FNArena’s Weekly Insights – November 3 2025

What About The AI Bubble Anxiety?

By Rudi Filapek-Vandyck, Editor

‘Bubble’ or mostly ‘misunderstood’?

The debate about the AI megatrend rages on with plenty of oomph and gusto — globally.

Post-August price action for AI beneficiaries on the ASX, including for Goodman Group ((GMG)) and NextDC ((NXT)), shows investors locally remain reluctant to take a firm view either way.

(No such hesitance seems apparent when looking at shares in Megaport ((MP1)).

Meanwhile, technology analysts are preparing for AI shifting into its next phase –coined ‘inference’– when AI adoption will spread across ever more segments and sectors that make up the broad economy.

As stated by analysts at Morgan Stanley recently:

“In a world where AI eventually becomes commonplace, the inference phase sees the deployment/adoption of AI models to use cases across both consumer and enterprise.

“This phase harbours significantly more interesting and long-dated potential and is still currently in the very early stages.”

That view coincides with the observation Australian businesses have only just started to embark on their own AI journey with this years August results season the first to see AI being mentioned regularly and in notable numbers.

In similar vein, it was only less than a month ago TechnologyOne ((TNE)) presented its world’s first fully embedded AI ERP software product, to be officially launched in Q1 next year.

Last week’s investor day by insurer Suncorp Group ((SUN)) was equally dominated by digital transformation and sizeable investment in AI capabilities.

As I suggested during my recent on-stage presentation to Sydney members of the Australian Investors Association (ASA):

The blunt statistics assure us about half of all marriages ultimately ends in divorce.

This still doesnt mean our advice to all newlyweds should be: get a lawyer, now!

Goldman Sachs and the AI bubble anxiety

Within this context, analysts at Goldman Sachs put together a sizeable report, including interviews with expert voices from outside the firm, in an educated attempt to determine whether AI is in a bubble or not.

Don’t look for an ironclad answer as the views shared and conclusions drawn remain polarised and diverse.

But the report contains many valuable insights to ponder and will no doubt make its way around the world to be read by investors looking to update their own ideas and perception.

Below are excerpts from the Goldman Sachs compilation, complemented with my own views and observations.

Enjoy.

Investors are worried

Goldman Sachs: “AI bubble concerns are back, and arguably more intense than ever, given a number of (worrying?) developments: a significant rise in the valuations of many AI-exposed companies, continued massive investments in the AI buildout, and the increasing circularity of the AI ecosystem, with model companies, infrastructure providers, and hyperscalers signing deals with each other that are blurring the boundaries between customers, suppliers, and capacity providers.

“Amid these developments and growing worries that the megacap tech rally may be masking signs of weakness in the broader market, whether AI bubble concerns are warranted –or overblown– is Top of Mind.

“We first ask GS US equity analysts Eric Sheridan (internet) and Kash Rangan (software). Sheridan notes that while some features of the current period rhyme with past bubbles, and the circularity of deals warrants caution, public market valuations and capital market activity levels remain below their Dot-Com peaks.

“He also points out that most of the Mag 7 continue to generate outsized free cash flows, engage in stock buybacks, and pay dividendsbehavior seldom seen during the Dot-Com Bubble.

“So, he seems less inclined to call the situation in the public market a bubble today, though he concedes that AI may just not be a bubble yet.

“Rangan, for his part, sees few signs of a bubble in his coverage universe with many software valuations –if anything– too depressed.

“He more broadly worries about companies increased reliance on debt to fund their AI ambitions.”

FNArena: A large majority of ‘the end of the AI bubble is near’ scaremongering that is spreading like a wildfire across social media platforms and into mainstream publications is based on the observation all prior technological break-throughs throughout history have triggered boom & bust cycles.

Hence, the modern-day megatrend by definition must follow the same script.

A lot of the critiques and warnings stem from ultra-simplistic cherry-picking of charts and data, with facts often misunderstood or misinterpreted.

Give a man a hammer and he starts looking for nails to hit, nest-ce pas?

In all fairness, the numbers involved they do look gigantonormous, but today’s economies are much larger than those from the past and so are the key companies behind most investments made.

Combine this with AI requiring a significant build-out of cloud and compute infrastructure and one should expect today’s numbers to be really, really large.

