Data Center Stocks Up 80% Since 2023, Cramer Argues

The debate over whether it is "too late" to buy data center and AI infrastructure equities regained traction after Jim Cramer's Investing Club column, published on CNBC on May 10, 2026, argued that selective exposure to the winners remains justified. Cramer highlighted persistent demand for generative-AI compute and cited large hyperscaler commitments as the primary driver of continued capacity expansion (CNBC, May 10, 2026). Market performance has already priced a significant portion of that growth: a basket of listed data center incumbents and AI hardware suppliers has rallied in excess of 60-100% since January 2023, according to consensus price series cited by market data providers. That advance has lifted valuations; the question for institutional investors is whether near-term fundamentals and longer-term structural growth justify a further premium versus broad benchmarks such as the S&P 500. This analysis dissects the underlying data, compares returns and fundamentals across peers, and outlines the key risks and catalysts institutional allocators should weigh.

Context

The data center and AI-infrastructure complex sits at the intersection of cloud software adoption, hyperscaler capex cycles and a secular upgrade to GPU-heavy architectures. In his May 10, 2026 column, Jim Cramer argued that the apparent late-cycle rally in stocks such as Equinix (EQIX), Digital Realty (DLR) and chipmaker NVIDIA (NVDA) reflects durable demand rather than speculative excess (CNBC, May 10, 2026). Hyperscalers — notably Amazon (AMZN), Microsoft (MSFT) and Alphabet/Google (GOOGL) — have publicly committed multi-year capex plans; for example, public filings and analyst notes in 2025–26 reflected multi-year cloud infrastructure allocations that underpin procurement of servers, networking and datacenter space. These commitments have resulted in a multi-year re-rating for operators and vendors, but the pace and distribution of spend are uneven, concentrated among a handful of hyperscalers and top-tier AI chip providers.

From a market-performance perspective, a composite index of pure-play data center REITs and AI hardware suppliers outperformed the S&P 500 in the 2023–2025 window. Benchmarked returns vary by segment: Equinix and Digital Realty industry peers recorded mid-single-digit to low-double-digit revenue growth in sequential quarters through 2025, while GPU-led vendors posted much higher growth due to AI demand. Even so, the outperformance has a clear comparator: the Nasdaq 100 and S&P 500 also rallied during this period, so relative valuation expansion remains the critical determinant of future returns. Institutional investors assessing entry points should therefore map expected incremental cash flows from new AI workloads against current multiples and the timing of hyperscaler procurement waves.

The structure of demand matters. Unlike previous waves where server refreshes were broad-based, the AI cycle is highly concentrated around specialized GPUs and custom acceleration hardware. That concentration causes asymmetric returns: a few winners (hardware suppliers and specialist colocation providers with tight locations) capture outsized economics, while broader-capacity providers face margin pressure from competition and commoditization. This structural feature underpins both the bullish thesis — sticky high-margin AI workloads — and the cautionary case — single-customer concentration and technology obsolescence risk.

Data Deep Dive

Specific datapoints illuminate the current state. CNBC's May 10, 2026 column summarises Jim Cramer's view that hyperscaler demand continues to expand (CNBC, May 10, 2026). Industry research firms have published supporting metrics: Synergy Research Group's cloud infrastructure spending reports (2025–2026) show sustained year-on-year growth in data center and cloud capex, with cloud providers increasing infrastructure allocations by high-single to low-double digits in 2024–25 (Synergy Research Group, 2025). Public company results parallel that aggregate trend: large data center REITs disclosed sequential revenue growth in most quarters through 2025, with some reporting mid-to-high single-digit organic growth (company quarterly filings, 2025–26).

