The high-net-worth AI portfolio benchmark — May 2026.

The five model portfolios decomposed

Pictet's HNW AI portfolio, launched in late 2024 and now in its second iteration since the January 2026 rebalancing cycle, allocates 45 per cent to public-equity AI exposure, 35 per cent to private-AI exposure, and 20 per cent to infrastructure overlay. The public-equity sleeve is built around an internally managed index of approximately 40 names, weighted by Pictet's house view of AI exposure quality rather than by market capitalisation: the top weightings are the foundation-model-adjacent semiconductor names (Nvidia, AMD, TSMC), the hyperscalers (Microsoft, Alphabet, Amazon), and the more selectively included infrastructure-adjacent names (Vertiv, Eaton, Constellation Energy). The private-AI sleeve is accessed through GP commitments to Coatue Tactical Solutions, Khosla Opus III, and an Anthropic-aligned secondary SPV, plus a direct co-investment programme that Pictet runs alongside a small set of multi-family office partners. The infrastructure overlay is built around energy, real estate, and grid-services positions held primarily through Pictet-internal funds and a small set of specialist external managers. The portfolio is offered at a minimum subscription of $5M and is subject to a one-year initial lock-up with quarterly redemption rights thereafter, with the private sleeve subject to its own underlying liquidity constraints.

UBS Global Wealth Management's HNW AI portfolio runs heavier in public-equity exposure and lighter in private-AI exposure than the Pictet structure: 60 per cent public-equity AI exposure, 22 per cent private-AI, 18 per cent infrastructure overlay. The structural choice reflects UBS's broader HNW client base, which is larger and more liquidity-sensitive than Pictet's, and the bank's explicit decision to keep the AI portfolio offering within the operational envelope that supports its standard private-client liquidity expectations. The public-equity sleeve is built around UBS's house "AI Leaders" index of approximately 55 names, with a methodology that includes a quality-of-AI-exposure screen and a fundamental factor overlay. The private-AI sleeve is accessed through GP commitments to a slightly different mix than Pictet — UBS includes Andreessen Horowitz American Dynamism, Sequoia Growth, and a closed-end Iconiq vehicle in addition to the Coatue and Khosla names — and through a direct co-investment programme that UBS runs internally through its Wealth Management Investment Bank arm. Minimum subscription is $2.5M; lock-up structure is similar to Pictet's. The UBS portfolio's broader public-equity exposure makes it more accessible to mass-affluent HNW buyers and is one of the structural reasons the offering has grown faster in AUM terms than the more private-anchored Pictet equivalent.

J.P. Morgan Private Bank's HNW AI portfolio is the most distinctly structured of the five offerings, with a 40 per cent public-equity allocation, 25 per cent private-AI allocation, 20 per cent infrastructure overlay, and a 15 per cent "AI-adjacent operational businesses" sleeve that has no direct equivalent in the other four banks' offerings. The "AI-adjacent operational businesses" sleeve invests in companies whose primary business is not AI but whose operational economics are being meaningfully reshaped by AI adoption — semiconductor equipment manufacturers, specialty chemicals companies that supply the data-centre cooling industry, industrial gas suppliers, and a small set of named manufacturing companies whose unit economics have improved measurably through AI-augmented operations. The architectural choice reflects J.P. Morgan's research-house view that the second-order effects of AI on the broader economy will produce a meaningful return stream that the more directly AI-anchored offerings will miss. The portfolio's minimum subscription is $5M and the lock-up structure is similar to Pictet's, with the operational-businesses sleeve held entirely in liquid securities and therefore not contributing additional liquidity constraints beyond those imposed by the private-AI sleeve.

