Indiscriminate Tech Selling Suggests A Degree Of Irrationality

Sell First, Ask Questions Later

Generally speaking, the mindset has been ‘sell first, ask questions later’ whenever a headline regarding some soft of significant AI advancement has crossed news wires. The nature of that announcement, and the sector impacted, hasn’t mattered especially much, with SaaS firms seemingly unable to get out of the market’s crosshairs for the time being.

However, the indiscriminate nature of the selling pressure seen makes little sense. If the argument is that software names will fall victim to AI advancements leading to firms bringing functions inhouse, that should – all else equal – be a boon for the hyperscalers, who will find their data centres in higher demand, and for chipmakers, who will also benefit from said demand. That, however, hasn’t proved to be the case, at least in terms of recent price action.

Broader Context

While that broad-based selling is counter-intuitive, drilling down to the software sector in particular suggests that pressure here may well be over-done.

For instance, the S&P software sector is currently experiencing a bigger drawdown than we saw during the pandemic, and in the aftermath of ‘Liberation Day’. In fact, the present 30-odd% drawdown is – excluding the aftermath of the dotcom bubble bursting – the third biggest recorded back to 1990.

Furthermore, the sector’s RSI indicates the most oversold conditions since the early-90s, again evidencing how rapid and violent the recent downside has been. While that, in itself, doesn’t mean that the sector can’t fall further, it does suggest that things could be getting rather overdone, given that the current environment is very different from those historical parallels.

Valuations tell a similar story. On a 12-month forward P/E basis, the S&P 500 technology sector now trades on a lower multiple than the consumer staples sector. Since the GFC, that has only happened a handful of times – namely, at the start of the pandemic, during the 2022 market sell-off, and briefly in April last year.

To be clear, this decline in tech valuations has little to do with earnings, given that we’ve just had back-to-back record quarterly results in the sector, and almost everything to do with the recent dive that we’ve seen in equity prices.

AI Argument Doesn’t Hold Up To Scrutiny

Zooming out a little, forgetting the recent market action, it’s important to question whether, on a logical basis, fears of AI advancement killing business models are overdone. I’d argue that they are.

While the speed with which models are improving is impressive, we are not in a world, nor are we likely near a world, where companies will be ripping up their SaaS contracts en masse, instead pivoting to solutions that have been ‘vibe-coded’ internally.

Not only would it be inefficient to do so (e.g. why would an insurance company want the rigmarole that comes with building & maintaining a proprietary CRM system when this can be outsourced more effectively), there are also various moats protecting software incumbents, particularly from a regulatory perspective.

Broader Market Implications

In any case, this relentless pressure on tech names does pose something of an issue for the broader market, with the 8 largest companies in the S&P 500 being tech, or tech-related names, and comprising just over a third of the index by weight.

Logically, this leads to two possible paths from here – either, software/tech names do indeed bounce, in turn boosting the market at large, taking spoos back to the top of the recent range around the 7,000 mark; or, the rotation out of software continues, with participants continuing to favour those names more exposed to the ‘real’ economy, in which case the equal-weight S&P (SPW) looks set to continue to outperform the cap-weighted benchmark.

In any event, the current software pressure spilling-over into a broader market rout seems a relatively slim possibility, particularly given the resilient nature of the fundamental bull case, amid robust economic and earnings growth.