Technology (XLK) vs Utilities (XLU)

XLK holds 76 stocks representing the technology sector of the S&P 500. April 2026 top holdings: NVIDIA 15.14 percent, Apple 12.53 percent, Microsoft 9.64 percent, Broadcom 6.22 percent, Micron Technology 4.12 percent. The fund is highly concentrated in mega-cap names: top 5 represent 47 percent of assets. AUM approximately $70 billion, expense ratio 0.08 percent. XLK is the primary US-listed pure technology sector exposure.

XLU holds approximately 30 stocks in the utilities sector. Top holdings: NextEra Energy 14.01 percent, Southern Company 7.23 percent, Duke Energy roughly 7 percent, Constellation Energy roughly 6 percent. AUM approximately $20 billion, expense ratio 0.08 percent. XLU has historically been the cleanest defensive sector exposure with high dividend yields (3 to 4 percent) and rate-sensitive bond-like behavior. The sector's composition has shifted in 2024 to 2026 toward utilities with substantial nuclear, renewable, and data center exposure, changing the fundamental character of the fund.

For three decades the XLK/XLU ratio captured the cleanest growth-defense rotation signal. Tech is long-duration: most of its earnings come from future growth that gets discounted heavily when rates rise. Utilities are short-duration with stable, regulated cash flows that are valued like bonds. When risk appetite is strong and growth expectations rising, XLK leads XLU. When risk appetite weakens and recession risk rises, XLU leads XLK as investors rotate to defensive yield.

The ratio captured the dot-com peak and crash (XLK/XLU peaked above 10 in 2000, fell to 3 by 2002), the 2008 to 2009 financial crisis (XLK/XLU fell from 4.5 to 2.5), and the 2022 inflation reset (XLK/XLU fell from 5.5 in late 2021 to 3.8 in October 2022). In each cycle, utility outperformance preceded or coincided with broader equity stress, and tech outperformance signaled risk-on regimes. The relationship was mechanically rate-driven: utilities benefited from falling rates, tech suffered; rising rates flipped the dynamic.

Starting in 2024, AI data center demand fundamentally altered XLU's character. Hyperscalers (Microsoft, Google, Meta, Amazon, Oracle) committed approximately $500 billion in annual AI infrastructure spending. Data centers consume vast electricity: a single hyperscale data center uses 100 to 300 megawatts continuously, equivalent to 100,000 to 300,000 homes.

US data center electricity consumption is projected to reach approximately 9 percent of total US electricity demand by 2030, up from 4 percent in 2024. The sudden surge in power demand reversed three decades of flat US electricity consumption growth. Utilities like Vistra, Constellation, and NextEra have signed multi-billion dollar power purchase agreements with hyperscalers. NextEra and Constellation have been among the strongest stock performers in any sector during 2024 to 2026. The result: XLU has gained roughly 20 to 25 percent over the past 12 months, similar to XLK's gains, despite being classified as defensive. The ratio has therefore been less informative than in past cycles because both sectors are riding the same AI capex narrative.

The structural shift can be visualized through the two largest constituents. NVIDIA at 15.14 percent of XLK earned $194 billion in fiscal 2026 data center revenue (up from $3 billion annualized pre-2023). The revenue comes from selling GPUs and accelerator silicon to hyperscalers building AI infrastructure. NextEra Energy at 14.01 percent of XLU has signed power purchase agreements with hyperscalers totaling roughly 30 GW of new commitments, supporting a multi-decade capex pipeline.

The two stocks are therefore both pure plays on the same underlying AI infrastructure cycle. NVIDIA captures the chip side of the cycle; NextEra captures the power side. When AI capex expectations rise, both stocks benefit. When AI capex disappoints, both fall. The cap-weighted XLK/XLU ratio is therefore less sensitive to AI capex shifts than its construction would suggest, because the dominant constituents in both legs are correlated rather than opposing. For pure rotation signals, equal-weighted versions of the indices or excluding the AI-exposed names would produce cleaner reads.

Despite the AI complication, rate sensitivity still matters at the sub-sector level. Utilities are heavily indebted (typical 50 to 60 percent debt-to-equity) and refinance regularly, so rising rates compress earnings via higher interest expense. Falling rates deliver opposite benefits. The 2022 to 2024 rate cycle compressed utilities significantly: XLU fell roughly 20 percent in 2022 alongside rising 10-year Treasury yields, before recovering with the 2024 to 2025 rate cuts.

