Weekly Warm-up: Quality Breaks Out; What's Next?
Quality’s breakout and widening advantage in earnings revisions reinforce our view that the mid-cycle transition has further to run. We highlight our updated quality + strong EPS revisions screen today. Warsh's framing of financial conditions could be key to the end of the broader correction. Quality Is Breaking Out as our Mid-Cycle Transition Extends. Our quality factor has reached a new 52-week high, gaining ~6% since mid-August, while the equal-weight S&P 500 has fallen nearly 4%. Leadership has been broad across quality measures, led by profitability, consistent with our view that earnings durability should matter more as the cycle progresses. Earnings Revisions Support Further Quality Outperformance. The spread in earnings revisions breadth between high and low quality has reached nearly 10%, the second-widest reading since 2010. Historically, quality has outperformed by an average of 7.5% over the following year at similar levels. Relative valuation for the factor remains near its historical median, reinforcing our preference for quality. Financial Conditions Are Doing the Fed's work, in-line with Warsh's Preference. Our composite of rates, the dollar and energy is at its tightest since 2022, with economically sensitive stocks down nearly 10% since the middle of August. We think some acknowledgement of this tightening by Warsh would be well received by rate-sensitive areas that have already de-rated notably. Conversely, an indifference to this tightening may invoke memories of Powell's 2018 "Auto-Pilot" remarks in reference to that tightening episode. Where are we in the broader equity market correction? Equity and other risk markets have been correcting for over 3 months. We were early to identify this correction as a migration from early cycle to mid cycle at the same time the AI trade was transitioning from enablers to adopters. Key observations include the poor breadth of the market and a very distinct and now obvious outperformance of the quality factor noted above. The finality of such corrections typically end with the high quality index catching down even as the factor itself holds up. This can happen quickly or more slowly. Be ready either way as this will likely mark the end of the correction, rather than something more ominous about the end of the business or capex /credit cycle. companies covered in Morgan Stanley Research. As a result, investors should be aware that the firm may have a conflict of interest that could affect the objectivity of Morgan Stanley Research. Investors should consider Morgan Stanley Research as only a single factor in making their investment decision. report.
US Equity Strategy | North America Weekly Warm-up: Quality Breaks Out; What's Next?
Quality Breaks Out; What's Next?
We have been arguing for several months that the market is moving from an early-cycle recovery led by lower-quality stocks into a more selective mid-cycle environment where earnings durability matters more. That transition is becoming clearer beneath the index.
Earnings remain strong, but the rise in rates and the more hawkish policy path priced since Jackson Hole have compressed valuations across much of the market. As a result, the S&P 500 continues to look more resilient than the average stock, while leadership has been shifting toward the higher-quality companies we have been recommending.
One of the clearest signs is the recent performance of our composite quality factor. It has risen to a new 52-week high, surpassing its late-July peak ( Exhibit 1 ). The move is particularly notable because quality lagged during the final leg of the summer rally. Since the equal-weight S&P 500 peaked on Aug. 14, our quality long-short composite has gained ~6%. The improvement has been broad across the quality sub-composites, led by profitability, as investors are being more selective about the companies that can sustain earnings, margins and cash flow through higher financing costs.
rolling recession and the beginning of a new economic cycle--a time when low quality outperforms. That ends with the transition to mid-cycle which we identified in June with the peak rate of change on earnings revision breadth for the low quality early cycle leaders (Semis, consumer discretionary and financials). It is typically accompanied by a more hawkish Fed which also arrived in June with the new Chair. The key takeaway for investors is that this is not the end of the business cycle but it does lead to new leadership; a transition that is usually painful due to the crowding in low quality stocks that happens in an economic recovery from recession.
One of the more important developments beneath the surface has been the widening gap in earnings revisions between high and low quality. On a three-month average basis, the spread in earnings revisions breadth has reached about 10%, the second-widest reading since 2010. In our view, that divergence is worth paying attention to given the historical relationship between revisions and subsequent price performance. The revisions spread has tended to lead relative performance by roughly 6–12 months. When it has been in the top quintile of its historical revisions range, as it is today, high quality has outperformed low quality by an average of ~8%% over the following 12 months. That compares with average outperformance of roughly 1% across all periods. Importantly, quality’s relative valuation sits near its historical median, trading roughly in line with low quality on forward P/E, so investors are not being asked to pay up for that revisions advantage.
The September CPI report reinforces why we want to own quality across a range of rate outcomes. Core CPI rose 0.33% month over month versus 0.2% consensus, with the year-over-year rate at 2.57% and upside in both goods and services. At the same time, we should be careful about extrapolating one month into a sustained acceleration in inflation.
