Narrations of a Sector ETF Operator: AI Can Still Carry the Market — But Now It Has to Earn Its Premium

Narrations of a Sector ETF Operator: AI Can Still Carry the Market — But Now It Has to Earn Its Premium

The equity market ended a difficult week doing something important: it absorbed a sharp bond selloff, oil above $100 for much of the week and increasingly hawkish inflation signals — and still finished higher. The S&P 500 gained 1.2% for the week and the Nasdaq advanced 2.0%, with Technology again providing the primary tailwind. Friday added some encouraging breadth as semiconductors, machinery, banks, builders, autos and travel-related stocks outperformed while crude oil and short-term Treasury yields declined.

The resilience matters because the valuation backdrop is more nuanced than the usual argument that AI enthusiasm has simply made the market too expensive. The S&P 500 was trading just under 19 times expected earnings late last week, its lowest forward valuation since 2023 according to LSEG data cited by Reuters. FactSet has also found that analysts raised aggregate Q3 S&P 500 earnings estimates by 1.2% during July and August, compared with the normal pattern of estimate reductions at this stage of a quarter. The market is certainly not cheap relative to long-run history, especially with Treasury yields above 5%, but earnings are doing more of the valuation work than they were earlier in the cycle.

That distinction becomes even more important when we look underneath Information Technology (VGT). The AI complex is not trading at one uniform valuation, and the differences between NVIDIA, Microsoft, Meta, TSMC and Micron help explain where the next phase of sector leadership may develop.

AI’s Valuation Test Is Becoming More About Earnings Than Multiples

NVIDIA (NVDA) is perhaps the clearest example. Consensus estimates currently place the shares around 22.6 times fiscal-2027 earnings, only a moderate premium to a broader market trading near 19 times forward earnings. That premium looks very different when set against NVIDIA’s operating trajectory. Fiscal second-quarter revenue surged 106% year over year to $96.2 billion, Data Center revenue increased 117%, and management guided fiscal Q3 revenue to approximately $108 billion, despite assuming no China Data Center compute revenue in the outlook. Gross margin guidance remains around 74%.

The valuation question around NVDA therefore is not primarily whether the stock carries an enormous market multiple. It is whether consensus earnings can continue growing rapidly enough to make a low-20s forward P/E sustainable. If AI infrastructure demand continues anywhere close to management’s current trajectory, NVDA’s premium to the S&P 500 looks relatively modest. The larger risk would be a deceleration in compute spending that causes forward earnings estimates to roll over at the same time the risk-free rate remains above 5%.

Taiwan Semiconductor (TSM) presents an even more interesting relative-value case. Consensus estimates put TSMC around 23 times 2026 earnings but only about 17.5 times 2027 earnings, which would place it below the broader U.S. market on the latter estimate. Meanwhile, TSMC generated $40.2 billion of second-quarter revenue and is guiding Q3 revenue to $44.6-$45.8 billion, with gross margins still expected between 65% and 67%. If those estimates hold, TSM offers exposure to the AI infrastructure boom without an obvious premium valuation relative to the S&P 500. Its discount reflects other risks — particularly geopolitical concentration and the capital intensity of semiconductor manufacturing — rather than weak underlying demand.

Micron (MU) sits at the extreme end of the valuation spectrum. Current consensus estimates imply roughly 14.8 times fiscal-2026 earnings and only about 6.8 times fiscal-2027 earnings. That is an extraordinary discount to the broader equity market, but it also tells us that investors remain skeptical that today’s memory profitability represents a normalized earnings base. Micron’s most recent guidance calls for approximately $50 billion of fiscal Q4 revenue, 86% gross margins and $31 of non-GAAP EPS, following record fiscal Q3 results. Wednesday’s report therefore matters enormously. If HBM and DRAM demand remain tight and management signals that current margins can persist longer than the market expects, the valuation remains exceptionally low relative to both the S&P 500 and most AI peers. If memory pricing is close to a cyclical peak, that apparent cheapness could prove misleading.