Tech's rate sensitivity operates through valuation multiples: long-duration cash flow streams discount heavily when rates rise. The 2022 episode saw XLK fall 28 percent at the trough as Treasury yields rose from 1.5 percent to 4.3 percent. Both legs of the pair are therefore rate-sensitive, but in different directions during normal regimes. The April 2026 environment with the Fed at 3.75 percent and 10Y yields at 4.30 percent is moderately restrictive; further rate moves in either direction would affect both XLK and XLU but with potentially asymmetric magnitudes.

The XLK/XLU ratio peaked in late 2007 at approximately 4.0, then fell to 2.0 by November 2008, a 50 percent decline that closely tracked the broader bear market. The ratio break occurred earlier than the formal recession start (December 2007) and earlier than the S&P 500 peak (October 9, 2007). The ratio declined sharply from October 2007 through March 2009 alongside the financial crisis.

The 2007 to 2008 episode demonstrated the indicator at its purest. XLU outperformance was driven by simultaneous Fed cutting (rates from 5.25 percent to zero) and risk-off rotation. Tech faced a double headwind: weakening earnings outlook and credit-cycle pressure on growth multiples. The recovery in 2009 to 2010 saw the ratio re-expand from 2.0 to 3.5 as growth confidence returned. The cycle was the cleanest example of the XLK/XLU ratio functioning as a growth-defense rotation signal.

COVID and the subsequent AI boom produced a unique XLK/XLU trajectory. The ratio held in a 4 to 5 range through early 2020, then surged to nearly 6 by late 2021 as tech benefited from work-from-home dynamics and the Fed's zero-rate policy. The ratio fell to 3.8 in October 2022 during the inflation-driven repricing, reflecting both XLK weakness (rates pressure) and XLU strength (defensive rotation).

From late 2022 the ratio recovered to approximately 4.0 by 2024, then expanded to 3.8 to 3.9 through 2024 as XLK benefited from AI capex acceleration. The unusual development: XLU also rose substantially through 2024 to 2026, narrowing the typical growth-defense gap. The combined effect: XLK/XLU ratio in April 2026 at 3.35 is below the 2024 average despite XLK being near all-time highs, because XLU has caught up. The signal is that growth-defense rotation has been replaced (or supplemented) by AI infrastructure rotation.

XLK at $155 is near its 52-week high of $156, having gained roughly 60 percent over the trailing 12 months. XLU at $46 has gained approximately 25 percent, also near recent highs. The 3.35 ratio sits below 2024 readings but above 2009 to 2017 averages of 2.5 to 3.0. The current reading reflects both AI tailwinds and the unusual situation where utilities are not behaving defensively.

The Iran war effect on the pair has been minor. XLK has been supported by continued AI capex acceleration; XLU has been supported by continued data center demand and modestly lower rates. Neither leg has shown traditional growth-defense divergence in response to the conflict. If the Iran war ends with Hormuz reopening, both XLK and XLU likely consolidate; if it escalates and produces broader equity stress, the traditional defensive XLU character could re-emerge if AI capex narrative weakens. The pair currently has lower information content than in past decades, but the AI capex interpretation is the most useful framework for reading it.

Historic divergence patterns. First, growth shocks: tech-specific issues (2000 dot-com, 2022 rate reset) produce XLK underperformance while XLU holds. Second, rate cycles: rising rates pressure utilities more than tech in short windows, then tech more than utilities in longer windows. Third, regulatory shocks: tech antitrust actions (Microsoft 2001, ongoing FTC actions) produce XLK-specific declines without XLU response.

New divergence patterns post-2024. AI capex disappointments (large hyperscaler capex cuts) would hit both XLK and XLU together because both are now AI-exposed. Defensive rotation during recession risk would lift XLU but with a lag versus past cycles, since AI capex is supporting utility revenues even in slower economic environments. The pair's signal value has therefore shifted from growth-defense rotation toward AI capex sensitivity. Watch for hyperscaler quarterly capex commentary as the dominant input over the next 2 to 3 years.

For practical use: track XLK/XLU ratio versus its 200-day moving average. Above 3.5 indicates aggressive AI capex pricing. Below 3.0 historically signaled defensive rotation; in the post-2024 era a similar level might signal AI capex disappointment rather than traditional risk-off.

Long XLK / short XLU is the cyclical risk-on bet, currently neutral-to-positive at 3.35. Long XLU / short XLK is the defensive bet, less compelling than in past cycles given AI capex tailwinds for utilities. A more useful 2025 to 2027 framework: track XLK/XLU together as an AI capex composite, with both legs benefiting when capex expectations rise. Pair the ratio with hyperscaler capex commentary, NVIDIA earnings, and electricity demand projections (EIA Short-Term Energy Outlook is a useful input). The traditional growth-defense interpretation will likely return when the AI capex cycle moderates, possibly 2027 to 2028.