Global Investment Outlook 2023 — Narrow Leadership and the AI Pivot

2023 was a year in which the headline numbers and the lived experience diverged more than usual. Major equity indices posted strong returns. Most companies inside those indices did not. The gain was concentrated in a handful of very large technology businesses, which — because indices are weighted by market capitalisation — was sufficient to carry the aggregate.

The year's most instructive event was the failure of several regional banks in March. It is frequently filed alongside 2008 as a banking crisis. It was a fundamentally different phenomenon. The 2008 failures came from credit: banks held assets that turned out to be worth far less than claimed. The 2023 failures came from duration: banks held government bonds of impeccable credit quality that had fallen in market value because rates had risen. The credit was fine. The interest-rate mathematics was not.

Meanwhile the discount-rate regime that had dominated 2022 began to matter less than a thematic one. The emergence of large language models as a commercially credible technology created the first investment theme since the rate shock capable of pulling capital in regardless of the cost of money.

For private markets, 2023 was the second consecutive year of constrained exits. That persistence, more than any single event, defined the institutional experience.

The breadth problem

The gap between index return and typical company return was unusually wide in 2023, and understanding why matters more than the specific numbers.

Major indices weight constituents by market capitalisation. A company worth two trillion dollars moves the index roughly two hundred times as much as a company worth ten billion. When gains concentrate in the very largest constituents, the index can rise substantially while the majority of its members are flat or down.

This produces several practical consequences that are easy to miss:

• Active managers underperformed for structural reasons. A manager holding a diversified portfolio necessarily underweights the largest constituents relative to the index. In a year when those specific names drive the return, underperformance follows arithmetically, independent of skill.
• "The market" became a misleading shorthand. Statements about market performance in 2023 were largely statements about a handful of companies.
• Index exposure became a concentrated position. An investor holding a broad index fund held a portfolio far more concentrated in a single sector and theme than the word "diversified" implies.

When a capitalisation-weighted index rises on the strength of a few constituents, buying the index is not diversification. It is a concentrated bet wearing diversification's clothing.

The banking stress was about duration

The March 2023 failures deserve careful treatment because the popular framing obscures the mechanism.

A bank takes deposits, which are liabilities repayable on demand, and invests them in longer-dated assets. When rates rise sharply, the market value of those longer-dated assets falls. This is not a credit judgement — a government bond bought at a 1% yield is still worth less than one bought at 5%, regardless of the certainty of repayment.

Accounting rules permit banks to hold certain securities at amortised cost rather than market value, on the reasoning that they intend to hold to maturity and will be repaid in full. That reasoning is sound provided the bank is never forced to sell.

The failure sequence therefore ran: rates rise → market value of held bonds falls, largely unrecognised in reported capital → depositors withdraw, for reasons including higher yields available elsewhere → the bank must sell assets to fund withdrawals → the unrealised loss becomes realised → capital adequacy is called into question → withdrawals accelerate.

Two features made 2023 distinct from prior bank runs. Deposits could be moved digitally, instantly, at scale — the run happened in hours rather than days. And concentrated depositor bases meant a small number of correlated decisions could move a very large fraction of a bank's funding.

The relevant lesson is not that banks were reckless in the 2008 sense. It is that an asset can be simultaneously perfectly creditworthy and a source of insolvency risk, if its duration is mismatched to the liabilities funding it and the holder loses the option to wait.

AI as the first post-rate-shock theme

Between the 2022 rate shock and early 2023, essentially every allocation question was downstream of one variable: what happens to the cost of capital. Sector preferences, stage preferences, and valuation discipline were all expressions of the same rate view.

The commercial arrival of large language models changed that. For the first time since the shock, a theme emerged that attracted capital on its own merits rather than as a rates expression.

This mattered structurally, not just thematically:

• It created demand for capital-intensive physical infrastructure — data centres, power, semiconductors — at a moment when most venture demand was for capital-light software. That is a different financing profile requiring different investors.
• It reopened large-round formation at a time when late-stage rounds were otherwise scarce.
• It produced a genuine bifurcation. Companies with credible AI positioning could raise on terms unavailable to anyone else, while comparable businesses without that positioning faced the prevailing tight market.

That bifurcation created an obvious incentive to reposition, and a correspondingly hard underwriting problem: separating businesses genuinely built on the technology from those describing existing products in new vocabulary. That diligence question defined the year's venture activity more than any macro variable.

The exit drought and its consequences

The most important thing about 2023 for institutional investors was not any event. It was the continuation of something from 2022: exits did not recover.

Public listings remained sparse. Strategic acquirers were cautious, facing higher financing costs and their own valuation uncertainty. Sponsor-to-sponsor transactions were constrained by a persistent gap between what buyers would pay and what sellers would accept, itself a function of the smoothed marks discussed in the 2022 report.