Airfares were up 22.5% year over year as higher oil prices passed through, computer prices rose 1.4% month over month amid AI-related chip cost pressures, and our economists suspect that part of the jump in owners’ equivalent rent was noise. These are meaningful price increases, but they do not necessarily imply inflation is broadening.
In fact, we expect the share of PCE categories running above 3% year over year to decline modestly ( Exhibit 4 ). Given Warsh’s focus on inflation breadth, the distinction between a strong monthly reading and the distribution of the underlying price pressures matters when assessing how long the hiking cycle ultimately needs to last. The other side of that debate is how much tightening the markets have already delivered. Our composite of Treasury yields, the dollar and energy is at its tightest since 2022 ( Exhibit 5 ). As discussed last week, the effect has been much more pronounced in the average stock than in the cap-weighted index.
So far, Warsh has largely attributed higher long-end yields to stronger growth, competition for capital and geopolitics. We agree that those forces matter, but the question now is whether he also acknowledges that the accompanying tightening in financial conditions is doing some of the Fed’s work. If Warsh acknowledges the restraint that markets have already put in place, there may be less reason for that restraint to persist for an extended period. If he instead keeps the focus on the inflation risks without noting the tightening already delivered, longer-dated yields and economically sensitive stocks could remain under pressure.
Chairman Warsh didn’t offer new guidance on monetary policy at the IMF meetings, which markets interpreted as broadly consistent with prior pricing. With many Committee members emphasizing patience, our economists expect the Fed to avoid surprising markets and remain on hold in October.
The equity market is already telling us where the tightening has been felt most acutely.
Since the middle of August, the equal-weight S&P 500 has declined 4%, while the cap-weighted index has outperformed. In equal-weight terms, Financials, Materials, Consumer Discretionary and Industrials have been sector underperformers—economically sensitive areas that are discounting the higher cost of capital.
The average stock has already taken much of the valuation adjustment, but if rates, the dollar and energy remain restrictive for long enough, the drag can move beyond multiples and into earnings growth. We don't think that is the most likely outcome today. The S&P 500 multiple has already re-tested 19x, breadth has returned to levels near the March lows and earnings remain strong, while our economists expect inflation breadth to narrow.
Bottom line, quality’s breakout is consistent with the mid-cycle transition we have been highlighting, and we continue to favor it. In our base case, a period of more stable long-end yields should allow earnings to remain the primary driver of returns; in the risk scenario where financial conditions stay restrictive, quality has historically also provided relative protection. That does not mean we would avoid cyclicals altogether. The de-rating has created better entry points in higher-quality cyclicals with improving EPS revisions.
Capital Goods stands out in this regard, with revisions breadth improving alongside broader manufacturing backlogs, as we have discussed. For the near term, we are watching how Warsh frames the tightening in financial conditions, whether the relative performance of Financials can stabilize through reporting season, and whether bond volatility continues to settle. We would want to see some follow-through on those measures before declaring that the divergence between the index and the average stock has been resolved.
Quality + Strong EPS Revisions Screen:
- Stock is a member of top 1000 by market cap.
- Stock is in top quartile of quality composite score.
- Stock's NTM EPS has improved from 3 months ago.
- Stock is rated as Overweight by Morgan Stanley analyst.
Last week we discussed the importance of financials relative performance and bond volatility as leading indicators for this correction to end. We have highlighted for months the weakening breadth as normal and expected in these mid cycle transitions. However, the divergence between breadth and the resilience of the S&P 500 remains unresolved and last week's price action suggests this divergence is starting to close from a meeting in the middle of the a falling S&P 500 and improving breadth (Exhibit 7). Ironically, in our view, the S&P 500 is among the higher-quality equity markets in the world which flies in the face of our quality preference at the factor and stock level. With the Nasdaq and Semiconductors taking the brunt of last week's pain, we think this can weigh on the S&P 500 a without breaking the leadership of the quality factor more broadly. It may also lead to better performance of other areas that have already been beaten up unfairly. Bottom line, we are getting closer to the end of the mid cycle and leadership transitions as the S&P 500 catches down. We advise investors to be ready for either a very rapid reconciliation or a slower bleed into November because in our view, that will be the time to add risk rather than shun it.
Following AI's $1.6 trillion Money Trail: Mapping the Revenue Generated by the Capex Boom
Below we highlight key takeaways from our US Equity Strategy & GVAT: Following AI’s $1.6 trillion Money Trail: Mapping the Revenue Generated by the Capex Boom. See the note for more details on index-level impacts, industry commentary, key risks and a single stock screen.
