AI is driving a historic, still‑ramping capex wave—on the order of $5–8tn globally to 2030 (BlackRock p.4) and $4.7tn 2026‑30 (UBS p.15)—with hyperscalers' annual spend expected to reach $500bn by 2026 (Barclays p.42) and ~1% of global GDP by 2030 (UBS p.16). Across houses, AI‑linked investment is now a macro pillar of US growth, in some cases contributing as much as consumption to 2025 GDP (KKR p.9) and tripling capex's normal growth contribution in 2026 (BlackRock p.5). But the build‑out is increasingly funded by debt rather than pure cash flow (Goldman p.27; T. Rowe p.7) and is colliding with hard constraints in power, grids, equipment, and critical minerals (BlackRock p.9; Goldman p.37; Brookfield p.6–7; UBS p.18–19). Consensus expects monetization to lag capex yet ultimately be large—AI revenues of $1.5–3.1tn by 2030 (UBS p.15–17) and potentially $10tn productivity gains over a decade (Brookfield p.5, p.20)—but is split on whether 2026 marks an early‑cycle surge or a maturing, riskier phase. The most decision‑critical numbers are: hyperscalers' capex load (60–70% of OCF, trending toward 70–78% by 2026: KKR p.4; UBS p.15); data centers' share of power (~3–4% of global electricity by 2030: Barclays p.37; 8–9% of US demand by 2035: Goldman p.45, UBS p.18); and the revenue needed—$1.7–2.5tn annually—to justify upper‑end capex at a 9–12% IRR (BlackRock p.6).
These houses frame AI as a multi‑decade capex and infra super‑cycle:
Constraints are framed primarily as investable bottlenecks. Power, grids, and copper shortages are linked directly to upside for utilities, grid owners, renewables, storage, and critical‑mineral producers (Brookfield p.7, p.12–15; HSBC p.12–14; UBS p.18–19; JPAM p.25–27). Private capital is cast as the key enabler, with infra debt offering "durable, inflation‑resilient yields" (Brookfield p.30; BlackRock p.13; T. Rowe p.12–13; JPAM p.6).
These reports agree AI capex is vast and still ramping, but focus on monetization, leverage and overbuild risk:
US‑centric capex, Asia hardware leverage: Most capex is US‑anchored (BlackRock p.4; UBS p.15; KKR p.4), but Asia dominates chips and data‑centre equipment, especially TSMC and broader Asian fab/spend (HSBC p.9–10; Goldman p.8; Brookfield p.5–7).
China's different model: KKR notes China's AI investment at 0.5% of GDP (direct) vs 5% US broad, with 8–10% capex‑to‑revenue vs 16–20% for US hyperscalers, implying more capital‑efficient but smaller macro punch (KKR p.50). UBS and HSBC see Asia (China, Korea, Taiwan, ASEAN, India) as key beneficiaries of AI infra via hardware, fabs, and regional data centers (HSBC p.9–10; UBS p.18).
AI capex is a macro‑scale, still‑ramping investment wave—on track for $500–600bn in 2026 alone and ~1% of global GDP by 2030 (Barclays p.42; UBS p.15–16)—but its continuation beyond 2026 hinges on two variables: whether monetization catches up to the capex curve, and whether power/grid and financing constraints remain manageable rather than binding. For investors, that argues for staying exposed to AI but tilting toward "picks and shovels" with tangible cash‑flow backing (power, grids, data centers, semis, critical minerals) and balance‑sheet discipline, while treating hyperscaler‑led AI growth as a powerful but increasingly constrained and funding‑sensitive engine rather than a risk‑free perpetuity.
| Source | Content |
|---|---|
| Brookfield Outlook | AI as a general‑purpose technology is expected to unlock up to ~$10T in productivity gains over the next decade, requiring ~$7T of infrastructure investment across the AI value chain, including data centers, power, semis and fiber [p.5, p.20, p.30]. Global hyperscaler capex is projected to rise ~50% from ~$270B in 2024 to ~$400B in 2025, framed as an early phase of a longer AI‑driven infrastructure “supercycle” extending to at least 2040 within >$100T of global infrastructure needs [p.5–7]. |
| Goldman Outlook | AI capex is large, highly concentrated in a few hyperscalers that now account for ~27% of S&P 500 capex, and has been underestimated by analysts every quarter for two years, with realized growth far above initial forecasts [p.8, p.15]. Hyperscalers’ AI investment is expected to remain “durable into 2026,” with continued upside risk to the durability of the broader AI trade [p.8, p.15]. |
| Blackrock Outlook | AI corporate capital spending ambitions of $5–8 trillion globally through 2030, mostly in the U.S., could make this the fastest and deepest capital buildout versus past waves like steam, electricity and ICT, with capital deepening reaching similar size “in half the time” [p.4]. Investment is heavily front‑loaded into compute, data centers and energy infrastructure over 2025–2030, with a transformation horizon modeled through 2040 as high capex intensity persists [p.4]. |
| Barclays Outlook | AI data‑centre and infrastructure spending is “among the largest infrastructure build-outs in modern history,” with hyperscalers’ annual capex more than doubling since 2023 and combined Meta/Amazon/Microsoft/Alphabet/Oracle capex projected to exceed $370bn in 2025 and $500bn in 2026, while OpenAI has announced investments “worth more than $1 trillion” [p.5, p.41–42]. AI capex already drives a significant portion of US GDP growth and is expected to keep supporting activity into 2026, although the pace of new commitments should slow as the focus shifts to executing announced projects [p.5, p.42]. |
| Goldman Outlook - Summary | AI-related capex is a major, ongoing investment wave dominated by five hyperscalers (Amazon, Google, Meta, Microsoft, Oracle) that already account for ~27% of S&P 500 capex, with their AI capex “extending into 2026” and capex “projected to rapidly ascend” as the focus shifts from infrastructure to applications [p.3–4, p.17]. AI capex is described as a central growth engine for earnings and for the US economy over 2025–2026 [p.3–4, p.17]. |
| JPM Outlook | AI investment is already macro‑significant, with data center capex around 1.2–1.3% of U.S. GDP and hyperscaler/cloud AI capex projected to rise from $80bn in 2019 to $588bn in 2026E, including very high growth years of 60% and 72% in 2024–25E [p.3–4]. The boom is framed as an ongoing, front‑loaded buildout through at least 2026, underpinning a structural transition rather than a short‑lived bubble [p.3–4, p.7]. |
| JPAM Outlook | AI and the “electronification of everything” fuel a once‑in‑a‑generation capex super‑cycle in power and digital infrastructure, with core infrastructure capex projected to outpace depreciation for the first time this century and to do so “for the foreseeable future” [p.3, p.24–25]. Power demand growth of 2%–4% annually across OECD markets and 1.7%–3.2% in the U.S. into the 2040s implies a multi‑decade AI‑linked buildout, though AI‑specific capex is not quantified as a share of total investment or GDP [p.25]. |
| HSBC Outlook | Exponential AI and cloud adoption is driving a multi‑year “rapid buildout of data centres,” underpinned by a strong, under‑appreciated capex cycle in the US and Asia [p.3, p.9–10, p.12, p.18–19]. McKinsey’s estimate of USD 5.2trn in data‑centre investments needed by 2030, alongside a 10.5% annual growth in the data‑centre market to USD 622bn and Asia Pacific’s 13.1% capacity CAGR to 2030, anchors the scale and duration of this wave [p.9–10, p.12]. |
