 2026 capital budget up to $60-64 billion, hyperscalers are pouring money into data centers, and the whole story of the year is supposed to be the AI buildout. And yet the Nasdaq dropped about 1.4% on Friday and finished the week down nearly 3%. The semiconductor ETF, SMH, fell roughly 9% over the week and is off something like 17% for the month.
So the capex is booming and the stocks are falling. That's the paradox, and it's worth sitting with for a second. When good news stops moving prices up, it usually means expectations got too high, valuations got stretched, or both. Sticky inflation isn't helping — June CPI cooled a bit, but Chair Warsh's Fed has stayed hawkish with rates parked at 3.50-3.75%, and there's fresh Middle East tension pushing oil around. Add a new Chinese AI model from Moonshot that spooked chip investors, and you get a market that's suddenly nervous about the exact names it loved a month ago.
I'm not here to call a top. I have no idea if this is a blip or the start of something. What I do know is that this kind of week is a gift, because it forces a question most people never ask: if the mega-cap tech trade wobbles, how exposed am I, really?
For most of us, the answer is: way more than we think.
The paradox, mapped
How record capex still produced a red week
Input
AI capex hits record highs
TSMC guides 2026 capex to $60-64B
Hawkish Fed
Rates held at 3.50-3.75%
Sticky inflation
Third straight upward revision
Moonshot AI shock
New Chinese model rattles chip names
Output
Nasdaq -2.9% for the week
SMH (semiconductors) fell ~9% over the same week
When good news stops moving prices, it usually means expectations — and valuations — got ahead of the fundamentals. Figures are approximate, week of July 13-17, 2026.
The IMF's Crosscurrents: Who's Winning the AI Boom
Before we get to portfolios, let's zoom out, because the macro backdrop explains why this concentration exists in the first place.
The IMF dropped its July 2026 World Economic Outlook update on July 8, and they gave it a title that basically writes my intro for me: "Global Economy in Crosscurrents of War and Technology." Global growth is pegged at 3.0% for 2026 and 3.4% for 2027 — steady on the surface. But the average hides a genuinely wild split underneath.
The IMF found that in the first quarter of 2026, the top four net exporters of AI-related hardware — Taiwan, South Korea, Thailand, and Malaysia — posted an average growth surprise of +4.4 percentage points, while the rest of the world missed by 0.3. Think about that gap. If you build the picks and shovels for AI, your economy is beating expectations by a mile. If you don't, you're treading water.
South Korea is the poster child. Even though it imports most of its energy — and energy prices are up thanks to the war — the IMF raised Korea's 2026 growth forecast to 2.6%, a 0.7-point upgrade, purely on the strength of AI-related exports. China got nudged up to 4.6%. Malaysia's at 4.7%, helped by data-center activity. India's cruising at 6.4%. Meanwhile the euro area got cut, because it just isn't plugged into the AI supply chain the same way.
A quick myth-buster
Some early write-ups of this report floated eye-popping numbers like 7.5% for Korea and 8%+ for China. Those aren't the IMF's actual country forecasts — Korea is 2.6% and China is 4.6%. The genuinely striking stat is the divergence: +4.4 points for AI-hardware exporters versus -0.3 for everyone else. Always check the headline number against the source table.
Here's why this matters for your portfolio. The same force lifting Korea and Taiwan is the force that's turned the S&P 500 into what it is today. AI demand has poured into a small cluster of enormous companies, and because the index weights by size, those winners have swallowed a bigger and bigger share of it. The macro divergence and the index concentration are the same story told two ways.
Your Index Fund Is a Tech Fund Now
Let me hit you with the numbers, because they're genuinely startling.