Key Takeaways
- Consensus has SPX capex/sales at 9%+ in 2027/28 vs 6% historically. We estimate AI investment exceeds 60% of 2027 capex across the SPX plus SPCX, NBIS and CRWV.
- This capex flows into supplier sales. We estimate AI investment will lift SPX sales growth by 200bps from 6.6% to 8.6% next year.
- Leases are synthetic capex. We introduce new hyperscaler lease models, increasing next year's total capex estimate by 14%–models available upon request.
- As capex growth moderates into 2028, where do we see opportunity? We screen for companies positioned to benefit from specific bottlenecks and gain market share.
- Our interactive capex map goes beyond identifying AI beneficiaries to quantify revenue exposure by company using MS analysts' insights. Please connect for data.
Summary
- New estimates from our sector teams map $1.6 trillion of 2027 compute investment from hyperscaler and neocloud end customers to single-stock revenues. Using our analysts' stock-level insights, we build a bottom-up estimate of AI capex-related revenue at the index level. Our tracking suggests a 65% absorption rate between AI Capex and S&P 500 companies, directly driving 5% of 2027 SPX sales and ~25% of SPX sales growth. More broadly, we see AI adopters and enablers driving a significant proportion of sales and EPS growth in 2027.
- A growing portion of capex is being financed through leases. Our new lease capex estimates add $195 bn in 2027 and $246 bn in 2028 to headline capex, increasing total 2027 investment by ~14% . Capex forecasts are increasingly not telling the whole story of capital investment because data centers acquired through leases, a form of debt-financed asset purchase, do not show up in cash flow statements as capex or management capex guidance. To address this, we forecast lease capex and incorporate it into our total capex estimates, which now reach $980 bn in 2026 (+~100% y/y) , $1.6 tr in 2027 ( +~60% y/y), and $1.8 tr in 2028 ( +14% y/y). We expect this investment to add an incremental 32 GW of compute capacity next year with an ROIC of 25-50%.
- We expect hyperscaler investment to continue to favor large-cap semis, led by NVDA as revenue per GW scales toward $40 bn, with continued strength in AVGO and memory. Following multiple compression since July, incremental upward revisions to AI capex could address concerns about cycle longevity. Within memory, DRAM remains a key bottleneck, yet international memory stocks trade at ~3x NTM P/E. We see the memory cycle entering its late stage, but not ending, in 4Q26, with moderating price increases and rising supply limiting further earnings surprises. In storage, STX and WDC benefit from HDD demand, and P from adoption of its flash storage technology.
- Increasing chip and server complexity creates opportunities for design software providers and server manufacturers to add more value, providing an earnings driver beyond compute capacity growth. Custom silicon investment is broadening the opportunity for Synopsys and Cadence across design software and IP while increasing system complexity supports Taiwan ODM earnings, with our team modeling 35–40% higher value-add per rack on VR200 versus GB300. In the US, DELL, HPE and SNX (through Hyve) are key beneficiaries across servers, networking and rack-scale infrastructure.
- Power constraints favor onsite generation and high-quality equipment suppliers across Machinery, Power, and Multis. EROC is most levered to hyperscaler spending within Machinery, with CMI and CAT also benefiting from backup and onsite generation demand. We expect upward capex revisions to extend power OEM backlogs at higher margins, benefiting GEV, INIO and BE. Across Multis, electrical and cooling suppliers (VRT, ETN, FPS, HUBB, TT, JCI, CARR) are expanding capacity as modular solutions support faster and simpler deployments amid increasing complexity and labor constraints, positioning suppliers to address customers’ focus on speed.
- With the cash-flow deficit among hyperscalers and neoclouds expected to widen next year, we anticipate financing needs to rise in 2027. We continue to see capacity for the higher-rated hyperscalers to fund their investment plans, even at higher borrowing costs, while lower-rated borrowers are more exposed to financing constraints. We also expect the financing mix to shift as companies raise capital for chips and servers, which lend themselves to shorter-duration, amortizing structures, with greater flexibility to tie financing to contracted demand.
Note: Hyperscaler and neocloud capex figures are cash capex, leases aggregated separately, SPCX inclusive of estimated terafab capex, AMZN AWS capex only. *Includes only gross profit for OEMs and ODMs to capture assembly portion of hyperscaler/neocloud capex - semiconductors and other components included in revenue for those suppliers. +Excludes $30bn of HBM memory sales from MU, $57bn of HBM from Samsung, and $49bn HBM from SK Hynix sold into accelerators to avoid double counting revenues captured in AI chips. Source: Morgan Stanley Research
How much does AI Capex drive S&P 500 revenues?