| MS - 2026 US Equities Outlook - The Rolling Recovery Is Here | AI capex is a powerful, ongoing secular tailwind driving a multi‑year “AI capex cycle” that began with narrow mega‑cap leadership and is now broadening to the S&P 493 [p.8, p.12–15, p.37–38]. The key explicit horizon is 2026–27, when AI‑driven margin uplift is modeled and when AI is expected to drive notable earnings broadening, implying the current capex wave persists at least through that window [p.14, p.22, p.54]. |
| KKR 2026 Outlook | AI-related capex is already “massive,” with hyperscalers reinvesting 60–70% of operating cash flow into capex and projected to peak near 78% by 4Q26, while the Magnificent 7’s capex+R&D is expected to reach ~$875bn in 2026, nearly 29% of all U.S. tech/IP investment [p.4–6]. In 1H25, AI-related software and information processing equipment contributed as much to U.S. GDP growth as consumption, and GPU sales/tech capex trends suggest elevated AI/data center capex persisting into 2026–27 [p.9, p.35–36]. Global data center infrastructure capex is estimated at nearly $7tn by 2030, indicating a multi‑year buildout with boom‑bust risk [p.26]. |
| Stifel Outlook | AI‑driven hyperscaler capex produced a “massive” 2025 surge that significantly boosted Big Tech earnings and Cyclical Growth and is expected to carry over into early 2026, but the y/y growth rate of hyperscaler capex “has likely peaked,” implying a fading macro impulse as GDP is driven by the period‑on‑period change in investment [p.14, p.17, p.25]. Fixed investment, at ~19% of GDP, was lifted by hyperscaler capex enough to offset slowing consumption in 2025, but cannot dominate the macro outlook beyond this late‑cycle 2025–26 window [p.10, p.25]. |
| RIC 2026 BAML | AI-related capex by hyperscalers is historically large, with capex running at ~60% of operating cash flow over the last 12 months—well below the ~140% OCF peak of early‑2000s telcos but enough to make hyperscalers more capital‑intensive than oil majors (72% vs 49% capex/OCF) [p.6–7]. The current AI capex wave is framed as a 2024–2026 buildout, during which tech shifts structurally toward more asset‑heavy business models, with returns likely to diminish as leverage and cost of capital rise over time [p.6–8]. |
| TRowe Outlook | AI investment is at “full throttle,” with hyperscalers running “aggressive, multibillion‑dollar annual capex programs” for data centers, GPUs, cloud capacity, and AI supercomputing hubs that have already “significantly boosted U.S. growth in 2025” and are expected to be reinforced by 2026 fiscal incentives [p.3; p.6]. AI data center chip TAM is projected to grow from around USD 200 billion in 2025 to USD 1 trillion in 2030 (AMD), implying a rapidly expanding, multi‑year hardware buildout [p.7]. AI‑linked investment is a major driver of the 22.5% growth in information processing equipment versus much weaker growth in other categories, underscoring its outsized and ongoing role in capex [p.14]. |
| UBS Year Ahead | Robust AI capex has already exceeded prior estimates threefold over two years and is projected to total USD 4.7tr globally between 2026–2030, with USD 2.4tr already planned from >40 announced projects in 2025 [p.15]. Annual AI capex is projected at USD 571bn in 2026 and to grow at a 25% CAGR in 2025–30, reaching USD 1.3tr (~1% of global GDP) by 2030—below historical investment booms of 1.5–4.5% of GDP, suggesting the cycle still has headroom [p.15–16]. |
| Source | Content |
|---|---|
| Brookfield Outlook | AI‑led automation is projected to generate “over $10 trillion” in economic productivity gains in the next decade, with value creation linked to capital‑intensive infrastructure buildout that enables efficiency gains across nearly all sectors [p.5, p.20]. Industrial and financial sectors are highlighted as key monetizers through cost reduction, labor‑shortage mitigation, supply‑chain optimization, and improved underwriting, loss forecasting and fraud detection [p.19–20]. |
| Goldman Outlook | Return on investment visibility for AI capex remains low, making it uncertain whether massive infrastructure spend will be justified by future revenues or productivity gains, especially given risks of cannibalizing existing businesses [p.8, p.15, p.22, p.27]. At the firm level, AI is already significantly streamlining workflows and decision-making, particularly in PE/VC-backed companies, though the aggregate macro productivity impact is not quantified and remains uncertain in timing and magnitude [p.33]. |
| Blackrock Outlook | Upper‑end AI capex (~$8 trillion) would require $1.7–2.5 trillion in incremental annual revenues through 2030 to deliver a 9–12% IRR, yet broker forecasts for hyperscalers’ revenues fall short of this, implying new AI‑driven revenue pools must emerge [p.6]. A 1.5 percentage‑point productivity‑driven uplift in U.S. growth could generate roughly $1.1 trillion a year in extra economy‑wide revenues—enough at the macro level—but there is no guarantee those gains accrue to AI builders, making monetisation and value capture uncertain [p.5][p.6]. |
| Barclays Outlook | AI is expected to enhance productivity and profitability across many industries over time, but the transformation is gradual and non‑linear, with significant rotations and a risk that current capex and valuations run ahead of realised returns [p.4–5, p.14, p.40–43]. Equity gains have been driven largely by multiple expansion rather than earnings, and there is explicit concern that, akin to the telecom overbuild, huge AI infrastructure investments whose GPUs have a useful life of only ~3 years may not earn adequate returns if monetisation or adoption is slower than expected, leading mainly to multiple compression rather than outright business‑model failure for leading firms [p.14, p.41–43]. |
| Goldman Outlook - Summary | Sustainability of AI capex is tied to “free cash flow and potential return on investment,” with particular focus on whether hyperscalers can “ultimately earn a return on investment” and whether AI enhances or cannibalizes their core businesses [p.10, p.16–17]. AI applications and monetization are highlighted as the crucial next phase after infrastructure build-out [p.4, p.17]. |
| JPM Outlook | AI capex is already translating into monetized demand for AI hardware, cloud services and software, with tech sectors delivering 36% of S&P 500 earnings and 56% of its capex growth over the last 12 months and Mag 7 EPS expected to grow 20.3% in 2026 vs 11.3% for the S&P 493 [p.7]. Economy‑wide productivity gains are “too soon” to see in the data even though consumers appear to benefit and business adoption is rising (9% integrating AI in production, 44% paying for AI platforms), leaving aggregate ROI and productivity uplift still uncertain at the macro level [p.3, p.7]. |