The GICS Information Technology sector by itself is now roughly a third of the S&P 500. The Magnificent Seven — Nvidia, Microsoft, Apple, Amazon, Alphabet, Meta, Tesla — were about 32.5% combined as of July 17, 2026. The top 10 holdings are around 37-38% of the entire index (that's actually down a touch from the ~40-41% record at the end of 2025, as the market broadened a bit in the first half of this year). And if you fold in the tech-flavored names sitting in Communication Services and Consumer Discretionary — Alphabet, Meta, Amazon — you're looking at closer to 44-45% of the index leaning on one theme.
Magnificent Seven weight0.0%Up from ~10% a decade ago, Jul 17 2026
Top 10 holdings0%Down from a ~40-41% record at end-2025
HHI concentration0vs. ~123 at the March 2000 dot-com peak
There's a formal way to measure this called the Herfindahl-Hirschman Index — the same tool antitrust regulators use to spot monopolies. The higher the number, the more concentrated the thing you're measuring. The S&P 500's HHI is sitting near 185-195 in 2026. At the March 2000 dot-com peak, it was about 123. We are meaningfully more concentrated now than we were at the top of the last great tech bubble.
Now, before the "it's different this time" crowd and the "it's exactly like 2000" crowd start yelling at each other — they're both a little right. It is different: the 2000 leaders were burning cash with no earnings, while today's Magnificent Seven are among the most profitable companies that have ever existed. Alphabet alone throws off around $160 billion in annual profit. These aren't Pets.com.
But — and this is the part the bulls skip — the math of concentration doesn't care how good the companies are. When a third of your index rides on seven stocks, a synchronized stumble in those names drags down every 401(k), every target-date fund, and every "diversified" index fund in the country at the same time. You saw a preview this week. When the mega-caps coughed, the equal-weight version of the same index (RSP) barely moved while the cap-weighted one fell.
If you own a plain S&P 500 fund, you're running a tech portfolio with 490 also-rans along for the ride. That's not a criticism of index investing — index funds still beat most active managers over time. It's just a fact about this index at this moment.
The $725 Billion Underneath It
The concentration has an engine, and it's a spending number. Four companies — Amazon, Alphabet, Microsoft, and Meta — collectively guided roughly $725 billion in 2026 capital expenditure, up about 77% from around $410 billion in 2025. If you own an S&P 500 fund, you're a co-signer on that bet whether you meant to be or not.
The question that separates a healthy buildout from a fragile one is how it's paid for, and the answer changed in 2026. For years the hyperscalers funded capex out of operating cash flow. Now they're borrowing: Amazon returned to the investment-grade bond market in July 2026 to raise at least $25 billion, and its free cash flow is projected to turn negative this year. Bank of America research flagged consensus estimates showing AI capex reaching roughly 94% of operating cash flow (minus dividends and buybacks) in 2025 and 2026, up from 76% in 2024.
Cash flow vs. debt
When capex is funded by cash flow, a bad year means slower growth. When it's funded by debt, a bad year means interest payments that don't care whether the AI revenue showed up on schedule. That's the difference that matters for the durability of these companies' earnings — and therefore your index returns.
Two honest caveats. Debt isn't automatically bad; borrowing at reasonable rates to build productive assets is normal. And the bull case is real: Google Cloud's contract backlog reached $460 billion, roughly double the $240 billion reported at the end of Q4 2025 — booked demand, not a hope. The signal to track each earnings season is the funding mix — debt issuance and free cash flow — not the headline capex number.
Enter your equity portfolio value and this estimates how much of your money effectively rides on the Magnificent Seven under a cap-weighted index versus an equal-weight approach:
Interactive · what your index fund actually owns
Your real Magnificent Seven exposure
Your equity portfolio value$100,000
Magnificent Seven weight in your S&P 500 fund32.5%
Default 32.5% reflects the cap-weighted S&P 500 in July 2026 — check your fund's actual weight for a precise figure.
Cap-weighted fund (e.g. VOO, SPY)$32,500
Equal-weight fund (e.g. RSP)$1,400
Extra Mag 7 dollars from cap-weighting alone
$0
Estimate for illustration, not financial advice. Actual fund weights vary by index and change daily.