Based on our mapping, we find that revenue from hyperscaler and neocloud customers makes up 5% of total S&P 500 revenue in 2027 and drives ~25% of S&P 500 y/y revenue growth. With capex growth expected to moderate from ~100% growth in 2026 to ~60% in 2027 and LDD in 2028, the growth contribution decreases but the magnitude of this spending will continue to be impactful. From 2026 to 2027 alone, there's an incremental $700 bn earmarked to be spent on AI. Our estimates can track ~65% of this capex flowing into the S&P 500 with the largest portion flowing to NVDA.
S&P 500 revenues ex-AI direct revenues are already growing a strong 6.6% and with AI capex impact, growth jumps to 8.6%. This underscores that S&P 500 revenue growth remains strong even without the direct impact of AI spending, which serves as a meaningful incremental growth driver.
How much do AI stocks drive the S&P 500 broadly?
For a different approach, we analyzed the breakdown of AI names within the S&P 500 versus non-AI names. Rather than isolating revenues tied to hyperscalers and large neoclouds as in the prior analysis, this approach captures the total sales and earnings contribution of companies broadly classified as AI-related. To determine AI names, we used the latest AI mapping (Global Thematics: The Thematic Map – Scaling Themes into AlphaSignals) and identified 42 names classified as "Core to Thesis" within our Enabler or Adopter categories.
AI earnings are outpacing non-AI earnings. Both the AI cohort and the SPX ex-AI cohort are growing earnings in the double digits, and both cohorts show 2Q26 as the peak quarter for earnings growth ( Exhibit 12 ). This peak could extend depending on how 3Q earnings season plays out relative to expected beat rates. Regardless, we appear to be near peak earnings growth — and markets trade on rate of change. The picture is further complicated by private AI investments showing up in non-operating profit, which skews the 1Q, 2Q, and full-year 2026 earnings picture.
With earnings growth now high but decelerating, market focus may shift toward who is incrementally improving EPS growth into 2027 — a dynamic that favors the AI cohort, where that growth is notably stronger. Taken together, AI remains a dominant theme in 2026's market, and we expect this to continue into 2027. That said, we think leadership will shift away from pure Enablers toward more blended Enabler/Adopters and Adopters, even as the AI theme broadly remains highly prevalent.
AI companies are growing 2027 sales at 28% versus 5% for non-AI companies ( Exhibit 13 ). Sales for the S&P 500 broadly are very strong and in line with historical regressions versus nominal GDP. S&P 500 2027 sales growth is running at 9%; stripping out AI investment (assuming a 65% SPX absorption rate) would bring that down to roughly 7% — still in line with historical regressions of nominal GDP and historical sales growth.
These points support our view that S&P 500 broadening is fully underway: growth would already be strong without AI investment, and AI spending is fueling incremental growth on top of that.
The growth numbers are significantly higher within AI names, but they're working off a lower absolute sales base. Sales for our "AI names" cohort represent 20% of S&P 500 total sales in 2026, rising to 23% and 26% in 2027 and 2028, respectively. The AI names category is heavily skewed by the Magnificent 7, which account for 14 of that 20 percentage points — roughly 70% of the cohort's sales. In terms of sales growth contribution, the AI cohort is driving 60% of 2027 S&P 500 sales growth ( Exhibit 14 ).
AI companies are growing 2027 earnings at 25% versus 4% for non-AI companies ( Exhibit 15 ). S&P 500 net income growth is tracking strongly at 12% in 2027, but that's a step down from 36% in 2026. The comparatively low 4% earnings growth in 2027 for SPX ex-AI is largely due to base effects from 23% growth in 2026 — which was historically very strong for the S&P 500 — and to the removal of some of the fastest-growing tech names from the ex-AI cohort.
Looking at contribution, AI names in the S&P 500 are driving 78% of 2027 earnings growth ( Exhibit 16 ). These AI names also represent a more significant and growing share of index-level net income, making up 37% of 2027 consensus. The Magnificent 7 drive about 75% of earnings within the AI cohort — similar to the sales breakdown.
Earnings Season Chartbook
3Q EPS is expected at +27% Y/Y and sales at +11% Y/Y.
- Financials started with a mixed start. Equities trading was strong while FICC was mixed; investment banking pipelines remain healthy but deal timing is uncertain. Capital pressure was generally manageable, and private credit interest remains strong as lenders scrutinize terms and returns. Fourth-quarter trading activity and the timing of transactions remain uncertain.