The software and platform side of the AI complex is more expensive. Microsoft (MSFT) trades around 26 times fiscal-2027 consensus earnings, a roughly 35%-plus premium to today’s broader-market valuation. But Microsoft is providing unusually strong forward evidence to support that premium. Management expects Intelligent Cloud revenue of roughly $41 billion in the current quarter, representing 33%-34% growth, while Azure revenue is expected to grow approximately 45% in constant currency. Microsoft has also said demand continues to exceed available capacity. This week’s launch of expanded Copilot capabilities, including its persistent Autopilot agent, strengthens the argument that Microsoft can monetize AI not only through Azure infrastructure but through its enormous installed enterprise-software base.

That makes MSFT expensive relative to the S&P 500, but the premium is at least attached to visible recurring revenue, cloud demand and potential ARPU expansion. The risk is straightforward: at 26 times earnings and with Treasury yields above 5%, Microsoft has less room for a guidance disappointment than a stock trading at the market multiple.

Meta Platforms (META) sits between those extremes. After the strong run associated with Muse, consensus valuation has moved to approximately 24 times 2026 earnings and 22 times 2027 earnings. That is a premium to the broad market, but not an extreme one given Meta’s growth and enormous existing monetization base. Management’s most recent guidance calls for Q3 revenue of $61-$64 billion, while 2026 capital spending is expected to reach $130-$145 billion. Meta nevertheless continues to expect 2026 operating income above 2025 levels. Muse potentially adds another monetization layer: analysts cited by Reuters have already begun modeling material revenue from subscriptions and agent-driven transactions if early adoption persists.

META illustrates the key debate facing the entire AI complex. The revenue opportunity can expand rapidly while the capital required to pursue it expands almost as quickly. Investors increasingly have to judge both sides of that equation.

The AI Trade Is Broadening — and That Helps Industrials

We’ve seen these dynamics emerge repeatedly in this week’s headlines. AI demand is spreading from training GPUs toward inference, agents, CPUs, memory, networking and cloud infrastructure. Akamai’s agreement with Anthropic — an $11.6 billion take-or-pay cloud contract with expansion options that could take the relationship toward $20 billion — is one of the clearest recent examples.

That broadening matters for Industrials (VIS). Friday’s durable-goods data showed core capital-goods orders rising 1.6% month over month, far ahead of expectations. AI data centers require electrical equipment, power generation, cooling, construction, engineering and transportation alongside semiconductors. Machinery and industrial names outperforming Friday could be an early sign that investors are beginning to recognize this second-order exposure.

Thursday’s ISM Manufacturing report will test that thesis. FactSet consensus calls for 55.0 versus 54.6 previously. Strong new orders would reinforce the industrial-capex story; another large increase in prices would be more problematic because the bond market could interpret strong manufacturing activity as further evidence that monetary policy still isn’t restrictive enough.

Energy Is Still Setting the Inflation Temperature

For Energy (VDE), crude remains one of the market’s largest macro swing factors. WTI declined 2.3% Friday as investors responded to reports of potential progress in U.S.-Iran diplomacy, but weekend headlines quickly introduced more uncertainty around the negotiations and the Strait of Hormuz.

Higher oil supports Energy earnings but simultaneously raises the discount rate applied to almost everything else. Lower oil could pressure VDE while improving the setup for Consumer Discretionary (VCR), homebuilders, transports, Real Estate and longer-duration Technology.

Oil has effectively become the market’s inflation thermostat. For sector investors, a sustained decline in crude could be nearly as important as a soft inflation report because it would reduce the probability that today’s price pressures become embedded in wages and consumer expectations.

Is the Market Vulnerable to a Correction?

The market remains tactically vulnerable, but valuation alone is not an especially strong correction argument at the moment.

A broad market at roughly 19 times forward earnings is not inexpensive with the 10-year Treasury above 5%, but it is materially less stretched than it was when multiples exceeded 20 times earlier this year. More importantly, some of the most important AI companies are not trading at dramatically higher multiples than the index. NVDA around 22-23 times forward earnings and META around 22 times 2027 estimates are very different valuation propositions from traditional speculative-growth stocks with little current profitability. TSM already trades below the market on 2027 estimates, while MU’s multiple implies substantial skepticism about the durability of today’s profits.

Microsoft is the more obvious premium-valuation case, but Azure’s expected 45% constant-currency growth provides a fundamental reason investors are willing to pay more.