Four Ways to Dilute the Tech (Without Dumping It)
Here's my core belief, so we're clear before the tactics: the answer is not to abandon tech. AI is real, the earnings are real, and betting against the most profitable companies on earth because a chart looks scary is how people miss decades of returns. The answer is to dilute — to own the tech, plus a bunch of stuff that isn't tech, so a bad month for Nvidia doesn't define your whole year.
There are four clean tools for this. Let me walk through each.
The four tech-resistant tools at a glance
Fund
Ticker
Expense Ratio
Mega-Cap Tech Exposure
Invesco S&P 500 Equal Weight
RSP
0.20%
Very low (~1.4% Mag 7)
iShares S&P 500 ex S&P 100
XOEF
0.20%
None (top 100 removed)
Vanguard Value
VTV
0.03%
Low
ProShares Dividend Aristocrats
NOBL
0.35%
Very low
iShares Core S&P Mid-Cap
IJH
0.05%
Zero (mid-caps only)
Vanguard FTSE Developed Mkts
VEA
0.03%
Low (international)
1. Equal-Weight ETFs (RSP, EQL)
The idea is dead simple: instead of weighting companies by size, weight them equally. In the Invesco S&P 500 Equal Weight ETF (RSP), every one of the 500 companies gets about 0.2%, rebalanced quarterly. Apple gets the same slice as a mid-sized regional bank. The Magnificent Seven, which are ~32.5% of the normal index, collapse to about 1.4% in RSP. That one rule change is the whole strategy.
The trade-off is honest. RSP charges 0.20% versus roughly 0.09% for a typical cap-weighted S&P 500 fund, and over the last decade it's lagged — because cap-weighting rides the winners and equal-weighting keeps trimming them. But flip to 2026 and it's held up far better during weeks like this one, when the mega-caps wobbled and the broader 500 didn't. Equal weight earns its keep when the market broadens; it drags when a handful of giants run.
If you'd rather balance at the sector level, the ALPS Equal Sector Weight ETF (EQL) puts roughly 9% in each of the 11 GICS sectors and rebalances quarterly. It's smaller and a bit pricier, but it's an intuitive way to make sure no single sector — read: tech — dominates.
Pros
Cuts Magnificent Seven exposure from ~32.5% to ~1.4% (RSP)
Built-in 'sell high, buy low' via quarterly rebalancing
Natural mid-cap and value tilt within the same 500 names
Historically outperforms when market leadership broadens
Cons
Higher fee than plain S&P 500 funds
Has lagged badly during past mega-cap runs
Higher turnover from constant rebalancing
EQL is smaller and less liquid — mind the spread
2. Exclusionary ETFs (XOEF)
This one's the newest and the bluntest. The iShares S&P 500 ex S&P 100 ETF (XOEF) does exactly what the name says: it takes the S&P 500 and deletes the largest 100 companies, leaving you the other ~400 at market-cap weight. Since basically all of the mega-cap tech lives in that top 100, XOEF surgically removes the concentration problem while keeping large, established American companies.
I like the concept a lot. BlackRock designed it as a building block — you can pair XOEF with OEF (the S&P 100 fund) to rebuild the full index in whatever ratio you want, dialing your mega-cap exposure up or down like a volume knob. The catch is that it launched in July 2025 and is still tiny, with only roughly $18-22 million in assets. At that size you need to watch the bid-ask spread and not treat it like a mainstream ETF — use limit orders.
Small fund risk
XOEF is the cleanest single-ticker "give me the S&P 500 minus the giants" tool that exists. But small AUM means wider spreads and some closure risk if it never gains traction. Treat it as a satellite, size it modestly, and check that assets are growing before you lean on it hard.
3. Factor Tilting (VTV, NOBL)
Value and dividends are the old-school way to sidestep tech, and they still work because expensive growth stocks — i.e., most of big tech — simply don't screen into these funds.