| JPAM Outlook | Returns are framed as higher, more stable cash flows for utilities and infrastructure owners, driven by long‑term contracts with creditworthy, relatively price‑insensitive customers like hyperscalers and industrials, underpinned by capital scarcity and regulated, inflation‑linked frameworks [p.26–27]. Monetisation runs through increased demand for power and infrastructure services rather than quantified AI productivity gains [p.6, p.25–27]. |
| HSBC Outlook | AI is already showing up in earnings and productivity, with US equity returns driven mainly by earnings growth (Mag7 and broader market) rather than P/E expansion and OECD estimates that AI alone can add 1–2.5% to labour productivity over 10 years [p.4, p.19]. Concerns about ROI and potential delays in AI earnings are acknowledged as sources of volatility, but double‑digit EPS growth forecasts (e.g., S&P 500 13% in 2026; Mag7 16%) and rapid monetisation in areas like China’s AI cloud (122.4% y/y in H1 2025) underpin the view that AI capex is broadly justified and not a bubble [p.4–5, p.9–10, p.19]. |
| MS - 2026 US Equities Outlook - The Rolling Recovery Is Here | Momentum around AI adoption is tied to “significant ROI and efficiency gain opportunities,” with AI modeled to add 40 bps to S&P 500 net margins in 2026 and 60 bps in 2027, upgraded from prior assumptions on growing evidence of quantifiable benefits [p.14, p.22, p.48–49]. Transcript analysis of ~7,400 earnings calls shows rising incidence of measurable AI impacts (15% of S&P 500 and 24% of AI adopters in 3Q25), with most realized benefits still skewed toward productivity and cost reduction rather than revenue growth, though revenue/customer‑growth use cases are emerging [p.48–53]. |
| KKR 2026 Outlook | AI-driven productivity is already visible, with S&P 500 real revenue per worker up 5.5% since ChatGPT after two decades of stagnation and U.S. GDP per employee projected to reaccelerate toward ~2% in 2026–27, supporting the case that capex yields efficiency gains [p.4, p.35–36, p.82]. Implied 10‑year S&P 500 EPS growth of ~16% versus an 11% median signals that much of the expected productivity/AI upside is already priced, and a bear case explicitly hinges on compute demand and AI monetization failing to keep pace with supply, exposing cracks in the AI capex cycle [p.5–6, p.14, p.59]. Speculative data center projects with weak unit economics and uncompetitive cost structures are flagged as likely value destroyers, underscoring ROIC concerns [p.26]. |
| Stifel Outlook | Extraordinary economic profits for AI‑linked Big Tech (e.g., Tech Hardware ~51% above S&P ROIC‑minus‑WACC and ~4.9x EV/Invested Capital) suggest current capex has so far been supported by high returns, but these sectors now trade at peak multiples with “no margin for error” into 2026 [p.6, p.17]. Asset‑price ratios (S&P 500 vs gold, oil vs gold) and productivity data (10‑year productivity growth back to Great Depression levels) show no evidence yet of a broad AI‑led productivity‑driven disinflationary boom that would structurally validate today’s investment and valuation extremes [p.32]. |
| RIC 2026 BAML | High AI capex is currently supported by very strong ROIC–WACC spreads in tech (19% vs 4% in oil majors) and concrete monetization/productivity examples, such as ServiceNow autonomously resolving ~80% of tickets and saving 400,000 labor hours a year, RBC cutting prep time from 45 to 5 minutes, Amazon’s Rufus engaging >250mn customers, and AI‑assisted research boosting discovery rates by 45% [p.7]. Productivity scenarios range from “marginal AI” (reverting to ~1.5% productivity after 2027) to “transformational AI,” which assumes an additional 1ppt annual productivity uplift through at least 2027, but authors stress that clear, broad-based gains are still limited and adoption is slowing, making future ROI sensitive to uptake [p.4, p.27]. |
| TRowe Outlook | The narrative is shifting from “what’s possible?” to “what’s profitable?” as speculative activity and bubble concerns increase pressure to “carve out clear paths to monetization” for large AI capex programs [p.6]. AI is expected to be “the biggest productivity driver for the global economy since electricity,” transforming sectors like healthcare, finance, manufacturing, and education [p.6]. Hyperscalers are in an “existential investment” phase, compelled to spend heavily to safeguard long‑term value amid rising competition, implying returns are partly strategic rather than fully proven near‑term cash flows [p.7–8]. |
| UBS Year Ahead | Monetization currently lags capex, with hyperscaler capex rising from ~40% to nearly 70% of operating cash flow between 2023 and 2025, but AI vendors’ longer-term annual revenues from end-users are estimated at about USD 1.5tr based on labor share, automatable tasks, and value capture assumptions [p.15]. Broader AI revenues across enabling, intelligence, and application layers are projected at USD 3.1tr by 2030 with ~30% six-year CAGR, while early adoption already lifts US GDP via information-processing investment (0.2ppt to 0.8ppt contribution 2019–2025) and generates ~1 hour per day of time savings per worker in studies, implying significant productivity and ROI potential if adoption scales [p.14–17]. |
| Source | Content |
|---|---|
| Brookfield Outlook | Capital needs for AI‑related power, data and manufacturing “far exceed what corporates and sovereigns can fund alone,” with sovereigns at record debt levels and large tech firms seeking well‑capitalized partners, pushing the system toward large‑scale partnerships, JVs, privatizations and private‑capital‑heavy structures [p.6, p.9]. Infrastructure private credit and debt financing of digital and energy assets are positioned as core funding channels offering “durable, inflation‑resilient yields,” but require disciplined underwriting to avoid credit‑quality erosion [p.28, p.30]. |
| Goldman Outlook | Hyperscalers have shifted from primarily cash-funded AI capex to heavier use of debt as capex plus buybacks/dividends have risen to ~95% of operating cash flow vs ~80% in 2019, prompting about $90bn of IG credit issuance YTD 2025 out of $1.5tn total [p.27]. Strong cash flow and low leverage mean no immediate systemic risk, but rising debt reliance, potential pressure on credit metrics, and uncertain ROI warrant close monitoring of AI-related credit through 2026 [p.27]. |
| Blackrock Outlook | AI builders are using debt to bridge a front‑loaded “financing hump” in which investment precedes revenues, leveraging up from currently low average debt‑to‑equity of about 0.54x among major cloud spenders [p.4][p.7]. Public debt is already at postwar highs while non‑financial corporate leverage has eased, so corporate balance sheets have room to lever up, implying greater corporate borrowing, a structurally higher cost of capital, and increased vulnerability to shocks such as bond‑yield spikes tied to fiscal concerns [p.4][p.7]. |