- Revisions breadth has cooled but remains positive. S&P 500 earnings revisions breadth (FY2) is +11% heading into the season, down from the late-June peak of +28%. The two weeks into the start of earnings season are seasonally weak for ERB before it historically moves higher once reporting starts. Short-term ERB has been weaker than normal, with Utilities and Comm. Services the strongest and Industrials and Tech the weakest over the last 4 weeks.
- 3Q26 estimates have revised higher into the season. They are up 1.8% over the past three months, against a historical tendency to fall ahead of reporting (2Q26 was +3.8%). The higher bar creates tougher expectations for T+1 price reactions, and early reads are soft. With ~10% of companies reported, the median T+1 reaction is -0.4% (-0.6% relative) and only 41% of reactions are positive.
- Calendar 2026/2027 earnings growth is now 33%/16%. 2026 earnings widely beat expectations in 1H26 (2Q26 EPS surprise of +26% vs. a 4.4% average) and were accelerated by non-operating private AI gains. Consensus sales growth is 11% for 2026 and 8% for 2027.
- Breadth of earnings strength is intact. The median Russell 3000 stock is growing sales by 8% and EPS by 14% through 2Q26.
- Mag 7 vs. SP493. Mag 7 earnings are expected to grow 37% in 3Q26 versus 19% for the SP493, both historically strong. The SP493 overtakes the Mag 7 in 2027 (16% vs. 10%).
- Analyst previews are positive but less so than last quarter. Net KPI previews are +18% and NTM EPS previews are +15%, vs. +27% for NTM EPS last quarter.
Key Charts
Price Reaction
- 3Q is only ~10% reported with the median T+1 price reactions at -0.4% on an absolute basis and -0.6% relative to the broad index.
- Breadth of positive returns is weaker thus far in 3Q26 at 41%.
Surprise
- 3Q Sales surprise at 1.1%, below the historical average of 1.3%.
- 3Q EPS surprise at 3.2%, below the historical average of 4.4%.
- Last quarter (2Q26) EPS surprise was +26%, strongly ahead of the historical average of 4.4%. Non-operating profits from private AI investments skewed 1Q/2Q higher but it was still a historically strong quarter ex-these gains.
Company Guidance & Revisions
- S&P 500 earnings revisions breadth (FY2 estimates) is positive at +11% heading into Q3 earnings season.
- The two weeks heading into EPS season are a seasonally week period. Short-term ERB has been weaker than normal into earnings season and we are monitoring it for first impacts as more names report.
- Utilities and Comm. Services have the most positive ERB in the 4 weeks leading into earnings season while Industrials and Tech have seen the weakest.
Sales
- Consensus expects 10.7% sales growth in 3Q.
- Consensus expects 11% 2026 sales growth and 8% 2027 sales growth
EPS
- Consensus expects 27% EPS growth in 3Q.
- Consensus expects 33% 2026 EPS growth and 16% 2027 EPS growth.
Margins
- Realized (LTM) EBIT margins are at 20.1% today and are expected to be 20.7% for FY2026 and 22.7% for FY2027.
- Realized (LTM) net margins are at 15.8% today and are projected at 16.3% for FY2026 and 17.1% for FY2027.
- 3Q consensus margins are at 15.8% for the S&P 500.
3Q Systematic Earnings (Full MS Coverage)
Summary:
- Previews: KPI previews are +18% vs +24% net positive last quarter and NTM EPS previews are +15% vs +27% last quarter on projected NTM impacts. Previews are currently net positive across all sectors with the exception of Industrials.
3Q Preview Tracking
Our Morgan Stanley Analysts submit a systematic earnings preview where they forecast two items: 1) KPI and 2) NTM EPS impact. Below we summarize the sector and industry summaries of these previews. The population is the Morgan Stanley US coverage, which is broader than the S&P 500 and roughly similar to the Russell 1000.
Methodology: Each analyst identifies the key performance indicator (KPI) for each individual stock and it may differ from NTM EPS. For example, the KPI for a Financial Services or Semiconductor company may be net assets or GPU demand and this may differ from NTM EPS (although, it is often similar). Each analyst chooses one of five forecasts for both KPI and EPS: very negative, negative, in-line, positive, or very positive. We roll up these previews into a net score by taking (positive-negative)/total reactions.
Key Performance Indicator (KPI) Previews
EPS Previews
Screen 1: Favored Stocks with Positive Surprise Expected
We screen for OW-rated stocks that are expected to beat on key KPI, or to see N12M consensus EPS revised higher post results.
Screen 2: Less Favored Stocks with Negative Surprise Expected We screen for stocks that are rated Equal Weight or Underweight by our analyst and are expected to miss on their key KPI. Note that the KPI may be different than N12M EPS.