TechMarketsupdated today

Is the AI boom a bubble or built to last?

Will the Invesco S&P 500 Equal Weight ETF (RSP) return more than the VanEck Semiconductor ETF (SMH), dividends included, from the Nov 4, 2026 close to the Jun 30, 2027 close?

Yes — Bubble 45% · No — Built to last 55%

Base rate and positioning: SMH has beaten RSP in most eight-month windows since 2023, but October broke the trend and a Fed still hiking into negative hyperscaler free cash flow makes this close to a coin toss

The bubble camp argues the money behind the build-out has run ahead of the revenue: Goldman Sachs estimates hyperscalers need about $308 billion a year of AI revenue just to break even, Michael Burry says the industry overstates profits by stretching GPU useful lives, and Vantage Point's Nick Ferres and moomoo's Tapas Strickland now add a Democratic Congress, AI regulation and state limits on Texas data centres to the risks. The built-to-last camp points to demand that has not broken: Amazon's Andy Jassy projects AWS growth above 40% in 2027 and 2028, AMD's data-centre revenue nearly doubled, and BlackRock says it 'still believes in AI' and favours firms selling into the build-out's bottlenecks. Both camps agree October repriced the cost of funding the boom; they disagree on whether demand pays for it.

Yes if: RSP returns more than SMH, dividends included
Window: Nov 4, 2026 → Jun 30, 2027, at the close
Source: Yahoo Finance, adjusted closes

The AI trade has had its worst stretch in more than a year. The Nasdaq Composite fell about 8.8% from its Oct 6 record to Nov 2, chip stocks dropped about 7% on Oct 30 alone, and Goldman Sachs's broad AI basket lost 15% in October and sits more than 24% below its peak. The worry is funding, not demand: hyperscaler capex is running at roughly $650-760 billion this year against a $1.3 trillion consensus for 2027, Alphabet posted its first negative free-cash-flow quarter as a public company, and Oracle had to sell $10 billion of stock 11% below market to keep building. On Nov 4, after Democrats won both chambers of Congress, AI and data-centre names lagged again while healthcare and renewables led. This question starts at today's close: if the build-out is a bubble, the equal-weight index (the average stock) should beat the semiconductor ETF by Jun 30, 2027.

Two total returns over one fixed window, read from adjusted closes: either RSP's is higher or it is not (a tie settles No).

Researched Nov 4, 2026 · tracked since Nov 4, 2026

What do you think?

Take a side

Bubblethis is “yes”

Capex is now more than the hyperscalers' operating cash flow, financed with record bond issuance while the Fed is hiking. If the AI revenue does not show up fast enough, spending gets cut and the chip suppliers take the hit first, while the average stock holds up.

Long

RSP Invesco S&P 500 Equal Weight ETFThe yes leg: the average S&P 500 stock, with little AI concentration. It wins this contract if money rotates out of the AI leaders into the rest of the index.

Long

XLV Health Care Select Sector SPDR FundHealthcare is the defensive sector investors rotated into on Nov 4 as AI names lagged; it holds little AI exposure and benefits if capital leaves the build-out.

Long

XLP Consumer Staples Select Sector SPDR FundStaples are what holds up if an AI unwind drags growth stocks lower; they lag badly if the AI rally resumes.

Long

VTV Vanguard Value ETFValue stocks are the other side of the growth-versus-value trade that AI leadership has dominated; they gain relative ground if AI multiples compress.

Short

SMH VanEck Semiconductor ETFThe no leg: the chipmakers selling into the build-out. Their revenue is the hyperscalers' capex, so they win if spending keeps compounding toward the $1.3 trillion 2027 consensus.

Short

NVDA NVIDIA CorporationThe single biggest beneficiary of AI capex and the largest SMH holding; the purest large-cap bet that the spending is durable.

Short

QQQ Invesco QQQ TrustThe Nasdaq-100 holds the hyperscalers doing the spending and the chipmakers selling to them; it is the broad, liquid way to own the build-out.

Short

VRT Vertiv Holdings CoPower and cooling for data centres: the second-order beneficiary of the build-out, and among the names that lagged on Nov 4 as Texas data-centre risk was repriced.

Worth knowing: This basket lags badly in an AI rebound, and October's drop means some of the bubble case is already in prices. The contract is relative: RSP can win while falling, if SMH falls more.

Built to last

Demand has not broken: AWS expects 40%+ growth, AMD's data-centre sales nearly doubled, and contracted backlogs exceed $1.5 trillion. October repriced funding costs, not the need for compute, and the chipmakers are paid in cash today.

Short

RSP Invesco S&P 500 Equal Weight ETFThe yes leg: the average S&P 500 stock, with little AI concentration. It wins this contract if money rotates out of the AI leaders into the rest of the index.

Short

XLV Health Care Select Sector SPDR FundHealthcare is the defensive sector investors rotated into on Nov 4 as AI names lagged; it holds little AI exposure and benefits if capital leaves the build-out.

Short

XLP Consumer Staples Select Sector SPDR FundStaples are what holds up if an AI unwind drags growth stocks lower; they lag badly if the AI rally resumes.

Short

VTV Vanguard Value ETFValue stocks are the other side of the growth-versus-value trade that AI leadership has dominated; they gain relative ground if AI multiples compress.

Long

NVDA NVIDIA CorporationThe single biggest beneficiary of AI capex and the largest SMH holding; the purest large-cap bet that the spending is durable.

Long

QQQ Invesco QQQ TrustThe Nasdaq-100 holds the hyperscalers doing the spending and the chipmakers selling to them; it is the broad, liquid way to own the build-out.

Long

VRT Vertiv Holdings CoPower and cooling for data centres: the second-order beneficiary of the build-out, and among the names that lagged on Nov 4 as Texas data-centre risk was repriced.

Worth knowing: Semis are the most volatile corner of the market: SMH can fall 20% on a single capex cut or a December Fed hike even if the long-run case is right. Size for the drawdown, not the destination.

5 sources, leaning both ways

yesEconomics Help — The AI Bubble Is in the Bond MarketChecked Nov 4: big-five capex at about 113% of operating cash flow in 2026, Alphabet free cash flow of -$9.1bn, Goldman's $308bn break-even AI revenue estimate, Burry's depreciation critique.
yesYahoo Finance (Reuters) — Market analysts react to Democratic sweep in US midtermsChecked Nov 4: Goldman AI basket -15% in October and 24% below peak; Ferres (Vantage Point) and Strickland (moomoo) flag AI regulation and Texas/Georgia data-centre risk; JPMorgan AM's Tai Hui says AI, not politics, drives markets.
noZTrader — Is the AI Bubble Bursting? What the October 2026 Tech Selloff Is Really PricingChecked Nov 4: argues demand has not broken (Jassy's 40%+ AWS growth outlook, AMD data-centre revenue nearly doubling); October repriced funding costs, favouring chip and server suppliers over debt-funded spenders.
noinvestingLive — Nasdaq's worst month since March 2025 puts returns on AI spending in focus, says BlackRockChecked Nov 4: BlackRock 'still believes in AI' but wants proof of returns; hyperscaler capex about $650bn vs $360bn a year earlier; ex-megacap Q3 earnings growth about 29% vs 18% for the Magnificent Seven.
mixedActionForex — Democrats Sweep Congress, Stocks Rebound and Dollar Sinks: What's Really Being Repriced?Checked Nov 4: Nasdaq-100 futures hit their lowest since late July before recovering; renewables and healthcare led while data-centre and AI infrastructure names lagged.