| Barclays Outlook | Hyperscaler AI capex has so far been financed mainly from internal cash flows, limiting immediate balance‑sheet stress even as spending more than doubles, but rising amounts and short hardware life mean investors will demand clearer, quicker returns [p.41–43]. Long‑duration, capital‑intensive infrastructure and energy‑transition assets needed to support AI (grids, storage, digital networks) face a large global investment shortfall—$65tn of needs vs an $11tn gap to 2040—highlighting ongoing financing challenges [p.29, p.36–39]. |
| Goldman Outlook - Summary | Since 2022, hyperscalers have largely funded AI capex from cash flows, but capex plus buybacks and dividends have recently absorbed ~95% of operating cash flow vs ~80% in 2019, prompting greater use of debt financing, including about ~$90bn of credit issuance by the five hyperscalers in 2025 through October versus ~$1.5tn of total IG issuance [p.17]. Despite currently strong cash flow and low leverage making AI-related debt “not an immediate concern” for credit investors, rising debt reliance through 2026 “warrants close monitoring” as it could pressure credit metrics and widen spreads, especially if ROI disappoints [p.16–17]. |
| JPM Outlook | AI capex is largely funded by profitable, cash‑rich tech and cloud firms with tech sector free cash flow margins near 20%, more than double late‑1990s levels, implying strong capacity to self‑fund investment rather than relying on fragile leverage [p.7]. Rising AI costs, however, increase the likelihood of corporate belt‑tightening—including hiring freezes or layoffs—to fund AI strategies, indicating potential internal strain even if balance‑sheet credit risk is limited [p.4]. |
| JPAM Outlook | Financing for AI‑related infrastructure has shifted heavily toward private markets, with private equity, infrastructure funds and private credit (grown from USD250bn in 2007 to USD2.5tn) funding data centers, networks and the power grids that support them [p.3, p.6]. The capex cycle is described as poised to deliver higher returns “not through leverage or risk escalation, but through genuine demand growth and capital scarcity,” implying manageable balance‑sheet strain [p.26–27]. |
| HSBC Outlook | n.a. |
| MS - 2026 US Equities Outlook - The Rolling Recovery Is Here | AI‑linked capex occurs in a market with strong cash generation and high free‑cash‑flow yields versus 1999, and is further supported by tax incentives (e.g., bonus depreciation, domestic manufacturing credits), suggesting spending is largely funded from robust internal cash flow rather than excessive leverage [p.15–20, p.23–27, p.33–34]. Capital discipline is emphasized: because AI and broader capex sit on private balance sheets, investment will be curtailed if ROI disappoints, implying a self‑correcting mechanism that limits prolonged balance‑sheet strain even if spending initially ramps [p.12–13]. |
| KKR 2026 Outlook | Hyperscalers are committing 60–70% of operating cash flow to capex, with roughly 6.7% of that capex funded by net new debt, prompting comparisons with past debt‑heavy cycles (telecom, shale) and a warning that leverage at these firms will face greater scrutiny by 2026 [p.4–5]. Such high reinvestment ratios constrain free cash flow and raise the risk that, if monetization lags, balance sheets and credit quality could be pressured despite current debt reliance being below prior overbuild episodes [p.4–5, p.26]. Speculative data center developers with weak ROIC and heavy capex burdens are singled out as likely “pans,” highlighting financing and solvency risk at the periphery of the AI buildout [p.26]. |
| Stifel Outlook | New AI‑related capital intensity is eroding hyperscalers’ free cash flow, forcing a shift from internally generated cash to external term‑debt financing, which in turn is widening hyperscaler credit spreads (from ~40 bps toward ~85 bps in the chart) and compressing forward P/E multiples (from ~29–30x toward ~18–20x) [p.8, p.13]. This balance‑sheet strain is framed as a key late‑cycle constraint on sustaining the AI capex boom and on maintaining Big Tech’s historically “asset‑light” high‑ROE model [p.6, p.13]. |
| RIC 2026 BAML | AI capex is being heavily debt‑funded, with AI‑linked big tech issuing $121bn in IG bonds in 2025 (vs a $28bn 5‑year average) and a further $100bn expected in 2026, including Oracle’s $38bn loan and 66% of its revenue backlog tied to OpenAI [p.8]. Despite this surge, S&P 500 ex‑financials net debt/equity is only 0.6x versus 1.25–1.4x in prior cycles, and this is characterized as normal re‑leveraging rather than a bubble so long as AI revenues remain strong, while higher leverage will eventually raise cost of capital and compress return spreads [p.7–8]. |
| TRowe Outlook | The magnitude and acceleration of AI capex now exceed the ability of even leading firms to fund it solely from operating cash flow, pushing them toward public bond markets and, increasingly, private credit to sustain very high capital intensity [p.7; p.12–13]. Debt funding introduces fixed interest payments, covenants, and more risk‑averse creditors who demand clear monetization paths, raising the risk that if AI revenue growth underwhelms—or rates rise—highly leveraged firms could struggle to service debt, potentially creating systemic risks for lenders and markets [p.7]. AI‑related infrastructure such as data centers and utilities is explicitly cited as a driver of private credit opportunities within an estimated USD 1.2 trillion broader private‑capital financing gap, implying growing balance‑sheet dependence on non‑bank credit channels [p.12–13]. |
| UBS Year Ahead | Capex as a share of operating cash flow for major hyperscalers has climbed sharply from around 40% in 2023 to close to 70% in 2025, implying higher free cash flow strain and dependence on ongoing investor support [p.15]. Risks flagged include the potential for circular dealmaking and cross-investments to create financial vulnerabilities, and a scenario where refinancing dries up or AI monetization disappoints could force capex cuts and raise default/financial stability risks [p.16, p.47–48]. |
| Source | Content |
|---|---|
| Brookfield Outlook | Electricity is characterized as “the bottleneck to growth,” with AI workloads using up to 10x the power of conventional compute per rack and another 5–10x increase expected as rack density rises, while data centers are projected to drive 17% of global electricity demand growth from 2025–2050 [p.6, p.11–12]. More than 70% of global transmission lines are over 25 years old, interconnection queues for renewables stretch close to a decade, and annual grid investment must exceed $600B by 2030 to avoid constraining AI and electrification, requiring an “any‑and‑all” mix of renewables, storage, nuclear and gas+CCS scaled rapidly [p.7, p.11–15]. |
| Goldman Outlook | Data-center power demand tied to AI is projected to grow 175%+ by 2030, lifting US data centers’ share of national power use from 3% to 8% and potentially pushing EU data-center demand toward levels that could equal a large fraction of current EU28 consumption, straining already aging grids (average US grid asset age ~40 years) [p.9, p.37, p.45]. Meeting this “generational” growth in power demand requires massive energy-transition and grid investment (~$12tn by 2030), with inadequate grid capacity and slow upgrades emerging as hard constraints on AI infrastructure rollout [p.9, p.37, p.46–48]. |