The vulnerability instead lies in the earnings denominator. Today’s multiples look reasonable only if the extraordinary forward earnings expectations embedded in NVDA, MU, TSM, MSFT, META and the broader AI ecosystem remain credible. If the 10-year stays above 5% while AI growth estimates are revised lower, investors would face both earnings compression and multiple compression simultaneously.

That is a much more dangerous combination than high valuation by itself.

Next Week Tests Both Sides of the Equation

Wednesday is the crucial day because investors receive direct evidence on both sides of the valuation equation.  FactSet consensus expects core PCE inflation to rise 0.3% month over month and 3.4% year over year, versus 3.3% previously. A hotter number would raise the discount rate again. Micron reports later that day and gives investors a direct test of whether AI-related memory earnings are still rising fast enough to offset that higher rate.  Thursday’s ISM Manufacturing report then tests the breadth of the capital-spending cycle, while Friday’s employment report will determine whether labor conditions are cooling sufficiently to relieve pressure on Fed policy. FactSet consensus calls for roughly 90,000 nonfarm payrolls, unemployment at 4.1% and hourly earnings rising 0.3% month over month.

For sector investors, the ideal outcome is no longer simply strong economic data. It is strong enough growth to sustain earnings without enough inflation to drive the 10-year materially higher.  That outcome would leave VGT as the primary leadership vehicle while improving the opportunity for VIS as AI capex broadens and providing relief to VCR, VNQ and VPU. A hotter inflation regime would strengthen the relative case for VDE and potentially VAW, but another sharp increase in Treasury yields would eventually become a problem even for Technology companies with outstanding fundamentals.

The AI thesis still looks strong. NVIDIA’s $108 billion quarterly revenue guide, TSMC’s accelerating sales outlook, Micron’s extraordinary memory profitability, Microsoft’s 45% Azure growth forecast and Meta’s continued revenue growth all argue that the earnings side of the trade remains real.  The more important question now is valuation relative to those earnings expectations and relative to the broader market. In several cases, the premium is surprisingly small. That gives AI leadership more room to withstand inflation pressure than headline index valuations might suggest.  But it also means the next phase of the trade becomes much less forgiving of disappointing guidance.

For the sector ETF operator, that is the pivot: AI does not necessarily need lower valuations to keep leading. It needs earnings growth strong enough to keep today’s valuations looking reasonable while the bond market demands a much higher return on capital.

 

 

Sources

  • FactSet market, economic and earnings data, September 2026.
  • NVIDIA Investor Relations, Fiscal Q2 2027 results and Q3 guidance, August 26, 2026.
  • Micron Technology Investor Relations, Fiscal Q3 2026 results and Q4 guidance.
  • TSMC Investor Relations, Q2 2026 results and Q3 guidance.
  • Microsoft Investor Relations, Fiscal Q4 2026 results and Q1 FY2027 outlook.
  • Meta Platforms Investor Relations, Q2 2026 results and Q3/full-year outlook.
  • Reuters/LSEG, U.S. equity-market valuation and market coverage, September 2026.
  • MarketScreener consensus valuation data, September 25, 2026.
  • S&P Global Market Intelligence, September 2026 Flash U.S. PMI.
  • U.S. Bureau of Economic Analysis, Bureau of Labor Statistics and Institute for Supply Management economic-release calendars.

Disclaimer:  This material is for informational and educational purposes only and does not constitute individualized investment advice, an offer to buy or sell securities, or a recommendation of any particular investment strategy. Sector ETFs and individual equities can experience substantial volatility and concentration risk.

Patrick Torbert

Editor | Chief Strategist

Patrick Torbert is a veteran financial market analyst who is currently the Editor and Chief at ETF Insight a NY based full-service content, TV, video podcast and digital marketing firm that represents several ETF issuers. Patrick brings 20+ years of experience from Fidelity Asset Management where he most recently served as an equity and multi-asset analyst.
Scroll to Top

Subscribe to our Newsletter

Stay updated with the latests analysis and insights from etfsector.com

If you haven’t received your newsletter email, check your spam/junk folder and add us to your contacts to ensure delivery.