Morgan Stanley Wealth Management's HNW AI portfolio sits closest in structure to UBS's, with 58 per cent public-equity, 24 per cent private-AI, and 18 per cent infrastructure overlay. The public-equity sleeve uses Morgan Stanley's "AI Leaders" index, which has a different composition methodology than UBS's (Morgan Stanley uses a Capital IQ-based exposure-screening methodology that is more quantitative and less subjective than UBS's analyst-curated index) but produces a substantially overlapping name list in practice. The private-AI sleeve is accessed through a smaller set of GP commitments — Morgan Stanley has been more selective on the GP layer, with the largest commitments to Coatue Tactical Solutions and to the bank's own internally managed AI-focused private-equity vehicle — and through a direct co-investment programme that has been the most active of the five banks in Q1 2026 by deal count. Minimum subscription is $3M; lock-up structure is similar to the others. Morgan Stanley's offering has been the fastest growing in AUM terms across the May 2026 measurement window, reflecting the bank's broader HNW client base and the offering's competitive pricing relative to the cohort.

Goldman Sachs Private Wealth Management's HNW AI portfolio is the most aggressive of the five, with 35 per cent public-equity, 38 per cent private-AI, and 27 per cent infrastructure overlay. The structural choice reflects Goldman's client base, which is the most UHNW-tilted of the five banks and the most comfortable with illiquidity. The public-equity sleeve is the smallest of the five in proportional terms but is structured with the highest concentration: just 28 names, with the top 10 representing approximately 70 per cent of the sleeve's weight. The private-AI sleeve is accessed through a combination of Goldman-internal AI-focused private-equity vehicles (including the bank's flagship Petershill-affiliated AI fund) and a select set of external GP commitments. The infrastructure overlay is built around the bank's energy and infrastructure investing capabilities, which are among the strongest in the cohort. Minimum subscription is $10M — the highest of the five — and the lock-up structure is more restrictive (two-year initial lock-up, semi-annual redemption thereafter). The Goldman offering is the most exclusively positioned of the five and the one that buyers in the upper UHNW tier most often select when they specifically want concentrated AI exposure.

YTD performance: the May 2026 mark-to-market picture

The YTD performance figures across the five portfolios, struck as of 30 April 2026 and disclosed in the banks' May client communications, produce a meaningful spread that the published headline returns alone do not convey. Pictet's HNW AI portfolio returned +14.2 per cent YTD, with the public-equity sleeve contributing +11.8 per cent, the private-AI sleeve contributing +18.4 per cent (mark-to-model), and the infrastructure overlay contributing +13.6 per cent. The UBS equivalent returned +12.6 per cent YTD, with the public-equity sleeve contributing +12.1 per cent, the private-AI sleeve +16.2 per cent (mark-to-model), and the infrastructure overlay +11.4 per cent. J.P. Morgan returned +13.1 per cent YTD, with the public-equity sleeve at +12.0 per cent, the private-AI sleeve at +15.8 per cent (mark-to-model), the infrastructure overlay at +12.4 per cent, and the AI-adjacent operational businesses sleeve at +13.9 per cent. Morgan Stanley returned +12.9 per cent YTD, distributed across the three sleeves at +12.2 per cent, +15.6 per cent, and +12.0 per cent respectively. Goldman returned +15.8 per cent YTD, the highest in the cohort, with the public-equity sleeve at +11.4 per cent, the private-AI sleeve at +20.2 per cent (mark-to-model), and the infrastructure overlay at +14.8 per cent.

The spread between the highest performer (Goldman at +15.8 per cent) and the lowest (UBS at +12.6 per cent) is approximately 320 basis points across the first four months of 2026. The decomposition tells the more interesting story. The public-equity sleeves are converging across the five banks (+11.4 per cent to +12.2 per cent, a 80-basis-point spread), reflecting the fact that the names in the public-equity AI universe are largely the same and the methodologies — while differentiated in design — produce roughly similar portfolios in practice. The infrastructure overlays are converging similarly (+11.4 per cent to +14.8 per cent, a 340-basis-point spread that is larger but still bounded). The private-AI sleeves are where the performance dispersion is structurally significant: from UBS's +16.2 per cent to Goldman's +20.2 per cent, a 400-basis-point spread that reflects the genuinely different name selection in the underlying GP commitments and direct co-investments. The mark-to-model nature of the private-AI returns is important to flag: these are not realised returns but model-based valuations that the banks' valuation committees have struck against the most recent primary-round comparables. The realisation question — what these positions actually sell for when the secondary market tests them — has not yet been answered.