Does this equal zero return on investment or by definition an abysmal return?

I’d argue there’s only so much that can be modelled and forecast in the early stages of new tech developments (hockey sticks do not combine well with linearity).

At the same time, megatech companies such as Alphabet, Meta and Microsoft are not that worried about returns on their infrastructure build-outs; they have higher goals, and so do governments in Washington and Beijing who fear outright obsolescence.

Simply put: this new megatrend is much more complicated than a simplistic one-on-one comparison with railroads or the Internet.

Key question; what if 2025 turns out to be the historical equivalent of 1995 or 1925?

Are we going to scream ‘bubble, bubble, troubles everywhere’ for another 4-5 years and then claim we were ‘right’ by the time the trend ends (assuming there’s a crash at the end of it)?

History shows if AI truly builds into the next mega-bubble, and then crashes, there won’t be many places for investors to hide.

Those whove lived through the GFC and the Nasdaq meltdown know exactly what I am referring to.

Be careful what you wish for.

Every multi-year bull market leads to pockets of exuberance and risk-taking excesses, and this time is no different.

But as again shown in the Q3 results season in the US, rumours about the death of the AI trade/trend are yet again greatly exaggerated.

In a comical twist, the more scepsis among investors about the validity and duration of the AI megatrend, the less likely it will quickly build into a megabubble that will destroy itself.

Froth, memes, and change

Goldman Sachs: “GS Chief Global Equity Strategist Peter Oppenheimer then digs deeper into the equity market parallels to past bubbles, also finding similarities but key differences — namely, the US tech leaders current strong fundamentals, balance sheets, and AI market positionwhich lead him to agree with Sheridan that the US equity market is not in a bubble yet.

“GS Senior US Equity Strategist Ryan Hammond also shares this view, finding little evidence of US tech sector froth outside of smaller pockets of the market like Quantum Computing.”

FNArena: The most difficult aspect generally for people to deal with is ‘change’ and the past decade, in particular for a traditional value-oriented market full of resources and other cyclicals, has challenged many over-simplified investment strategies and approaches to listed companies.

The ASX has always had a number of companies whose typical valuation would not suit investors looking for low PEs, but for a long time such companies could simply be ignored and dismissed.

Think REA Group ((REA)) and Cochlear ((COH)), for example.

Over the past decade, these companies and a number of newcomers have only become more important for the share market overall.

Imagine this: TechnologyOne –on a FY26 PE multiple of 88x– and Pro Medicus ((PME)) –on a FY26 PE multiple of 171x– are now both part of the ASX50, together with Car Group ((CAR)), ResMed ((RMD)), WiseTech Global ((WTC)) and Xero ((XRO)) — all high PE stocks.

Even without considering CommBank ((CBA)) shares –nowadays making up 12% of the ASX200 on 27x times FY26 forecast EPS– can anyone else see how this by default translates into a higher average multiple for the index?

A second observation I make is the average investor does not understand infrastructure investing.

This then leads to rather ridiculous assessments such as no profits can ever be made from building data centres or why is the return on investment (ROI) negative during the build-out phase.

I asked ChatGPT recently two questions: when did Transurban ((TCL)) list on the ASX and how long before the toll road operator reported its first profit?

The answer: 1995 and 19 years.

Has every toll road ever build always lived up to projections? No. Transurban owns at least one of the local disappointments.

This by no means implies there cannot be an overbuild in capacity and supply in the future, but so far demand is still greater and supply limitations and delays plentiful.

Once the investment phase is complete, infrastructure assets tend to yield returns in the double-digits, and data centre operators such as NextDC are today valued in line with such forecasts.

The Grand Debate about US megacaps is not dissimilar, but at its core there’s a fair argument to be made valuations to date have remained in rather close correlation with earnings growth and cash flows.

Nobody will declare today’s equities are ‘cheap’ or ‘underpriced’ –not as far as the top performers are concerned– but this does not equal ‘excessively priced’ or ‘unimaginable bubbles’.

Theres another important observation (as also touched upon by Goldman Sachs) and that is markets will try to identify both winners and losers from the new megatrend.

The danger here is value-seekers might end up with AI losers that will prove ‘cheaply’ valued for good reason.