Hardware vendors tied to AI compute also reported outsized performance in recent quarters. NVIDIA surpassed $1 trillion in market capitalisation in 2023 and remained one of the largest market-cap companies into 2026, driven by accelerating data-center GPU revenue (public market data, Bloomberg consensus, 2024–26). That scale has translated into material production priorities for foundries and supply chains. Comparatively, legacy server and networking vendors have seen more muted growth; their revenue growth often tracks overall enterprise IT spending, which was flat-to-modest in the same period. These contrasts — very strong growth at GPU and hyperscaler-facing vendors versus modest growth for broader IT suppliers — create a dispersion that investors must model explicitly.

Valuation metrics reflect those divergent fundamentals. As of early May 2026 market snapshots, pure-play data center REITs trade at occupancy- and cash-flow-driven multiples that are below high-growth hardware peers, while GPU and AI chip suppliers trade at higher revenue multiples reflecting earnings leverage and scarcity value. For institutional allocators comparing year-on-year (YoY) growth, data center REITs posted roughly mid-single-digit YoY revenue gains in 2025 while AI hardware vendors posted YoY growth often in the 30–100% range across sequential quarters (company reports and market research, 2025). Relative to the S&P 500, which recorded single-digit EPS expansion in the same time window, the AI hardware cohort has materially outpaced earnings growth.

Sector Implications

The current environment favors differentiated scale and proximity to hyperscaler demand. Colocation operators with direct connectivity to major cloud regions and firms offering AI-optimised racks command pricing power, while generic small-footprint operators face pricing pressure. For example, providers that secured long-term commitments from hyperscalers report higher contracted revenue visibility and lower churn risk compared with peers relying on enterprise customers. This segmentation implies that a broad, undifferentiated allocation to the data center complex will underperform a targeted selection of hyperscaler-facing names over the medium term.

Supply-chain and capacity constraints also reshape competitive dynamics. Foundry allocations for GPUs and specialised accelerators are finite in the near term; that creates bottlenecks for certain hardware vendors and can delay broader enterprise uptake of AI workloads. Conversely, data center operators with room to expand capacity adjacent to hyperscaler campuses or with pre-approved grid and permitting paths can monetise higher per-rack pricing in tight cycles. The net effect is that growth and returns are rewarded to players with pre-existing scale, strategic locations and long-term commercial contracts.

Comparative performance versus peers and benchmarks highlights the portfolio construction challenge for institutions. An allocation that mimics a broad technology index (e.g., Nasdaq 100) will expose investors to the biggest winners but also to sector cyclicality. A focused allocation to data center REITs may underdeliver absolute returns but provide income and lower volatility. These trade-offs should be explicit in any investment policy statement guiding allocations to the data center sector or related thematic buckets.

Risk Assessment

Concentration risk is chief among the hazards. Hyperscaler concentration means that a single procurement pause or a shift in architecture (for example, adoption of new accelerators that render existing GPUs less valuable) can disproportionately affect revenue streams for both hardware vendors and specialised colocation providers. Historical precedents in technology cycles show steep drawdowns when demand re-prices; investors should model downside scenarios that include a 20–40% correction in demand growth and stress-test earnings and cash-flow coverage metrics accordingly.

Valuation risk compounds the concentration problem. Much of the outperformance through 2023–25 has been driven by forward-looking expectations of multi-year adoption curves; if those expectations slip — due to macro re-tightening, slower enterprise adoption, or better-than-expected efficiency gains that reduce physical rack requirements — multiples could compress rapidly. Interest rate sensitivity remains material for data center REITs: higher discount rates increase the cost of capital for capex-heavy expansions and compress NAVs. Credit and financing metrics in 2025–26 show elevated scrutiny from lenders, and that creates refinancing risk for heavily levered operators.

Operational risks include energy pricing, grid availability and permitting delays. Data centers are capital- and energy-intensive; volatile power costs or localized grid constraints can impair utilisation and margin. Moreover, geopolitical risks — export controls on AI chips, sanctions regimes affecting supply chains — can create abrupt supply shocks. These operational variables warrant specific scenario analysis and a focus on contractual protections such as minimum revenue guarantees or take-or-pay arrangements in commercial leases.

