Macro scenario stress tests

Hyperscaler capex is running at roughly US$750–800bn for 2026, about 93 per cent of operating cash flow (33 per cent in 2023). Consensus has US semiconductor earnings growing 98 per cent; NVIDIA is 7–8 per cent of the S&P 500 above US$5 trillion.

Financing has moved down the ladder: Oracle US$55.7bn capex against US$32bn operating cash flow; J.P. Morgan estimates US$1.5 trillion of new technology debt over three years; NVIDIA announced chip-collateralised financing vehicles on 10 August.

July 2026 was the rehearsal: Korea −24 per cent, small-cap tech −12 per cent, networking names −26 per cent in a month. October added Alphabet −12.6 per cent in a day and Intel −29 per cent.

Transmission and result

Look-through AI exposure is about 10 per cent of the portfolio (~$104k): roughly 30 per cent of VGAD, 30 per cent of Ashmore, 20 per cent of Plato and 15 per cent of Barwon; MFS, Langdon, Talaria and Perennial hold essentially none.

Largest modelled losses VGAD, Ashmore and Barwon; the Australian sleeve loses about $48k through risk-off and small-cap de-rating. Offsets: MFS roughly flat, Talaria, Langdon and the hedged defensives — the portfolio loses about half of what VGAD alone would.

What to watch. Nasdaq-100 and semiconductor index 50-day below 200-day; hyperscaler 2027 capex guidance cuts; AI-debt spreads (Oracle, CoreWeave, NVIDIA-collateralised vehicles); Korea and Taiwan leading declines; equal-weight S&P outperforming cap-weight.

Mitigation. VGAD is the cheapest reduction (28 per cent in ten AI-linked names): a hedged equal-weight or value-tilted global index would cut look-through AI exposure by roughly a third. Ashmore's semiconductor concentration is the less obvious exposure. Langdon and MFS are the natural offsets and argue against cutting Langdon entirely.

The Fed hiked on 16 September and held on 28 October with two hawkish dissents; the DXY sits near 101–102, still roughly 15 per cent overvalued on Morningstar's model; the Treasury's buyback programme keeps the debt-dynamics tail risk alive.

AUD/USD at 70.4 US cents with the RBA at 4.60 per cent and a further hike priced for 3 November compresses the rate differential that has supported the dollar. Gold's fall to a 12-month low in October cut against the thesis for now.

Transmission and result

Translation losses on the six unhedged funds of roughly $28–40k are offset by commodities and miners (Perennial +$8k, MQAE +$4k) and the hedged sleeve → net about +1 per cent. Most exposed: Plato, MFS, Talaria.

The dangerous combination is B plus C (foreign selling of US assets): the currency cushion disappears and the outcome moves toward −20 per cent.

What to watch. DXY vs its 100-day average near 99.5 with the descending trendline intact; AUD/USD reclaiming 72.5 cents; gold; Treasury auction tails and the buyback programme; Fed-cut vs RBA-hold timing.

Mitigation. The hedge ratio is the dial: 44 per cent unhedged brought to roughly 30 per cent (hedged classes of Plato or MFS where they exist, or hedged index exposure) halves the B loss on those funds at the cost of the B2 cushion.

A dollar spike in a risk-off episode: AUD −12 per cent, equities −8 per cent locally. September 2026 was this configuration in miniature, except that the AUD rose.

Transmission and result

The unhedged sleeve becomes the hedge (+$20k across Plato, MFS, Langdon, Talaria and Barwon) and the portfolio loses only about 2 per cent. This is the more common pattern and the reason 44 per cent unhedged is not obviously wrong for an Australian investor.

What to watch. AUD/USD breaking below 68 cents with equities falling; the DXY above 104.

Mitigation. None needed: the unhedged sleeve is the cushion. Hedging more (the B fix) gives this up.

The scenario has been rehearsing itself: the Fed hiked in September, the RBA hiked to 4.60 per cent on 29 September with another priced for 3 November, and the ASX 200 fell four consecutive weeks to 8,665 before October's rotation. Unemployment 4.6 per cent with CPI overshooting is the uncomfortable combination.

The March 2026 episode is the fuller template: an oil shock plus rate fears took the ASX 200 down 10 per cent peak to trough in three weeks; MQAE's NAV fell 9.7 per cent, Spheria 12.6 per cent in the month, Langdon roughly 15 per cent in the quarter.

Transmission and result

A recession bear market costs about 17 per cent of the total portfolio. The Australian sleeve loses the most in dollars (small, micro and resources are 22 per cent of the book); VGAD and Langdon lead global losses; only Talaria, Lazard and cash hold up.

About 93 per cent growth assets, weighted beta near 1.0, no fixed interest, gold or absolute return; cash plus the currency cushion are worth about 4 points of protection.

What to watch. ASX 200 losing 9,000, then the 200-day near 8,800, then the March low 8,457; MQAE NAV below 11.83; Spheria below 1.81; MQAE 50-day below 200-day; Australian credit spreads and bank provisioning; RBA language.

Mitigation. Duration is the missing asset: 10–15 per cent in Australian government or investment-grade bonds at ~5 per cent yields offsets a meaningful part of C while the rate cycle is at or near its peak. Trimming small, micro and resources from 22 toward 15 per cent reduces beta more cheaply than any other change; Perennial is the largest single scenario-C loss for its size.