| Blackrock Outlook | AI and traditional data centers could consume 15–20% of current U.S. electricity demand by 2030—and potentially up to a quarter—testing the limits of power systems, grids, fossil‑fuel and materials industries amid interconnection backlogs and slow permitting in the West [p.9]. Land and energy, rather than chips, are identified as the real constraints, with the risk that AI capex plans are scaled back if power and grid bottlenecks bite, while China’s rapid buildout of generation and transmission plus cheap solar and batteries gives it an advantage in deploying AI infrastructure [p.9]. |
| Barclays Outlook | AI data centres are projected to consume nearly 4% of global electricity by 2030 and already put “much pressure on electricity prices,” with power both a CPI component (2.6% weight) and a key input cost, creating inflation and margin risks if supply and grid capacity lag demand [p.6, p.37]. The build‑out of renewables (now >30% of global electricity, with solar additions in 2024 exceeding coal, gas and nuclear combined) and grid‑scale storage (BESS capacity +113% to 126 GW in 2024) is framed as necessary infrastructure to accommodate AI‑driven electrification, implying that inadequate grid and generation expansion could increasingly constrain AI capex and deployment [p.36–38]. |
| Goldman Outlook - Summary | Elevated capex needs for grid modernization, renewables, and data center power demand—partly driven by AI data centers—are already outpacing operating cash flows for most utility issuers, contributing to a cautious sector view and highlighting power and grid infrastructure as a financial bottleneck [p.16]. Utilities’ strained cash-flow coverage is flagged as a key constraint linked to AI-related power demand [p.16]. |
| JPM Outlook | Power availability is flagged as a key bottleneck, with a possible “supply crunch on power” cited as a trigger that could puncture AI momentum and help cause a recession or bear market [p.4]. Power constraints are also listed alongside slower adoption and hardware obsolescence as risks for AI‑linked equities, and the “insatiable need for power by data centers” is highlighted as both a constraint and an investment theme for infrastructure and utilities [p.7, p.10]. |
| JPAM Outlook | Power and grid capacity are central constraints: electricity demand in OECD markets is expected to accelerate to 2%–4% per year (U.S. 1.7%–3.2%) into the 2040s, while new generation and transmission face long permitting timelines, regulatory complexity and supply‑chain bottlenecks that slow the system’s ability to expand [p.25, p.27]. These constraints create a period of capital scarcity in which existing generation, transmission and distribution assets become more valuable and limit how fast AI‑driven data‑center buildouts can progress [p.26–27]. |
| HSBC Outlook | Electricity is identified as a key bottleneck, with “electricity‑hungry” data centres, EVs, and growing cooling demand straining capacity and forcing utilities to invest heavily in grid upgrades and new generation [p.3, p.12]. AI’s “seemingly insatiable energy needs” increase the requirement for renewables, nuclear, and storage, while co‑locating power sources with data centres and doubling global renewable capacity by 2030 (plus 23% CAGR in storage to 2035) are framed as necessary to avoid power‑system constraints on AI capex and as major investment opportunities for utilities and power infrastructure developers [p.13–14]. |
| MS - 2026 US Equities Outlook - The Rolling Recovery Is Here | Intensifying demand for compute and therefore power in the AI capex cycle makes power availability and “time to power” critical constraints, fueling a “race to acquire power” that elevates the importance of rapid‑to‑deploy solutions such as natural‑gas turbines, Bloom Energy fuel cells, nuclear, and crypto‑to‑data‑center conversions [p.15]. Generation capacity and speed‑to‑energize are identified as binding bottlenecks and key to data‑center deployment pace [p.15]. |
| KKR 2026 Outlook | Chronic shortages in transformers, switchgear, electrical equipment, and related labor—driven by years of underinvestment, the energy transition, and weather—intersect directly with power‑hungry AI/data center demand and threaten to constrain or delay projects [p.19]. AI training clusters can require 100–200 chillers per cluster, vastly exceeding traditional commercial cooling needs and implying heavy electrical load and infrastructure demands, while rising U.S. power demand (including from data centers) is a key driver of projected natural gas demand growth to 138 Bcf/d by 2032 [p.23, p.70–71, p.122]. These factors support a view that the power system is a binding constraint and key cost driver for AI capex. |
| Stifel Outlook | n.a. |
| RIC 2026 BAML | Power and grid constraints are a central supply risk: expected U.S. power additions have quadrupled from 182 TWh in 2021 to 803 TWh, while the four largest high‑voltage equipment suppliers face ~3‑year backlogs, creating long lead times for data center connections [p.7]. Efficiency gains can paradoxically increase total power use (Jevons paradox), and a slower infrastructure buildout would lengthen payback periods on AI capex and depress AI equity multiples [p.7]. |
| TRowe Outlook | n.a. |
| UBS Year Ahead | Rapid AI data center buildout is driving US electricity demand higher, with data centers’ share of US power use rising from 0.7% (2015) to 3.2% (2024) and projected 8.6% by 2035, adding power demand comparable to Sweden’s annual consumption by 2030 [p.18]. US wholesale power prices are expected to be 23% higher on average in 2025 vs. 2024, and around USD 500bn of global grid investment is projected in 2026, framing power and grid capacity as real but investable constraints on AI infrastructure rather than immediate hard caps through 2030 [p.18–19]. |
| Source | Content |
|---|---|
| Brookfield Outlook | Land and siting emerge as constraints, with industrial/logistics land being repurposed for data centers where power access is available and valuations rising (e.g., a logistics parcel selling at 1.5x logistics land value for data center expansion) [p.24–25]. Permitting and development speed implicitly constrain buildout via decade‑long interconnection queues and the need to “de‑bottleneck” grid and power infrastructure, while industrial adoption of AI is limited by execution and operational‑expertise requirements rather than plug‑and‑play deployment [p.7, p.19]. |
| Goldman Outlook | AI and clean-energy hardware depend on highly concentrated supply chains, with ~60% of rare earths produced in China and ~90% of leading-edge AI/5G semiconductors manufactured in Taiwan, creating strategic vulnerability and driving economic-security-focused capex [p.20, p.45]. Labor and execution capacity are also binding: US and European power industries need 750,000+ additional workers by 2030 while renewables are >2.5x more labor-intensive than fossil fuels, and long permitting and construction lead times for gas and especially nuclear further constrain how fast AI-related power and infra can be deployed [p.48]. |