A second consecutive year of low distributions compounded rather than repeated the problem:

• Institutions relying on distributions to fund new commitments could not recycle capital. Fundraising conditions tightened further, including for strong managers.
• Fund lives extended. Assets held longer than underwritten changed the return profile, since internal rate of return is time-sensitive even when the multiple is unaffected.
• Secondaries shifted from opportunistic to structural. In 2022, secondary sales were largely distressed rebalancing. By 2023 they had become a routine portfolio-management tool, with continuation vehicles emerging as a mainstream way for managers to hold quality assets longer while offering existing investors an exit.

This last development is a genuine structural change in how private capital operates, and it originated in a liquidity constraint rather than in a design decision.

Normalisation of the down round

A behavioural shift completed in 2023 that had begun the previous year.

In 2022, a down round carried significant stigma. Companies went to considerable lengths to avoid one — cutting costs deeply, raising structured rounds with punitive terms in exchange for a flat headline valuation, or extending runway rather than testing the market.

By late 2023 the calculus had changed, for a simple reason: when a large share of the cohort has repriced, repricing is no longer a signal about the individual company. It is a signal about the vintage.

The practical effect was healthier than it sounds. Clean down rounds are generally better for founders and employees than flat rounds carrying heavy structure. A liquidation preference stack accumulated to avoid a lower headline number can eliminate common equity value at exit entirely, while the headline number preserved nothing but appearances.

The normalisation also cleared a backlog. Companies that had avoided the market for eighteen months could finally transact, because the price they would receive was no longer read as an indictment.

The same mechanism in three places

A striking feature of 2023 is that one mechanism — the effect of a higher discount rate on long-dated cash flows — produced three superficially unrelated events. Recognising them as one thing is the year's most useful analytical exercise.

In banking, institutions holding long-dated fixed-rate securities saw those holdings fall in value as rates rose. The securities were high quality and the credit risk was minimal. The problem was duration: a long-dated bond bought at a low yield is worth less when yields rise, and an institution that must sell before maturity realises that loss. Nothing about the assets was risky in the sense banks are usually assessed on.

In private markets, the same rate move had repriced long-duration equity claims — companies whose value depends on cash flows years out. The 2022 report describes the mechanism; 2023 is where the consequences propagated to exits, which had themselves been the mechanism for realising those distant cash flows.

In public markets, the concentration in a small number of very large companies reflected, in part, the same variable. Companies with near-term cash flow and strong balance sheets were less affected by a higher discount rate than companies whose value sat further out, which meant the market's leadership narrowed toward exactly the businesses least exposed to the mechanism.

Why this matters: an investor who understood the 2022 duration mechanism had, in principle, a framework that explained all three of 2023's headline events. Almost nobody applied it across all three, because the events arrived in different sectors with different vocabularies — a banking crisis, an exit drought, a market concentration story.

The same arithmetic appeared as a bank failure, a fundraising problem and an index concentration. They were reported as three stories because they happened to three different constituencies. They were one variable.

The general lesson is about the cost of specialisation. Analytical frameworks are organised by asset class, and a mechanism that operates across asset classes is therefore systematically under-recognised — each specialist sees their own instance and treats it as sector-specific. Duration is the clearest example in this archive, and the 2024 report's account of constraint migration is the same problem in another form.

What an allocator could act on

Ask what an index actually holds. Index concentration reached a point in 2023 where a market-cap-weighted allocation was substantially a position in a handful of companies. That is checkable from published index weights and requires no view. An investor holding an index for diversification should verify periodically that it is still providing any.

Check duration on the balance sheet, not only in the portfolio. The banking stress demonstrated that duration risk sits wherever long-dated assets are funded by short-dated liabilities — which is a description of a bank, and also of several other structures. The question is not what the asset is but whether the holder can hold it to maturity, and that depends on the liabilities.

Treat a second year without exits as a different condition from a first. One constrained year is cyclical. Two consecutive years begins to consume the resources — runway, LP patience, fund life — that would have made the constraint survivable. The 2023 US venture report describes the deferrals expiring; the general point is that duration of a constraint matters independently of its severity.

Read carrying values with the exit environment in mind. A private mark is an estimate of what an asset would realise. Two years without transactions means two years without evidence for that estimate. The absence of exits is not neutral for valuation — it removes the mechanism by which valuations are validated.

Recognise that a normalised down round is a functioning market. The removal of stigma from down rounds was genuinely healthy. A market where repricing carries a reputational cost is a market where repricing is deferred, and deferral compounds. 2023's normalisation cleared a backlog that had been accumulating since 2022 precisely because the price was no longer read as an indictment.

What 2023 established

• Index concentration became a recognised risk rather than an academic observation.
• Duration risk was re-learned as a balance-sheet problem, not only a portfolio one.
• Thematic allocation returned alongside macro allocation, after a period in which rates dominated everything.
• Secondaries and continuation vehicles became permanent infrastructure of private markets rather than distressed-market features.
• Private marks faced sustained scrutiny, as a second year without exits made the gap between carrying value and realisable value harder to defer.