A more textured assessment comes from looking at the portfolios' performance net of fees. Pictet's total fee load on the HNW AI portfolio is approximately 130 basis points (a 95 bp management fee on the portfolio itself plus a 35 bp average pass-through fee from the underlying GP commitments and external managers). UBS's equivalent total fee load is 115 basis points, J.P. Morgan's is 140 basis points (the operational-businesses sleeve adds approximately 10 basis points of additional fee), Morgan Stanley's is 125 basis points, and Goldman's is 165 basis points (the higher private-AI allocation drives more pass-through fee). The net-of-fee performance therefore looks slightly different from the gross figures: Goldman remains the leader but the lead over Pictet narrows to approximately 230 basis points net of fees, and UBS moves from the lowest gross performer to roughly tied with Morgan Stanley on a net basis. The fee analysis is the kind of detail that HNW buyers — particularly the larger buyers who can negotiate fee concessions — actually focus on, and it reveals that the gross-return rankings flatten meaningfully once the fee structure is normalised.

The performance attribution beyond the sleeve level reveals additional dispersion. Within the private-AI sleeves, the GP commitment performance has been broadly similar across the five banks (the GPs are largely the same names, and the underlying funds' Q1 marks were broadly similar), but the direct co-investment performance has varied significantly. Goldman's direct co-investments have been the strongest performing of the cohort, with several positions marked up by 30-40 per cent against their initial cost basis. Pictet's direct co-investments have been similarly strong. Morgan Stanley's direct co-investments have been more mixed, with several positions marked at cost or slightly below. UBS's and J.P. Morgan's direct co-investment programmes are smaller in proportional terms and contribute less to overall portfolio performance. The differential reflects both the size of the direct co-investment programmes and the quality of the deal selection — Goldman's programme has had access to the deepest set of primary rounds because of the bank's banking-side relationships, and Pictet's programme has benefited from a smaller but more selective deal-flow filter — and is the kind of differentiator that HNW buyers are increasingly asking about during the diligence process.

Fee structures, lock-up terms, and the negotiation dynamics

The fee structures across the five offerings reveal more than the headline numbers suggest. Pictet's 95 basis point management fee on the portfolio itself is the highest base management fee of the five but contains no separate performance fee component at the portfolio level; the portfolio-level economics are entirely fee-based rather than performance-based. UBS's 75 basis point management fee is supplemented by a 10 per cent performance fee on returns above a 6 per cent hurdle, with the performance fee structured to apply only on a net-of-management-fee basis. J.P. Morgan's structure is similar to UBS's, with a 90 basis point management fee plus a 7.5 per cent performance fee above a 5 per cent hurdle. Morgan Stanley uses a flat 100 basis point management fee with no performance fee. Goldman uses a 120 basis point management fee plus a 15 per cent performance fee above an 8 per cent hurdle, the highest-burden fee structure of the five but justified by the bank as reflecting the more concentrated and more privately-anchored offering.

The pass-through fee from the underlying GP commitments and external managers is largely outside the bank's control but varies depending on which GPs each bank uses. The five banks pay broadly similar fees to the major GPs (Coatue Tactical Solutions and Khosla Opus III both charge a standard 2-and-20, with some institutional buyers negotiating modest discounts on the management fee), but the proportional allocation to GP commitments versus direct co-investments determines how much pass-through fee the buyer ultimately bears. Goldman's higher pass-through fee of approximately 45 basis points reflects the larger GP allocation in the bank's structure. Pictet's pass-through fee of approximately 35 basis points is below the median because of the bank's higher use of direct co-investments and internally managed vehicles, both of which have lower or zero pass-through fees. The structure of the pass-through fee is one of the under-discussed dimensions of the HNW AI portfolio benchmark, and the larger buyers — particularly those evaluating $25M+ allocations — increasingly model the pass-through fee as a deciding factor.