| Blackrock Outlook | Land availability, permitting backlogs, and interconnection queues are highlighted as key non‑chip constraints on data center expansion, alongside the need for critical minerals to support the associated power and infrastructure buildout [p.9]. Opportunities are seen “where constraints will likely bite the most”: power systems, grids, critical minerals and firms positioned to benefit from permitting reform, underscoring how these bottlenecks shape AI capex pacing and location [p.9]. |
| Barclays Outlook | Critical minerals and rare earths are a key structural bottleneck, with China controlling around 70% of global rare earth production and much refining capacity, allowing it to weaponise exports and raising cost and security risks for AI‑related technologies and semiconductors [p.8, p.22–23]. In response, the US is attempting to build a vertically integrated domestic mine‑to‑magnet and chip ecosystem via government‑backed projects in rare earths, lithium, copper, cobalt, germanium and advanced fabrication (MP Materials, Lithium Americas, Trilogy Metals, Intel), but this will take years and likely implies higher cost structures [p.22–24]. |
| Goldman Outlook - Summary | n.a. |
| JPM Outlook | A “supply crunch on critical materials” is mentioned alongside power as a risk that could slow AI capex and trigger broader macro weakness [p.4]. |
| JPAM Outlook | n.a. |
| HSBC Outlook | Labour is highlighted as a second major bottleneck, as US net immigration has “plummeted” and new activities require re‑skilling, potentially constraining execution of AI and re‑industrialisation projects [p.3, p.5]. Critical minerals and rare earths—essential for renewables, storage and thus AI‑enabling power systems—face concentrated supply and export restrictions, underscoring diversification needs and posing a structural constraint, while specific issues like land and permitting for data centres are not discussed explicitly [p.13–14]. |
| MS - 2026 US Equities Outlook - The Rolling Recovery Is Here | n.a. |
| KKR 2026 Outlook | Supply‑demand imbalances in industrial equipment and construction‑related labor, including persistent shortages of electrical components, transformers, and contractors, create real‑world bottlenecks for data center and infrastructure build‑outs tied to AI and the energy transition [p.19]. Cooling infrastructure needs (100–200 chillers per AI cluster plus liquid cooling directly on servers) also imply significant demands on specialized equipment, water, and site suitability, effectively tightening constraints on viable locations even before considering local permitting or land issues [p.23]. Broader industrial and infrastructure constraints are embedded in their “Security of Everything” and real‑assets themes as structural capacity limits that can slow AI‑related construction [p.18–20, p.42]. |
| Stifel Outlook | n.a. |
| RIC 2026 BAML | Local political and permitting frictions around land, water, and cost are already constraining data center deployment, with efforts to cancel or slow projects in Mesa (AZ), Colorado Springs (CO), and Prince William County (VA) as grid prices and resource concerns rise [p.7]. |
| TRowe Outlook | n.a. |
| UBS Year Ahead | Tightening supply of critical materials—especially copper—is highlighted as a constraint, with copper demand expected to rise close to 3% and prices to exceed USD 13,000/mt in 2026 as the deficit widens to 87,000mt (from 53,000mt in 2025), driven in part by AI-related electrification and grid buildout [p.19]. Broader raw-materials tightening and the interaction with the energy transition are emphasized as key resource-side pressures [p.19]. |
| Source | Content |
|---|---|
| Brookfield Outlook | AI, together with digitalization, deglobalization and decarbonization, is framed as a structural—not cyclical—driver of an infrastructure “supercycle,” with AI‑related digital and energy infrastructure financing described as a “defining growth driver” over the next decade [p.2, p.5, p.30]. Electricity availability is depicted as a gating factor for national competitiveness and GDP growth, implying that macro outcomes and future growth are increasingly contingent on sustaining large‑scale AI‑linked power and infrastructure capex [p.11–12]. |
| Goldman Outlook | AI capex has become a major pillar of business investment that offsets slowing consumer spending, stagnant housing, and a weakening labor market, potentially masking underlying fragility in the “real economy” [p.28]. A marked reversal in AI-related investments or a sharper deterioration in the labor market is flagged as a plausible trigger for a hard landing, making growth increasingly dependent on continued AI capex momentum into 2026 [p.28]. |
| Blackrock Outlook | AI capex is so large and concentrated in a handful of big tech firms that “micro is macro,” with AI investment contributing to U.S. growth in 2026 at roughly three times its historical average and helping growth hold up even as the labor market cools [p.3][p.5]. A slowdown or rollback of AI capex—whether from monetisation shortfalls or physical constraints—would therefore pose significant cycle risk to both macro growth and market performance, given the outsized role of AI in driving recent returns [p.3][p.5][p.9]. |
| Barclays Outlook | AI is a “strong tailwind” without which the US economy would have stagnated or shrunk in 2025, and AI data centres now drive a significant portion of US GDP growth, making the cycle increasingly dependent on continued capex momentum [p.5, p.42]. AI‑related mega‑caps have contributed a disproportionate share of global equity returns, turning AI into a risk‑on, high‑volatility complex where any stall in capex or monetisation could trigger severe snap‑backs in markets and a reversal of the wealth effect that has been supporting demand [p.13–15, p.32–33, p.42–43]. |
| Goldman Outlook - Summary | AI capex is a core pillar of 2025–2026 growth, counterbalancing slowing consumer spending, a stagnant housing market, and emerging labor-market weakness, creating an “uneasy equilibrium” where AI-driven investment may mask fragility in the underlying real economy [p.11, p.18]. A “marked reversal and broad unwind of AI-related investments” is cited as a potential trigger for a hard landing, underscoring how dependent the macro outlook and markets have become on continued AI capex [p.18]. |
| JPM Outlook | Given the scale of AI investment, a slowdown—whether from earnings misses, power/materials crunch, or liquidity shock—could cause a recession and/or bear market or at least pressure AI‑linked wealth gains that have been supporting consumption [p.4]. Tech and AI‑linked firms now account for a large share of S&P 500 earnings and capital spending growth, leaving markets increasingly dependent on continued AI capex and associated earnings expansion and vulnerable to any AI‑related “missteps” [p.7]. |
