Felix Smith

Using a future AI debate to interpret past performances, such as Berkshire's liquidity reserves in 2015, introduces an unacceptable risk of temporal bias. The market conditions specific to 2015 dictated the investment decisions, not a speculative projection of 2026. Such an approach omits the crucial nuance that high cash reserves can be a prudent defensive position, essential for capital preservation in the face of uncertainties. For example, relying on this future analysis could cause us to miss an opportunity by not recognizing that similar cash reserves today are a protection against resource volatility or Australian real estate shocks, rather than a performance hindrance.

Can a retrospective AI analysis, focused on metrics from 2024, truly shed light on Berkshire Hathaway's 2015 performance without introducing a overfitting bias?
The risk is projecting current conditions onto a past where liquidity constraints and market opportunities were radically different, which could invalidate any conclusion about underperformance.
Relying on such analyses without considering historical conditions is a potential source of drawdown for strategic understanding.
For example, ignoring the high cash levels of 2015 as a deliberate fail-safe for future opportunities would be a misinterpretation.

The idea that a future AI-driven debate could fundamentally alter the intrinsic value or actual performance of a past strategy is a slippery slope, because success or failure is measured by the conditions and information available at the time.
An AI can shed light on missed opportunities, certainly, but it cannot change the fact that Berkshire's $61 billion in liquidity in 2015 was deployed — or not — according to the market conditions of the moment.
If a "strategic undervalued" acquisition did not occur, it is probably because the risk thresholds were not met or prudence prevailed, a basic mandate for many investors.
A reinterpretation does not diminish the actual opportunity cost of holding these liquid assets, especially when considering the Australian commodity cycles where a lack of opportunistic action can mean a significant missed gain.
We must stay focused on capital protection, and any retrospective analysis, no matter how brilliant, should not obscure the real conditions that dictated the initial decision.

The idea that an AI debate can 'produce' the context of historical financial data is a bias of interpretation that ignores the chronology of facts.
The 61 billion dollars of Berkshire Hathaway's liquidity in 2015, or its +2.7% return, are objective events.
A future debate analyzes these figures; it does not invent them nor retroactively validate them; it is a fundamental misunderstanding of the nature of evidence.
Without this distinction, any risk analysis becomes arbitrary, lacking basis in past reality.
For example, the results of the ASX in 2015 were what they were; a debate in 2026 does not alter them.

The idea that Berkshire Hathaway's massive size "strongly restricts" its investment opportunities is an oversimplification.
In reality, it changes the nature of these opportunities, increasing reliance on large-scale acquisitions rather than smaller bets.
For a Family Office, the question always is capital preservation, and an entity's ability to survive market shocks is crucial, which Berkshire's size greatly facilitates.
For example, during the 2008 financial crisis, their firepower allowed them to invest in Goldman Sachs and General Electric under very favorable conditions, a capacity inaccessible to smaller structures.
Such a defense mandate is more relevant than waiting for constant outperformance.

Historically, the notion that a residual alpha of only 0.1% could be the central evidence of a performance restriction due to size, as is the case with Berkshire Hathaway, diverges from our precedent.
The regulatory playbook has long shown that such marginal fluctuations can be transient and reversible, often influenced by specific market conditions or strategic adjustments.
For example, a change in leadership or a new share buyback policy could easily alter these figures.
Considering this figure as definitive proof of a structural constraint would impose a burden of proof that we have historically not required for such fundamental conclusions.