Something I think that will increasingly become a feature as AI development progresses.

Success is not guaranteed

Goldman Sachs: “But David Cahn, Partner at Sequoia Capital, sees it differently. He argues that the only way to justify the massive data center buildout forecasted by 2030 –which he estimates will cost several trillion dollars– is Artificial General Intelligence (AGI).

So, in his words, if you believe there’s a data center bubble and there’s going to be an overbuild of capacity, then you want to invest in consumers of compute.

“If you’re a consumer of compute, having an overcapacity of compute means your gross margin goes up and your COGS goes down.

“Byron Deeter, Partner at Bessemer Venture Partners, is less concerned about the amount of capex pouring into the space, with AI reshaping businesses in ways previously unimaginable.

“While he acknowledges that valuations are high today, he sees them as largely justified by AI firms underlying fundamentals and revenue potential.

“And he views the ecosystems circularity less as artificial inflation and more as a reflection of strategic interdependence across the AI value chain.

“So, he believes that AI bubble concerns are overblown, and that this isn’t a ‘hope-and-hype’ cycle like the Dot-Com Era.”

FNArena: Echoing the thoughts and writings of Nassim Taleb: the world is a lot more complex than in the way we tend to summarise and describe it.

Last week, Queensland insurer Suncorp Group ((SUN)) used its investor day to update financial analysts about its digital transformation, including sizeable investments into migrating data onto the cloud, upgrading its ancient platform and integrating AI to shorten claims processing and improve on customer experience.

Suncorp’s strategy highlights one key aspect that is too often forgotten by todays skeptics and critics: companies do not only invest in AI to achieve higher efficiency and better margins; insurers like Suncorp do it to stay relevant and on top of the competition.

And to successfully use and apply AI into their operations, core data need to migrate into the cloud. This is also why CommBank ((CBA)) earlier this year made the same move.

Estimates are rubbery during the best of times, but it is estimated only half of all businesses in Australia is using cloud computing today (same globally).

This does not mean the other half will join in a hurry –many are smaller businesses– but it does add to forecasts of ongoing strong demand for cloud computing infrastructure (i.e. data centres).

Its called a megatrend for good reason.

Making (accurate) forecasts is oh so difficult

Goldman Sachs: “GS Head of the Global Economics team Joseph Briggs is perhaps the most optimistic about the AI capex boom, arguing that the economic value generated by AI will ultimately justify the spend.

“Specifically, he estimates that generative AI will create [US]$20tn in economic value (in present-discounted value terms), [US]$8tn of which will flow to US companies, as it unlocks significant productivity gains.

“And although the range of outcomes remains wide, he notes that plausible revenue estimates generally exceed current cumulative AI investment forecasts even before factoring in AGIs potential emergence, though whether the firms making the investments today will be the ultimate beneficiaries remains less clear.

“But Gary Marcus, Professor Emeritus at New York University, remains skeptical about the promise of the technology in its current form and the amount of capex being spent on it.

“Marcus explains that generative AI remains far from AGI today despite some incremental improvements over the past two years, with the technology still essentially ‘autocomplete on steroids’.

“While he is encouraged by the industrys recent shift away from believing that large language models would be the solution to AGI and toward more promising approaches like neurosymbolic AI, he still sees significant challenges ahead on the road to AGI.”

FNArena: I am old enough to remember the early enthusiasm from the Commodore 64 and Atari gaming.

In hindsight, both were revolutionary for different reasons, but (looking back) they were no more than small, embryonic steps.

For those not old enough: the ’64’ in the Commodore name refers to the ‘unusually large memory for a home computer’ that became available in the early 1980s.

Back then that was 64 kilobytes. To put some context around this: today’s typical laptop has a RAM memory of 16 megabytes; which is 262.000 times larger.

With today’s software requirements, we would not even be able to send a simple email or access the Internet with the computing power from back then.

And similar with the Internet that spread from the early 1990s, and the mobile phone later on, it is plain impossible to predict or imagine exactly how this new technology develops and what will be developed next as a result of it.

But in ten years’ time, when we look back on yesterday, today and tomorrow, it will all look logical and straightforward.

Sometimes it’s too early to make concise predictions, other than there will never ever be another world without AI.

There is no need for a killer application as demanded by so many skeptics; AI is like electricity.

