The tech see-saw
What used to be a two-seat board is now a four-way tilt
Since COVID, a multi-year rush of US-led, pro-growth fiscal spending carried the global economy, sweeping markets with it, through one of the strongest stretches in modern memory. But a rush doesn't lift everyone evenly. As the winners kept winning, some stratospherically so in both size and speed, the market simplified into a single question: tech or not-tech, or US /non-US, like a see-saw with two seats.
That framing is no longer good enough.
The mega-cap trade has fractured internally, and the stakes have become increasingly quad-polar. The honest picture now looks less like a playground plank and more like the four-armed see-saws you see in carnivals - a central pivot with four seats, each arm able to rise or fall independently. Software and hardware can win at each other's expense within tech; the rest of the economy can reclaim ground without tech collapsing; or a genuine recession can drag every arm down together. But anyone who's sat on a see-saw knows how it ends: the bigger kid can't stay suspended in the air for too long. Sooner or later, the heavier end comes back down to the ground. Mega-cap tech is comfortably the biggest kid on the plank, and we don't think the laws of the playground stop applying just because the arm is longer this time..
By the numbers
Microsoft's market cap fell ~US$823bn in the first six months of 2026 - then bounced back ~US$680bn in July alone (including the biggest one day gain ever for a stock.
That single month's bounce is roughly the entire market cap of Microsoft at the end of 2017, or NVIDIA's entire market cap in Q1 2023 - swung in a matter of weeks
1. Software wins
AI's economics settle at the application layer. Distribution and data moats capture the return; infrastructure spend growth eases as the buildout matures.
• Enterprise AI software revenue outpaces infrastructure capex growth
• Capex intensity peaks and starts to ease
2. Hardware wins
The bottleneck stays physical. Compute, chips and energy remain scarce enough that infrastructure keeps capturing outsized economics as the application layer commoditises.
• Infrastructure and capex spend keeps compounding, no sign of peaking
• Chip and energy supply constraints persist or worsen
3. Rest of economy wins
Tech normalises without breaking. Growth eases toward sustainable levels, the AI premium compresses gently, and capital rotates to sectors starved of it.
• Market breadth widens as non-tech earnings revisions turn positive
• Mega-cap multiples de-rate gradually, not abruptly
4. Recession
None of the above matter. A genuine contraction – caused by war, weather, a credit event, or exhausted consumers - drags every arm of the board down together.
• Labour market and yield curve signals turn decisively negative
• Consumer spending data rolls over broadly, not just at the margin
Our view: Leaning into #3 and #4
We think the market is too optimistic - not just on the near-term earnings trajectory of the mega-cap winners, but also on it continuing over the next three to five years. If profit growth for tech's biggest winners keeps compounding at today's elevated levels over that horizon, the value being created has to go somewhere: it either broadens out to more companies, or it concentrates further into fewer hands. We don't think it comfortably does both, and we're positioned around quadrants three and four rather than one or two.
The market has started answering that question. Over the past quarter, breadth has genuinely improved - our active managers have seen strong outperformance as capital rotates beyond the mega-cap handful, consistent with quadrant three, tech normalising as the rest of the economy reclaims ground. But we're treating quadrant four as the other live outcome, not a separate tail risk. If broadening reverses and the winners keep taking share instead, the companies losing that battle don't just underperform quietly - they have to cut costs aggressively to stay competitive, and cost-cutting across enough of the economy is itself a recession input.
In our view, concentration and recession risk are less two alternatives than they are two stages of the same scenario.
More like this