The Vanguard Value ETF (VTV) is my favorite here on pure cost: a 0.03% expense ratio and a portfolio led by names like Berkshire Hathaway, JPMorgan, ExxonMobil, Johnson & Johnson, and Walmart. It's tilted to financials, energy, healthcare, and industrials — precisely the stuff a tech-heavy index underweights.
For a quality-and-income angle, the ProShares S&P 500 Dividend Aristocrats ETF (NOBL) only holds companies that have raised their dividend for 25 straight years or more. That screen naturally excludes most of big tech (which mostly hasn't been public that long or doesn't prioritize dividends), equal-weights its holdings, and caps any sector at 30%. The cost is steeper at 0.35%, and it'll lag hard when tech is ripping — that's the deal you're making.
4. Geographic & Size Diversification (IJH, VEA)
This is the one people forget, and it might be the most powerful, because it moves you into parts of the market that have zero overlap with the mega-caps.
On size: the iShares Core S&P Mid-Cap ETF (IJH) tracks the S&P MidCap 400 — about 400 mid-sized companies, none of which are the Magnificent Seven, because by definition they're too small to be mega-caps. It costs 0.05% and gives you a slice of the market a plain S&P 500 fund essentially ignores.
On geography: the Vanguard FTSE Developed Markets ETF (VEA) owns roughly 4,000 companies across Europe, Japan, Canada, and Australia at a 0.03% fee. And here's the valuation kicker — developed Europe ex-UK trades at roughly 16x forward earnings versus roughly 23x for the US, for broadly similar future earnings streams. That's a big discount. I'm not saying that gap closes tomorrow — the prior decade was a US blowout — but owning only one country is itself a concentration bet.
My Take: Dilute, Don't Abandon
Okay, here's where I stop hedging and tell you what I'd actually do.
I'd run a barbell. Keep about 60% in a market-cap core — a normal S&P 500 or total-market fund — because I genuinely don't want to miss the AI boom, and cap-weighting is the cheapest, most tax-efficient way to own the winners. Then split the other ~40% across the four tools: something like 10% equal-weight (RSP), 10% mid-cap (IJH), 10% developed international (VEA), and 10% value or dividends (VTV or NOBL).
What does that do? It quietly drags your effective tech weight down from ~33% toward the low-to-mid 20s, without ever making you sell the thing that's been working. If AI keeps ripping, your core carries you. If the mega-caps break — like they threatened to this week — your satellites cushion the fall. You don't have to be right about the timing, which is the whole point, because nobody is.
Play with the mix yourself below — the same barbell logic used above, with the actual sector math behind it.
A word on the crypto angle, since it always comes up when people talk about "concentration." There's a fashionable institutional pitch right now arguing that crypto is the natural hedge against an economy where AI concentrates wealth into a few hands. I find the framing interesting and the conclusion self-serving when it comes from companies whose balance sheets depend on you believing it. Crypto is a legitimate diversifier for a small, risk-tolerant sleeve, but it is not a substitute for the boring equity diversification above, and I'd treat any "AI wealth concentration = buy my coin" argument with a healthy raised eyebrow.
Diversification isn't exciting. It will, by design, always leave you with something in your portfolio that's underperforming — that's proof it's working, not evidence it's broken. The goal isn't to beat the S&P 500 next quarter. It's to still be invested, and still sane, when the concentration everyone's ignoring finally matters.
Frequently Asked Questions
Amazon, Alphabet, Microsoft, and Meta collectively guided roughly $725 billion in 2026 capex, up about 77% from around $410 billion in 2025 — and a growing share is funded with debt rather than operating cash flow.
This is educational content, not financial advice. I'm a researcher, not your advisor, and I don't know your situation, taxes, or risk tolerance. Every fund and figure here changes daily and past performance doesn't predict future results. Do your own research and talk to a licensed financial professional before making changes to your portfolio. See the full disclaimer.
Sujit Karki|||13 min read