Oct 15, 2026 · 41 min
Firmus collapse tests AI infrastructure valuations
Stock Take: Lessons From a Failed Float
The failed float exposes how financing structures, fast-obsolete hardware and optimistic growth assumptions can obscure risk in AI infrastructure businesses.
- 1Firmus’s failed listing raises questions about NeoCloud valuations, circular financing and concentrated revenue.
- 2GPU depreciation and changing technology challenge the durability of AI infrastructure economics and headline EBITDA.
- 3Dicker Data and Lovisa show how capital allocation, leadership changes and hindsight shape judgments about long-term winners.
Don't miss
The hosts connect Firmus’s failed float to the broader question of whether NeoCloud economics can survive GPU depreciation, changing technology and financing risk.
The brief
Gaurav Sodhi, Graham Witcomb and Nick Cummings examine Firmus’s failed float, asking whether its proposed valuation reflected a flawed deal or broader enthusiasm for AI infrastructure.
The hosts distinguish NeoCloud operators from traditional data centres, then probe NVIDIA’s role, circular financing and revenue concentration as risks migrate through the AI supply chain.
GPU resale values, rental rates and EBITDA come under pressure from rapid technological change, raising the possibility that Firmus marks a late stage in the investment cycle.
The conversation turns to Dicker Data’s share-price surge, acquisition and leadership changes, weighing dividends and capital allocation against a newly improved outlook.
Lovisa’s CFO departure prompts governance questions, while the hosts explain why strong past performance did not make the retailer an obvious long-term winner beforehand.
What was said on this episode
24 statements · 10 positive · 11 negative · 3 neutral
Firmus sought an approximately $50 billion valuation despite lacking meaningful revenue or earnings.
“they were trying to float their business at, I believe, close to a $50 billion valuation”
Listen at 2:29
NeoCloud business models depend on scarce NVIDIA chips and may fail if alternatives become widely available.
“in a world where chips are more freely available, alternate chips are more freely available, this whole NeoCloud model doesn't even make sense”
Listen at 4:04
NVIDIA uses NeoClouds to diversify revenue beyond hyperscaler customers.
“I think NVIDIA sees these NeoClouds as sort of another way that they can. earn revenue”
Listen at 5:21
A financing withdrawal or AI-trade problem could quickly expose substantial AI-sector commitments and debt.
“if the financing gets pulled or if there's some sort of problem with the AI trade, that all these commitments and all this debt sort of comes home to roost pretty quickly”
Listen at 7:12
Rapid GPU depreciation can make NeoCloud companies loss-making despite positive EBITDA.
“these GPUs depreciate really quickly. So once you get down to profit, they're actually loss making”
Listen at 8:14
Previous-generation NVIDIA chips have retained resale values and lease rates.
“Nvidia's previous generation chips are actually holding up not only their resale value, but also their lease rates have not dropped at all”
Listen at 8:49
Advances in chip technology can quickly make prior-generation chips obsolete.
“the chips advancements in chips can make the you know last general generation obsolete pretty quick”
Listen at 10:14
Firmus’s failed listing may signal that easy money has ended and financing will become harder.
“this is the first sign that the easy money is gone and it's only getting harder from here”
Listen at 12:09
The broader market may not currently be in an AI bubble.
“I'm starting to think we're not in a bubble”
Listen at 13:23
Firmus reportedly sought a valuation above more advanced American peers despite unproven technology and operations.
“they were trying to value themselves above American peers, even though the American peers are further along the monetization journey”
Listen at 14:48
AI infrastructure is receiving trillions of dollars of private-market funding, mainly through debt.
“there's a lot of funding going into this space it's not it's not like it's not being funded at all like there's trillions of dollars going into it”
Listen at 17:57
Dicker Data’s share price approximately doubled over six months.
“over the last, I guess, six months, it's just gone gangbusters. It's pretty much doubled.”
Listen at 19:56
Dicker Data’s Setcar acquisition was its largest, costing approximately $100 million.
“So this is... Dicker's largest acquisition, but it's still relatively small compared to the company's market cap. It was about $100 million deal.”
Listen at 21:15
Dicker Data’s Setcar acquisition appears sensible and is expected to increase earnings.
“So it looks like a pretty sensible deal. It'll be earnings accretive”
Listen at 22:46
Dicker Data reduced its dividend payout ratio from 100% to an 80–100% range.
“the payout ratio has gone from 100% to between 80% and 100%”
Listen at 25:10
Dicker Data’s relatively low acquisition multiple reduces the deal’s risk.
“the fact that it's a fairly cheap multiple also removes some of the risk”
Listen at 27:49
Brett Blundy is the most important figure in LaVisa because he chairs and owns 40%.
“The most important bits of LaVisa are chairman and 40% owner Brett Blundy.”
Listen at 30:27
Brett Blundy’s departure would dramatically change LaVisa’s investment case.
“If he left. I think it would dramatically change the investment case.”
Listen at 30:47
LaVisa’s core advantage is rapidly designing and distributing on-trend jewellery at scale.
“The core benefit that the visa has is that it can design on-trend jewellery and do it en masse. and get it out to stores really quickly.”
Listen at 31:03
LaVisa’s new CEO represents an unfavorable cultural departure from previous leadership.
“I don't like the new CEO. I think listening to that last conference call, it was a very big departure from the CEOs that have come before.”
Listen at 33:14
LaVisa’s recent leadership change is probably unfavorable.
“my early feeling is that this is not a good change”
Listen at 34:17
Investors should not sell or rapidly reverse their LaVisa positions.
“I would not be flipping my stake in LaVisa or selling it.”
Listen at 35:38
LaVisa is buyable at a sufficiently attractive price despite some business-model fragility.
“I would buy it today. I think I need a pretty good price.”
Listen at 38:59
LaVisa’s business model has some fragility, particularly in Australia.
“there is a bit of fragility to this business model”
Listen at 39:05
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.