Fazen Markets Perspective

From a contrarian institutional lens, the most actionable distinction is between narrative winners and contractually protected cash flows. Cramer's view, as relayed on May 10, 2026, emphasises the narrative — that the AI cycle is large and persistent (CNBC, May 10, 2026). Our view adds a second axis: quantify the portion of future cash flows that are contracted or highly probable versus those that depend on incremental, discretionary hyperscaler spend. Where a name can demonstrate >60–70% contracted or committed utilisation over multi-year horizons, valuation premiums are easier to justify.

Another non-obvious insight is that hardware scarcity can compress investment opportunities for new entrants but also accelerate client consolidation. Scarcity benefits incumbent suppliers in the short term but can spur hyperscalers to vertically integrate or to sign exclusive supply deals that reduce open-market opportunities. Institutional investors should therefore discount implied TAM (total addressable market) growth assumptions for potential share losses to vertically integrated hyperscalers over a 3–5 year horizon.

Finally, timing trumps theology. Headlines about perpetual AI-driven growth are less useful than cadence of capex spend, quarterly bookings and visibility on multi-year contracts. Institutional buyers should prioritise companies that disclose backlog, contracted revenue, or long-term purchase agreements and should consider staggered entry points tied to observable milestones such as hyperscaler bookings or major foundry allocation announcements. For more on connected themes and modelling frameworks, see our AI infrastructure and topic coverage.

Outlook

Forward-looking indicators point to continued, but more selective, upside potential. If hyperscaler demand maintains a mid-to-high single-digit compound annual growth rate for physical rack and specialised GPU deployments over the next 24 months, high-quality hyperscaler-facing operators and scarcity-priced hardware vendors can deliver outsized returns relative to broader benchmarks. Conversely, if demand growth reverts to pre-AI refresh levels, rerating risks increase substantially. The path dependency of capex — the lead time for new data center builds and the multi-quarter production cycles for chips — means that visible changes in hyperscaler procurement are likely to transmit to earnings only with lags; investors should track booking and backlog data closely.

For portfolio construction, a barbell approach that pairs yield-oriented, lower-volatility data center REITs with a targeted sleeve of high-conviction AI hardware suppliers can reconcile growth and risk. Position sizing should reflect concentration risk, financing profiles and contract visibility. Active monitoring of quarterly bookings, public procurement disclosures, and supplier foundry allocation reports will remain the most reliable real-time indicators of where to add or reduce exposure.

FAQ

Q: How material is hyperscaler concentration to portfolio risk? Answer: Hyperscaler concentration is substantial; top three cloud providers account for the majority of incremental hyperscale capex and AI procurement. A single pause or reprioritisation by one hyperscaler can reduce near-term demand for colocation space and GPUs by high-single-digit to double-digit percentages, depending on the vendor's customer mix. Historical examples (enterprise cycles and cloud refresh waves) show that concentrated demand can amplify downside volatility and elongate recovery times.

Q: Are valuation premiums justified for AI hardware names? Answer: Valuation premiums are justified to the extent they reflect durable margins, supply scarcity and multi-year contractual visibility. If a hardware vendor can demonstrate multi-year purchase agreements, sustainable gross margins above peers, and preferential foundry allocation, premiums are easier to defend. However, premiums become vulnerable if product cycles are shortened by architectural changes or if hyperscalers internalise procurement.

Q: What are practical risk-management actions for institutions? Answer: Use staggered entry, require contractual visibility (backlog, long-term contracts), and stress-test cash flow models under demand corrections of 20–40%. Hedging interest-rate exposure and scrutinising balance-sheet liquidity and refinancing timelines are also essential for capex-heavy operators.

Bottom Line

Cramer's May 10, 2026 column underscores a genuine growth story, but institutional deployment requires disciplined, data-driven differentiation between contracted cash flows and narrative-driven expectations. Selectivity, contractual visibility and scenario-based sizing are the decisive tools for institutional investors.

Disclaimer: This article is for informational purposes only and does not constitute investment advice.