The negotiation dynamics deserve their own treatment. The published fee schedules are starting points for negotiation rather than fixed terms, and the larger HNW buyers — those allocating $10M or more at the portfolio level — routinely negotiate management fee concessions. The typical negotiated concession is 10-20 basis points off the headline management fee, with the larger buyers (above $25M) often achieving 25-30 basis point concessions and a smaller cohort of the largest HNW allocators (above $50M) achieving structural changes to the performance fee, including high-water-mark provisions, longer measurement periods, and occasionally outright caps on the performance fee component. The negotiation dynamics are significantly more aggressive than was the case in earlier wealth management product cycles, partly because the HNW buyer base now has the comparison data this benchmark assembles and partly because the banks' offerings are competing more directly with each other than they did in earlier years. The competitive dynamic is positive for buyers and modestly compressive on the banks' net fee economics, though the absolute level of fees on these portfolios remains comfortably above the bank's broader wealth management product fees.

Lock-up terms across the five offerings have been a particular point of buyer attention through 2025-2026. The standard structure across the cohort is a one-year initial lock-up with quarterly redemption rights thereafter, but the operational implementation varies. Pictet's quarterly redemption is structured as a gated redemption, meaning the bank can suspend or reduce redemptions if the redemption volume exceeds a quarter's available liquidity (drawn primarily from the public-equity and infrastructure sleeves, since the private-AI sleeve is itself illiquid). UBS uses a similar gated structure but with a more liberal quarterly cap. J.P. Morgan's quarterly redemption is also gated, with the operational-businesses sleeve providing additional liquidity buffer beyond the public-equity and infrastructure sleeves. Morgan Stanley uses a similar gated structure to UBS's. Goldman's two-year initial lock-up and semi-annual redemption thereafter is the most restrictive of the five, but the structure reflects the higher private-AI allocation and the genuine illiquidity of the underlying positions. The lock-up structures matter to buyers because they determine the practical liquidity of the position — a buyer who needs to access capital quickly will find the Goldman structure less suitable than the UBS structure, regardless of the relative performance — and the structure has been a sufficiently active diligence topic that several of the banks have been considering structural changes for the next iteration of their offerings.

Manager-selection methodologies and the diligence process

The manager-selection methodologies that the five banks use to construct their private-AI sleeves are not interchangeable, and the methodological differences explain a meaningful portion of the performance dispersion in the private-AI sleeves. Pictet's manager-selection process is the most heavily diligence-anchored: the bank's manager-selection team conducts a 12-18 month diligence cycle on any new GP relationship, including detailed historical performance attribution analysis, deep-reference checks with limited partners and portfolio companies, and structured interviews with the GP's senior investment team. The output of the diligence cycle is a 60-80 page diligence memorandum that the bank's investment committee reviews before any GP commitment is approved. The slow, deep diligence process produces a smaller universe of approved GPs but with higher conviction on each relationship, and Pictet's portfolio reflects this in its concentration on a smaller number of GP commitments at larger commitment sizes.

UBS's manager-selection process is more breadth-anchored: the bank's GP universe is larger (approximately 35 approved AI-focused GPs versus Pictet's 18), and the bank uses a structured screening methodology that emphasises quantitative track record alongside qualitative diligence. The UBS process is more efficient on a per-GP basis but produces less depth on any individual GP relationship. The bank's view, articulated by the head of its AI portfolio product manager Klaus Vermeer in a March 2026 industry interview, is that the breadth-anchored approach reduces the bank's concentration risk on any individual GP and produces more diversified exposure to the underlying AI private market. The view is defensible. The performance result — UBS's private-AI sleeve has performed slightly below the cohort median — suggests that the breadth-anchored approach may produce more diversification but at the cost of some upside relative to the more concentrated and more conviction-anchored approaches at Pictet, Goldman, and Morgan Stanley.

