The Week Ahead (9-13 Feb): AI Agents Prompt SaaS Extinction Fears + US Job Cuts

11:00am by Roland Head

5 comments 327 reads

Welcome back to The Week Ahead. It’s been a volatile seven days in the market, especially for investors with exposure to software and Big Tech.

I’ve put together some charts to highlight some of the big moves and look at the stories behind them. I’ll wrap up by explaining why investors may also want to keep an eye on the real (US) economy over the coming months, in addition to monitoring the AI story.

Silver & gold lose shine

After falling sharply last week, gold and silver steadied for a few days before resuming their decline with a sharp fall on Thursday:

While many of us (including me) have some exposure to precious metals through UK-listed miners, the massive surge higher in late January had a speculative feel to it. I don’t think it’s surprising to see some of it unwinding.

Identifying whether this is a temporary blip or a turning point in the long-running gold bull market is beyond my pay grade. But for some context, the gold price is still around 70% higher than it was one year ago.

Big tech bets on $655bn AI spend

This week has also seen the AI story evolve in two separate and seemingly opposing directions.

On the one hand, four of the biggest names in tech are now collectively planning to spend c.$655bn on AI capacity during their 2026 financial years:

Amazon.com (NSQ:AMZN) said this week that capex would rise by around 50% to $200bn in 2026, roughly 30% more than market forecasts;
Google owner Alphabet (NSQ:GOOGL) increased 2026 capex guidance to $175-185bn, well above consensus expectations of c.$120bn;
Last week, @Meta Platforms (NSQ:META) said capex would double to c.$135bn in 2026, while Microsoft (NSQ:MSFT) is on track to spend around $140bn in its current financial year, which ends in June.

Markets rightly want to see evidence that these vast spending commitments will drive future profit growth and generate attractive returns. Investors’ views appear mixed, though.

While Alphabet and Meta have largely been given the benefit of the doubt after reporting good commercial progress – for example improved ad effectiveness – Amazon and Microsoft have faced greater scepticism.

Microsoft in particular appears to be under pressure –…

Unlock with Email

Unlock with Google

Already have an account?
Login here

Neil McQuillan

59 mins ago

Perhaps unfairly, I would suggest that writing software code rarely fits into any of these categories. In many ways, it is deeply generic. I wonder if that’s why coding has been the most successful AI application so far.

I would definitely take a different view here while some patterns are generic, building applications that are genuinely high-performance, scalable, and deliver great user experience remains difficult and highly skilled work. If it was easy SASS businesses won't print such high margins, they would have been eroded away due to competition.

What AI has changed is the tooling. The performance of Claude Code and other agentic development platforms such as Warp is genuinely altering the game. Most software that creates real economic value is still complex and highly specific. The fact that AI can now contribute meaningfully in this space is a smoke signal for the rest of the economy. The world we know is going to change radically. e.g. I would take the direct opposite view if AI can produce software systems which is an engineering skill, then the change it can bring elsewhere is going to massive.

I also think software professionals have adopted AI faster simply because we’re used to high-velocity technical change. We’re culturally primed for it. Perhaps this is turkeys voting for Christmas — but it’s also one of the most exciting and progressive shifts the industry has seen in decades.

The SaaS sell-off, in my view, is largely about licensing models. Per-seat pricing breaks down when each human operator can suddenly do far more work. Fewer seats will be needed, and pricing models will have to evolve accordingly.

One caveat: I’ve found agents tasked with open-ended execution to be unreliable. They tend to hallucinate outside their knowledge boundaries and can produce inconsistent results from identical inputs. However, agents used to generate fixed software that automates processes are far more promising. That’s where SaaS platforms can drive dramatic productivity gains per seat.

If software production continues to get cheaper — which I believe it is — then new SaaS entrants with lower cost bases may start to nibble at the heels of incumbents. That said, brand power still matters.

If software production becomes cheaper (which I believe is happening) then perhaps new SASS platforms can chew the heals of those with high costs. I'm not sure here... but time will tell. The brand power of Salesforce (NYQ:CRM) Adobe (NSQ:ADBE) ServiceNow (NYQ:NOW) etc. is significant.

Ultimately, nobody knows how this plays out. Given that uncertainty, being selectively “out” of the SaaS sector probably makes sense for some investors. I’m staying close to the space — waiting for the trend to turn.

Mark Mitchell

43 mins ago

Any improvement on the average coding is welcome. In the last decade or so of my IT career, I noticed I was 'called out' at night more often as the standard of code went downhill for whatever reason - possibly due to management thinking throwing cheap bodies at it would improve the software or possibly the lip service paid to Agile rather than doing it right did not help. LOL.

Kelvin Prescott

31 mins ago

"The SaaS sell-off, in my view, is largely about licensing models." that's the issue, for sure. But, I think that the existing vendors will be able to adapt their licensing models pretty quickly to convert to pricing based on transaction volumes or other (theoretical) models of value add. The ones that have a strong core software product and deep integration into a wider ecosystem may well even benefit. As the majority of code in every software product is never used - because humans don't know about or have the skills to exploit them - an intelligent licensing model that uses transactions rather than seats as the basis of pricing could potentially yield higher overall revenue

6 mins ago

I concur, the P/E and stock market very much value fixed recurring revenues over variables, in a deal I was involved in recently the P/E house pretty much ignored the variable and commission based revenues, going for 5.5x ARR, that said the EBIT valuation was 20x so it seemed fair. However there was some talk about discounting the variables into bundles to shift the valuation up for the next deal, I think that's wrong as I agree with your view that usage based charges are a way to increase profitability as people adopt new functionality.

3 mins ago

I feel your pain, driving speed over quality often leads to success so you can see why that happens. In my experience first to market is the ultimate long term winner. I always remember the early iphones, no mobile internet, patchy phone connectivity but the ultimate long term winner.

Difficult circle to square. I definitely do not think that AI will increase software quality and there will be a lot more of it, so the number of bugs will scale out.

The Week Ahead (9-13 Feb): AI agents prompt SaaS extinction fears + US job cuts

Silver & gold lose shine

Big tech bets on $655bn AI spend

Unlock the rest of this article with a 14 day trial

About the Author

Roland Head

5 comments

Unlock the metrics for all these stocks and more...

Unlock the metrics for all these stocks and more...