Make it available and entrepreneurs, financiers and innovators will write the next chapters.

And yes, indeed, speculators will eagerly participate too.

What to do as an investor?

Goldman Sachs: “So, what does all this mean for how investors should be positioned? Within tech, Sheridan sees value in stocks well-positioned to benefit from potential AI disruption and underappreciated growth stories, while Cahn and Deeter see substantial investment opportunity in the mostly-private AI application layer.

“But Oppenheimer also sees value in looking beyond the tech sector given the extreme level of market concentration today and ever-present risk of a market correction.

“He continues to recommend diversifying across regions, factors, and sectorsa strategy that has paid off this year.”

FNArena: Don’t put all your eggs into the AI basket. Investing is not about making big, bold bets (if you do, do it with a small portion only).

Managing risk remains all-important in today’s context.

Separate the risk of a share market pull back from the AI megatrend, which most likely will have a lot longer to run and develop further.

Probably best to limit your time and attention on social media. The noise over there can be quite overwhelming and there’s nothing as easy as having an uneducated opinion.

If you’re still uncomfortable with having exposure to your early-stage AI-beneficiaries (which in Australia are mainly connected to data centres) then consider the next phase sees companies making investments and announcing plans and progress.

There will be winners and losers.

As they say, there’s no escaping the future. But you can prepare for it.

More reading:

https://fnarena.com/index.php/2025/11/03/insurance-sectors-ai-challenges-have-arrived/

https://fnarena.com/index.php/2025/10/15/rudis-view-tech1-delivers-ai-first/

https://fnarena.com/index.php/2025/06/19/fnarena-visits-nextdcs-s3-data-centre/

The FNArena service includes a dedicated AI section:

https://fnarena.com/index.php/tag/gen-ai/

Review All-Weather Model Portfolio

The financial year ending on June 30th 2025 featured the return of Donald Trump in the White House and of extreme market volatility.

The second half of the year also saw doubt creeping into general sentiment towards AI and demand for data centres.

All in all, a gain of 13.85% (pre-fees) for the twelve months is not something to be unhappy about, right?

FY25 review of the All-Weather Model Portfolio: https://www.fnarena.com/index.php/download-article/?n=4B38C0EF-A173-8CE6-736A7AFC7B19FC49

Model Portfolios, Best Buys & Conviction Calls

This section appears from now on every Thursday morning in a separate update on the website. See Rudi’s Views for the archive going back to 2006 (not a typo).

FNArena Subscription

A subscription to FNArena (6 or 12 months) comes with an archive of Special Reports (21 since 2006); examples below.

(This story was written on Monday, 3rd November 2025. It was published on the day in the form of an email to paying subscribers, and again on Wednesday as a story on the website).

(Do note that, in line with all my analyses, appearances and presentations, all of the above names and calculations are provided for educational purposes only. Investors should always consult with their licensed investment advisor first, before making any decisions. All views are mine and not by association FNArena’s see disclaimer on the website.

In addition, since FNArena runs a Model Portfolio based upon my research on All-Weather Performers it is more than likely that stocks mentioned are included in this Model Portfolio. For all questions about this: contact us via the direct messaging system on the website).

Want to subscribe to FNArena?

For further information, please visit our website: www.fnarena.com

What is the Weekly Email?

You are receiving this email because you have either trialed or subscribed to the FNArena Australian Broker Call service.

Terms and Conditions

The content of this information does in no way reflect the opinions of FNArena, or of its journalists. In fact we don’t have any opinion about the stock market, its value, future direction or individual shares. FNArena solely reports about what the main experts in the market note, believe and comment on. By doing so we believe we provide intelligent investors with a valuable tool that helps them in making up their own minds, reading market trends and getting a feel for what is happening beneath the surface. This document is provided for informational purposes only. It does not constitute an offer to sell or a solicitation to buy any security or other financial instrument. FNArena employs very experienced journalists who base their work on information believed to be reliable and accurate, though no guarantee is given that the daily report is accurate or complete. Investors should contact their personal adviser before making any investment decision.

© FNArena 2025. All Rights Reserved. No portion of this email may be reproduced, copied or in any way re-used without written permission from News Network. All subscribers should read our terms and conditions.