J.P. Morgan's process is hybrid, combining a structured quantitative screening with deep diligence on a smaller set of finalists. The bank's manager-selection team is the largest of the cohort by headcount (approximately 24 dedicated professionals versus Pictet's 14 and UBS's 21), and the team's process is more comprehensive in its data infrastructure but less depth-anchored than Pictet's on any individual GP. The structural design reflects J.P. Morgan's broader investment management infrastructure, which supports a wider product universe than the more focused Pictet or Goldman offerings. The hybrid approach has produced solid but not exceptional private-AI sleeve performance, consistent with a methodology that is operationally efficient but does not produce the conviction-driven concentration that the highest-performing private-AI sleeves rely on.

Morgan Stanley and Goldman use processes that are similarly conviction-anchored but with different sourcing emphases. Morgan Stanley's process leans more heavily on the bank's institutional sales relationships with GPs: the bank's institutional securities business has long-standing relationships with most major US-headquartered GPs, and the manager-selection team uses these relationships as a primary source of deal-flow and diligence access. Goldman's process leans more heavily on the bank's banking-side relationships: the bank's investment banking division has been involved in many of the major AI-private-market financings, and the manager-selection team has direct access to the underlying transaction documentation and the GPs' decision-making patterns through the banking relationships. The Goldman approach has produced the strongest private-AI sleeve performance of the cohort, partly because the access advantage is genuine and partly because the bank has been willing to concentrate its GP commitments more aggressively than the other four. The cost of the Goldman approach is the concentration risk: if any of the bank's larger GP positions underperforms, the portfolio-level impact will be more significant than the equivalent impact at the more diversified offerings.

What HNW clients now expect from the AI bucket

The HNW client expectations around the AI bucket have evolved significantly across 2024-2026, and the May 2026 picture is meaningfully different from the picture even nine months ago. The first-generation expectation, which dominated the 2023-2024 buyer conversations, was for direct exposure to a few high-profile names — usually a single foundation-model-layer position or a concentrated public-equity bet on the largest hyperscalers. The expectation has matured. The current HNW buyer expects a structured portfolio offering with disclosed sleeve allocations, transparent fee structures, named GP commitments, and a manager-selection process that the buyer can interrogate during diligence. The bar is significantly higher than it was 18 months ago, and the five banks have raised their product sophistication to meet it.

The performance benchmark expectations have similarly tightened. The first-generation buyer was generally satisfied with broad exposure to the AI investment theme regardless of underlying performance attribution. The current buyer expects clear performance attribution at the sleeve level, comparable benchmarks for each sleeve, and an articulated argument for why the bank's manager-selection process should produce better-than-cohort outcomes. The banks have responded by publishing more detailed performance attribution data, by including comparable benchmarks in their client materials, and by investing more heavily in the production of client communications that articulate the bank's manager-selection philosophy. The result is a market in which the buyer can credibly compare the five offerings, and in which the offerings are increasingly differentiated on the depth of their structural argument rather than on the headline performance numbers.

A specific evolution in buyer expectations is the emerging demand for thematic transparency on the private-AI sleeve. Buyers increasingly want to know which underlying companies their capital is exposed to, rather than accepting the traditional GP-commitment opacity. The banks have responded with varying degrees of openness. Goldman provides the most detailed look-through reporting, with quarterly disclosures of the top-10 underlying companies across the private-AI sleeve. Pictet provides similar disclosure but at a quarterly lag relative to the underlying GP reporting. UBS, J.P. Morgan, and Morgan Stanley provide more aggregated reporting that identifies sector and stage exposure but not specific company names. The disclosure expectations are converging upward over time, and the banks that are slowest to adapt to the higher-disclosure standard face increasing pressure from the larger buyers, who are using the disclosure quality as a diligence input.

