The high multiples of some SAAS companies had bullet proof moats priced in. I wonder if the sharp price reaction is simply that AI is now making the moats appear a little less bullet proof. For instance, is it now a little less impossible that a challenger can replicate PME’s vast web of integrations?

From what I have read and have slowly synthesised over the last few weeks, it’s also about the margins.

if, as seems inevitable, SaaS companies have to use AI to stay relevant, they will have to pay for the cost of that compute. Currently nearly free but likely set to get expensive as AI companies ratchet up the cost and try to become profitable.

This will lead to margin compression and hence the multiple SaaS used to attract is a thing of the past.

How this plays out for each SaaS company is difficult to understand, probably for anyone, hence the indiscriminate selling.

@Chagsy yes, it is fascinating to follow this and I've been thinking along similar lines - not that I have any answers.

My point of this post is to try and put the current attention this is getting in the mainstream public into a long history of innovation, research, testing, approval and commercialisation.

Regarding margin capture, first, there are several ways this might play out, as I think about it. (On reading the following, I think it might be less than clear!)

Model 1: From my understanding, LLMs are converging faster and faster, so the AI companies will need to go downstream to create specialised applications specific to each industry verticle or function, in order to capture margin. We are seeing this, for example, with the specific tools being annouced by the likes of Antropic. In this model, the AI companies come into competition for margin with legacy SaaS players. If the competition is meaningful, margins will compress. (Later on, I explain why I don't think this is the battle that will play out.)

Model 2: New "AI native health specialists" build new tools from scratch and secure a foothold with customers from which they build out over time. Same as Model 1, but new "AI-native start-ups" are distinct from the LLM owners. As such, they'll also be subject to any squeeze from the LLM-owners on token costs, so it is a risk for them too,

Model 3: Incumbent SaaS players integrate AI into their offerings. Depending on their pricing power, they will seek to pass on any AI-token costs to customers, ... like a fuel price or other input cost pass-through in others industries. The token cost will be tranparent. If token cost becomes an issue, customers will presumably be able to select how much "computer power" they want to use, perhaps according to the clinical priority.

As with many other verticle integrations, the ability to capture margin ebbs and flows over time.

So coming back to $PME: is their product, the user experience, the worflow integration, the regulatory compliance assurance, the ability to seamlessly integrate a range of proprietary tools, the security, the seamless control of new releases and features, ,,, and other things ... does it all add up to a strong enough moat? If it does, then I can see that they might be able to pass on token costs or other rent-seeking behaviour from the AI/LLM companies. In that case the'll be able to defend margins, although we might see demand elasticity by the end users depending on just how expensive the tokens get.

So, one thing I am watching for specifically, is does some genius develop a proprietary, analytical capability - likely using AI - that is able to deliver not just image analysis, but analysis and integrated clinical reasoning with an order of magnitude greater efficiency? The quest for this has been going on for years, and there have been many milestones along the way.

At this point it is worthwhile recapping the journey of technology developing in medical image analysis. A very rough history to date has been:

1970s-80s: computer assisted detection (CAD), rules based and heuristic (e.g. edge detection)

1990s: first computer aided detectiin tools in mammography and X-rays (lung nodules). The clinical impact was modest due to high false-positive rates.

2000-2011: Machine learning and "support vector machines" applied to tumor classification, texture analysis, lesion segmentation.

2004-2010: First CAD FDA approvals ,... but clincal value remained debated.

2012-2016: Deep learning revolution;

2012: AlexNet and ImageNet Brealthough

2015-16: First neural networks match or exceed clinician (skin cancer, 2016), dibetic retinopathy (Google Dep Mind, 2016), lung nodule detection.

2015-2016 marked the inflection when AI could match specialist level diagnostic accuracy for certain tasks, e.g. triage

2017: First FDA clearance for autonomous AI: IDx-DR (diabetic retinopathy)

2018-19: Explosion on radiology AI startups. Compaines include triage tools, stroke detection, PE detection, fracture detection.

2019: AI in stroke triage

Breathough moment: shift from research-base to reimbursible tools

2020-2022: Workflow and enterprise inegration (yo, $PME and several others ... vendor neutral software players in the ascedence!)