| JPAM Outlook | AI‑driven data‑center expansion is one of several pillars (alongside industrial onshoring and the energy transition) behind structurally higher power demand and the multi‑decade infrastructure capex cycle, so a slowdown in AI buildouts is highlighted as a key long‑term demand risk that could undermine growth expectations for utilities and infrastructure [p.25, p.27]. Demand is portrayed as diversified across multiple drivers, tempering but not eliminating macro dependence on AI capex [p.25–27]. |
| HSBC Outlook | AI‑related investment is presented as a key driver of US and global growth, central to the strong capex cycle and a major reason economists have underestimated US growth and earnings [p.3–4, p.18–19]. Risks to the cycle include ROI disappointments, delays in data‑centre and electricity buildout, and policy/monetary shocks, which could cause volatility or corrections but are not seen as sufficient to end the bull market while AI‑linked earnings “keep rolling in” [p.3–5, p.19]. |
| MS - 2026 US Equities Outlook - The Rolling Recovery Is Here | AI‑driven efficiency and capex are embedded as central drivers of the 2026–27 S&P 500 earnings path and market broadening, with stronger AI margin uplift underpinning both the base case and bull‑case scenarios [p.1–2, p.14–15, p.22]. Macro risk is material: a renewed inflation spike forcing rate hikes in late 2026 could trigger multiple compression, especially in high‑beta, long‑duration, AI‑exposed areas, while private‑sector capital discipline means AI capex could be cut quickly if ROI falters, abruptly weakening a key growth pillar [p.12–13, p.22–23]. |
| KKR 2026 Outlook | AI-related investment in software and information processing equipment has recently contributed as much to U.S. GDP growth as consumer spending, making AI capex a central macro driver [p.9]. Market valuations similarly embed high dependence, with implied long‑term S&P 500 EPS growth at ~16% and a bear case that features an AI capex bust if compute demand and monetization underperform, potentially triggering a broader market correction [p.5–6, p.55–59]. A potential boom‑and‑bust in AI/data centers—analogous to telecom or shale—is identified as one of the key downside risks to both growth and asset prices [p.5, p.26, p.59]. |
| Stifel Outlook | U.S. growth in 2025 is unusually reliant on hyperscaler AI capex, with rising fixed investment (19% of GDP) offsetting slowing personal consumption (68% of GDP), but with capex growth “likely peaked,” the AI boom is seen as insufficient to prevent a fragile 2026 macro outlook if the consumer weakens [p.1, p.10, p.25]. Cyclical Growth and Big Tech equity performance is similarly heavily tied to this late‑cycle AI capex pulse, leaving markets exposed when the rate of capex growth normalizes or reverses [p.14, p.16–p.18, p.20, p.25]. |
| RIC 2026 BAML | AI equity performance and valuations are increasingly tied to expectations of sustained AI capex and adoption, with AI stocks (AIQ) outperforming the Nasdaq and S&P 500 even as measured AI adoption among large firms has slowed after peaking at 15%, implying that “big upside needs faster adoption” [p.1, p.4, p.10]. Bubble‑risk metrics (BRI at 0.25 vs 0.8 “pop” threshold) and easing global financial conditions (forecast central bank cuts) suggest that macro and policy backdrops still support AI capex, but slowing adoption or tighter financial conditions would raise cycle risk and could trigger capex retrenchment and valuation compression [p.1, p.4–6]. |
| TRowe Outlook | AI‑related capex has “significantly boosted U.S. growth in 2025,” with 2026 fiscal incentives expected to “strengthen that tailwind,” and is a key force behind a “two‑speed” economy where AI‑linked investment and fiscal expansion help avert recession while housing and consumption lag [p.3; p.14]. If many AI firms become highly leveraged and sector growth slows, systemic credit risks could emerge that impact lenders, investors, and the broader market, tying macro and market stability to sustained AI capex and revenue growth [p.7]. |
| UBS Year Ahead | AI-related capex has become a primary engine of equity performance—IT and communication services now make up 36% of MSCI AC World, and the top 9 US tech stocks contributed 72% of Russell 3000 gains over the past year—while AI investment in information-processing equipment and software has raised its contribution to US real GDP growth from 0.2ppt to 0.8ppt between 2019 and 2025 [p.14]. AI disappointment is identified as a key macro risk: a stall in AI investment due to poor monetization, technical setbacks, or funding strains could trigger capex cuts, slower adoption, and possible financial stress, underlining that current growth and market leadership are increasingly dependent on continued AI capex [p.16, p.47–48]. |
| Source | Content |
|---|---|
| Brookfield Outlook | Primary beneficiaries include data centers (“AI factories”), hyperscalers, power generators, grid and transmission operators, renewable and storage developers, nuclear and gas+CCS platforms, semiconductor and compute manufacturers, and fiber/digital‑infrastructure owners, all positioned to capture AI‑driven capex [p.5–7, p.11–15, p.24–25, p.28–30]. Industrial companies that modernize with AI and financial services/financial‑infrastructure firms leveraging AI for underwriting, loss forecasting and fraud detection are highlighted as additional winners from productivity gains, while infrastructure debt investors gain from financing AI‑related assets with inflation‑resilient yields [p.19–20, p.30]. |
| Goldman Outlook | The AI capex wave primarily benefits hyperscalers and is spreading to semiconductors, agentic-AI software, data management, cybersecurity, fintech, and digital infrastructure, with elevated valuations for AI-linked infra (median EV/EBITDA 11.7x vs 10.2x for broader infrastructure) [p.8, p.37]. Smaller and mid-cap “picks and shovels” providers on the front line of AI infrastructure and innovation, as well as firms positioned in energy transition, power/grid upgrades, and economic-security themes (including alternative chip and critical-mineral supply), are highlighted as key beneficiaries of both AI build-out and its constraints [p.8, p.19, p.20, p.37, p.45, p.48]. |
| Blackrock Outlook | Structural beneficiaries include “picks and shovels” tied to AI infrastructure—power systems, electricity grids, critical‑minerals producers, data networks, connectivity and permitting‑reform winners—where constraints are most binding [p.9][p.13]. Listed infrastructure, providing the power and networks AI relies on, trades at a deep discount to global equities despite multi‑decade demand from AI and the energy transition, and BlackRock emphasizes opportunities in infrastructure equity and private credit that finance the buildout [p.9][p.13][p.15]. |