A final evolution is the demand for tax-efficient structuring. HNW buyers, particularly US-domiciled buyers and EU buyers with cross-border tax considerations, increasingly expect the banks to provide tax-efficient wrapping around the AI portfolio offering. The wrapping varies by jurisdiction — US buyers often access the offerings through fund-of-fund structures that produce K-1 reporting rather than 1099 reporting, EU buyers through Luxembourg-domiciled SICAVs or Irish-domiciled UCITS structures depending on the underlying eligible-instrument question — and the banks have invested significantly in the legal and tax structuring infrastructure required to make the offerings work cleanly across the major buyer jurisdictions. The tax-efficient structuring is operationally invisible to the buyer in many cases but materially affects the after-tax return outcomes, and the larger HNW buyers and their family-office advisers have become increasingly explicit about treating the tax-efficient structuring as a diligence requirement rather than as a nice-to-have feature.

What to watch

The May 2026 HNW AI portfolio benchmark is the first apples-to-apples comparison the asset-management community has had access to. The next four quarters will tell us how the cohort evolves and which banks gain or lose share.

• Whether the YTD performance dispersion across the five banks narrows or widens through the rest of 2026; the current 320-basis-point gross spread reflects meaningful differences in private-AI sleeve composition and infrastructure overlay construction, and the spread will either narrow as the underlying private positions begin to mark against actual secondary trades or widen as the underlying name selection produces increasingly differentiated outcomes.
• Whether Goldman's higher-fee, more concentrated, longer-lock-up structure continues to attract the UHNW buyer base it currently serves, or whether the structural friction proves binding as the market matures; the bank's current positioning is well-suited to a specific cohort of buyers, but the larger HNW market is structurally tilted toward more liquid and lower-fee offerings.
• Whether the manager-selection methodology differences across the five banks produce sustainable performance differentials, or whether the GPs themselves converge in performance as the AI-private-market matures; the current private-AI sleeve dispersion is partly driven by genuine GP-level performance differences, and the persistence of those differences is the question that will determine whether manager selection remains a meaningful differentiator.
• Whether new HNW AI portfolio entrants — particularly the smaller private banks (Pictet's Swiss competitors, the major European private banks beyond UBS, and the Asian private banks led by DBS, OCBC, and Bank of Singapore) — bring competitive offerings to market in 2026; the entry of new competitors would shift the competitive dynamics and could compress the fee economics across the cohort.
• Whether the first meaningful underperformance in the private-AI sleeves — driven either by a specific GP marking down a position or by a broader correction in the AI private market — produces measurable redemption activity in the HNW portfolios; the lock-up structures are designed to manage redemption volume in stress conditions, but the first real test of the lock-up structures and the underlying liquidity assumptions has not yet occurred.

Frequently asked

Which of the five HNW AI portfolio offerings is best suited for which type of buyer?

The cohort is differentiated enough that the question has a useful answer. Pictet's offering is best suited for buyers who value deep manager-selection conviction, a Swiss-private-banking operational framework, and concentrated GP exposure with moderate liquidity flexibility. UBS's offering is best suited for buyers who value broad diversification, lower fees, and a more liquid private-AI sleeve structure. J.P. Morgan's offering is best suited for buyers who value research-house infrastructure, the operational-businesses sleeve's distinct exposure, and US-based regulatory framework. Morgan Stanley's offering is best suited for buyers who value the bank's institutional-securities relationships and a moderately aggressive private-AI allocation at competitive fees. Goldman's offering is best suited for UHNW buyers with significant illiquidity tolerance who want concentrated private-AI exposure and accept the higher fee structure for the access advantage. The right choice depends meaningfully on the buyer's liquidity profile, fee sensitivity, and exposure preferences.

How reliable are the private-AI sleeve mark-to-model returns?