2020: COVID-19 accelerates AI in CT-based pnemonia classifiers

2021-2022: Platform consolidation, integration of AI tools in PACS

2022: "Transformer architecture" applied to medical imaging

2023: LLMs assist in reporting writing and clinical decision support; early integration into enterprise radiology systems

2023-25: Multimodal models combining imaging, clincal notes, pathology, genomics >>> dawn of unified diagnostic AI assistants

2025-26: AI transitioning from narrow detection / classification into multimodal clinical reasoning.

There is little doubt that the pace of innovation is accelerating. However, in order to capture value, any entity has to gain regulatory approval for the advancement, integrate the innovation into the clinical workflow, and this requires the ability to release changes into the live clinical environment - itself a critical capability - and of course the whole sales and procurement cycle and system investment lifecycle process.

So, I think the current major software vendors who are working hard to integrate AI tools into their platforms, have a strong moat. And that's partly because this isn't a new thing for them. They've already spent 1-2 decades integrating technical innovations - including ML and AI - into their commercial products. The commercial imaging software vendors, of which $PME is one, have been doing this as their core capability. And I believe developing that capability is going to be something that is very hard for an AI company, or even an AI-native-healthtech start up, to do.

I'm not ruling it out, but I think we will see it coming if we keep an eye out for it.

Good question @UlladullaDave

For me, the great concern of the day, AI, sums it up pretty well - "A PE of 300 is completely decoupled from current financial reality". And it's so many standard deviations from the norm it's probably a 1 in 100 year event (or worse), that you can actually do well owning it from there.

Wonderful business...but was on an insane valuation. I'd be interested when it gets below $90 but there will likely be some other fat pitches in good companies being thrown if it does.

It has been great to read all the content and differing opinions from Strawpeople about some of the great tech companies on the ASX who have taken a hammering over the past few weeks. I have used the past few weeks of the SAAS/Resources stocks plummeting with a bit of introspection and taking on some of the learnings from what Ian Cassel's many podcasts have taught me. I feel that right now is probably the best time for me to re-evaluate my portfolio (both on SM and RL) to start re-balancing to companies that will provide me with the best long term returns overall.

The thing that best aligns with my thinking at the moment, was some of the commentary around XRO and AI competition here on the forum. While LLM's and their offshoots might be able to do the job, I am not yet confident that it will be able to provide customers with the same level of trust, reliability and accountability that a dedicated software solution with SME support can offer to your business. I do come at this from a niche position as I currently work in a role where accuracy matters and even a 1-2% hallucination would be unacceptable outcome.

While I don't know much about the likes of PME, WTC or to a lesser extent XRO/TNE, I agree with many here that the market drawdown to date has provided a good opportunity to get into companies which I NEVER dreamed of owning given the massive earnings multiples that they were trading. I feel like ive got a lot of reading to do over the next few months to get myself up to speed, but will be taking some positions on SM to help nudge me into doing the work

I think @UlladullaDave has hit the nail on the head. At the end of the day, it comes down to the valuation relative to the growth (specifically in profitability and cash generation), and those valuations will no doubt have to come down in light of the emerging risks posed by AI.

@thunderhead agree, but I'd argue they had to come down at some point. The market was primed for a catalyst that swept a re-rate through and AI was it. Not doubting that AI will be a negative for some software businesses and a positive for others, but the valuations were so silly on a lot of these SaaS businesses that they were looking for an excuse to crash.

The thesis for so many of them was basically they will become totally entrenched, have no competition and become massive rent seekers. Capitalism ain't perfect, but that always struck me as unlikely to happen at an industry wide level and if it did it would probably have ended up with the heavy hand of regulation coming at it.

The AI panic does seem to be a catalyst but I've been wondering if interest rates are also playing a role. A few years ago there was some debate over whether inflation was transient and maybe raising interest rates wasn't the solution. Then it seemed like inflation was a problem, but interest rates were going to tame it.

I'm don't have a great grasp of the macro side of things but it does seem like we won't be seeing low interest rates again any time soon. The market was adjusting to this but the AI shock has acted as a trigger to send investors heading for the door.