| Barclays Outlook | Key beneficiaries include AI‑enabling semiconductors and data‑centre infrastructure (e.g., Intel and Nvidia’s partnership), critical‑minerals producers with US backing (MP Materials, Lithium Americas, Trilogy Metals), and energy and nuclear firms positioned to supply reliable power, including advanced nuclear fast‑tracked for defence and AI infrastructure [p.22–24, p.36–39]. Additional “picks and shovels” span grid and storage infrastructure and AI adopters in sectors like healthcare, defence/cybersecurity, and automation/robotics, which can monetise productivity gains and often benefit from long‑duration, inflation‑linked cash flows tied to electrification and data‑centre demand [p.17, p.36–39]. |
| Goldman Outlook - Summary | Key beneficiaries include the large US tech “hyperscalers” and Magnificent 7, which drive and benefit from AI capex, as well as small- and mid-cap “enablers,” the “picks and shovels” of the AI boom positioned on the front line of AI innovation [p.3–5]. Additional winners span global semiconductor and AI hardware ecosystems—especially EM leaders such as TSMC with massive planned AI-related investments—and policy-supported tech sectors in markets like Japan, while utilities are flagged more as pressured by AI-related power and grid capex than as clear beneficiaries [p.7–8, p.16]. |
| JPM Outlook | Key beneficiaries include AI “innovators” (large tech, Mag 7), “enablers” such as industrials and utilities that supply infrastructure and power, and “adopters” like financials and health care using AI to drive efficiency [p.2, p.7]. On the picks‑and‑shovels side, semiconductors, data centers, and power/grid infrastructure stand out, with private equity, infrastructure, and private credit heavily involved in data center and associated infrastructure build‑outs, and VC activity concentrated in AI semis, robots/autonomous machines, LLMs, and sector‑specific AI applications [p.7, p.10]. |
| JPAM Outlook | The clearest beneficiaries are energy utilities—especially vertically integrated utilities that own generation, transmission and distribution—because they can offer end‑to‑end power solutions to hyperscalers and industrials, capturing long‑term, inflation‑linked contracts in monopolistic or quasi‑monopolistic markets [p.27]. Private‑market “picks and shovels” across data centers, core infrastructure and power grids are also cited as key winners from the AI‑driven capex wave and associated capital scarcity [p.6, p.24–27]. |
| HSBC Outlook | Primary beneficiaries of the AI capex and constraint‑driven buildout include chipmakers, semiconductor equipment makers, data‑centre developers, power equipment and smart‑grid suppliers, utilities and power‑infrastructure developers, and industrials tied to re‑industrialisation [p.3, p.9–12, p.18–19]. Asia’s AI hardware and data‑centre ecosystem (China, South Korea, Japan, ASEAN, India) is positioned as a major regional winner, while software, cloud, intelligent automation, and select aerospace/defence players also benefit from deployment at scale [p.9–12]. |
| MS - 2026 US Equities Outlook - The Rolling Recovery Is Here | Key beneficiaries of the AI capex wave and related constraints include semiconductors (especially memory, where a “supercycle” is beginning), power and energy‑infrastructure providers tied to “powering GenAI” (natural‑gas turbines, Bloom Energy fuel cells, nuclear, crypto‑site conversions, select Utilities), and Industrials leveraged to the emerging US capex cycle and tax incentives [p.15, p.33–34, p.37–38, p.41–42]. An AI value‑creation heat‑map highlights industry groups with particularly attractive AI upside—Healthcare Equipment & Services, Transportation, Consumer Services, Software & Services, Capital Goods, Automobiles & Components, and Staples Distribution & Retail—as structural winners from AI‑enabled productivity and margin gains [p.53–55]. |
| KKR 2026 Outlook | Major beneficiaries include “picks and shovels” tied to AI infrastructure and constraints: data centers, HVAC and chiller manufacturers, electrification and industrial equipment suppliers (switchgear, transformers, electrical components), grid and energy infrastructure, and natural gas/LNG producers supporting rising power demand [p.18–19, p.23–25, p.68–71]. Real assets such as infrastructure, energy, and asset‑backed finance are favored as collateral‑backed ways to play AI‑driven power and equipment needs, while significant long‑term value is also expected to accrue to firms that successfully apply AI to their own operations rather than just enabling it [p.5, p.18–20, p.68–70]. |
| Stifel Outlook | Primary beneficiaries of the AI capex wave are Cyclical Growth sectors dominated by Big Tech and hyperscaler‑linked industries—Semiconductors & Equipment (e.g., NVDA), Technology Hardware & Equipment (e.g., AAPL), Media & Entertainment (META, GOOG), and Autos & Components (~88% TSLA)—which currently exhibit massive economic profits and the highest valuation multiples vs the S&P 500 [p.14, p.17]. Tactically, these AI‑exposed groups are seen as crowded and late‑cycle, with recommended hedges in Defensives such as Healthcare, Staples, gold, Waste, and certain Software as ways to manage the risk from a potential AI‑capex and valuation rollover in 2026 [p.1, p.14–p.16, p.20–p.21]. |
| RIC 2026 BAML | Direct “picks and shovels” beneficiaries are most evident in power and grid equipment, where high‑voltage electrical equipment providers enjoy ~3‑year order backlogs due to AI‑driven grid upgrades, and in hyperscaler and semiconductor ecosystems where rising capital intensity and sustained compute demand (LLM FLOPs growing 4.4x annually, faster TPUs) underpin earnings [p.7–8]. |
| TRowe Outlook | Primary beneficiaries include chipmakers and semiconductor enablers that supply training and inference hardware, hyperscalers running cloud platforms and data centers, and “physical AI” infrastructure segments spanning power infrastructure, cooling, networking, automation, and robots/drones [p.5; p.7–9]. Industrials, materials, and energy sectors are positioned to gain from OBBBA’s focus on energy grids, data centers, roads, bridges, and industrial capacity, while private credit and private equity benefit from growing capital needs for AI‑related infrastructure such as data centers and utilities [p.8; p.12–13]. Leadership is rotating from original AI mega‑caps toward these “picks and shovels” enablers, where earnings potential is viewed as underappreciated [p.8–9]. |
| UBS Year Ahead | Key beneficiaries span the AI stack: the enabling layer (GPUs, memory, interconnects, SSDs, networking, cloud, and AI data center infrastructure), the intelligence layer (model and ML platform providers), and the application layer (copilots, assistants, customer-service tools, R&D applications, humanoids), with a balanced allocation across all three advocated [p.17–18]. Power and resource constraints create additional “picks and shovels” winners in US utilities, grid and power infrastructure, and critical materials such as copper, which benefit from rising electricity demand, grid investment (~USD 500bn in 2026), and tightening raw-materials markets [p.18–19, p.26–27]. |
This site synthesizes publicly available 2026 outlook reports for informational purposes only. It is not investment advice. Views expressed are those of the original authors. No affiliation with or endorsement by the cited institutions is implied.