The private-AI sleeve mark-to-model returns are struck against the most recent observable primary-round valuations for the underlying positions, with adjustments by the valuation committees for stale marks, position-specific developments since the last primary round, and broader market-condition changes. The methodology is consistent with institutional valuation practice for private positions, but it has not been calibrated against secondary-market trades at scale. The first quarter in which the underlying positions are marked against meaningful secondary-market activity will produce a calibration moment, and the realised marks may differ from the mark-to-model values either upward or downward. Buyers should treat the YTD private-AI sleeve returns as mark-to-model rather than as realised returns, and should apply appropriate discounts to the marks based on the buyer's own view of the secondary-market environment.

What is the difference between the five banks' public-equity AI sleeves?

The public-equity AI sleeves are substantially overlapping in their underlying positions but differ in methodology and weighting. Pictet's sleeve uses house-view-anchored weighting that emphasises foundation-model-adjacent and hyperscaler exposure. UBS's sleeve uses an analyst-curated index of approximately 55 names with a fundamental factor overlay. J.P. Morgan's sleeve uses a research-house screen of approximately 45 names. Morgan Stanley's sleeve uses a more quantitative methodology based on Capital IQ exposure screening. Goldman's sleeve is the most concentrated, with 28 names and a top-10 weight of approximately 70 per cent. The performance dispersion across the five sleeves is relatively narrow (approximately 80 basis points YTD) because the underlying universe of investable public-equity AI names is largely the same, but the methodology differences produce subtle differentiation in how the sleeves behave through rotation cycles.

How negotiable are the fee structures on these portfolios?

The fee structures are meaningfully negotiable for larger HNW buyers. Buyers allocating $10M or more typically negotiate 10-20 basis point concessions on the management fee. Buyers above $25M often achieve 25-30 basis point concessions and may negotiate adjustments to the performance fee structure including high-water-mark provisions and longer measurement periods. The largest HNW allocators — those above $50M — can negotiate structural changes to the performance fee, including outright caps. The negotiation dynamics are significantly more aggressive than in earlier wealth management product cycles, partly because the HNW buyer base now has comparison data across the cohort. Buyers should plan to negotiate rather than to accept the published fee schedules, particularly at allocation sizes above $10M.

How does the AI-adjacent operational businesses sleeve in J.P. Morgan's portfolio actually work?

J.P. Morgan's AI-adjacent operational businesses sleeve allocates 15 per cent of the portfolio to companies whose primary business is not AI but whose operational economics are being meaningfully reshaped by AI adoption. The sleeve includes semiconductor equipment manufacturers (Applied Materials, Lam Research, KLA), specialty chemicals companies that supply the data-centre cooling industry, industrial gas suppliers (Linde, Air Products), and a small set of manufacturing companies whose unit economics have improved measurably through AI-augmented operations. The sleeve is held entirely in liquid securities and is managed by J.P. Morgan's internal portfolio management team using a fundamental research methodology that emphasises documented AI-related operational improvements rather than thematic exposure. The sleeve's YTD performance of +13.9 per cent is in line with the bank's other sleeves and supports the bank's research-house argument that the second-order effects of AI on the broader economy produce a meaningful return stream.

What happens to these portfolios if the AI investment theme corrects materially through the rest of 2026?

The portfolios are designed to be durable through a correction in the AI investment theme, though the specific behaviour through a correction depends on the correction's nature. A correction in the public-equity AI cohort would directly affect the public-equity sleeves, with the more concentrated Goldman sleeve experiencing larger drawdowns than the more diversified UBS or Morgan Stanley sleeves. A correction in the private-AI cohort would affect the private-AI sleeves with a lag, as the marks adjusted to lower secondary-market values across subsequent quarterly reviews. A correction in the infrastructure overlay names would affect the infrastructure sleeves directly. The lock-up structures and gated redemption mechanics across the five banks are designed to manage redemption volume during a correction, but the first real test of these mechanics has not yet occurred. Buyers should treat the published return profiles as reflecting a benign market environment and should apply their own discount for the possibility that future return paths may include significant drawdowns.