Latest / KDAV Radio / Better Offline - Monologue: Australia's Dodgiest AI IPO Is A Pale Horse

Transcript

Better Offline - Monologue: Australia's Dodgiest AI IPO Is A Pale Horse · KDAV Radio

• 0:02
  Speaker 1: Hello and welcome to this week's Better Offline monologue, I'm
• 0:07
  Speaker 1: your host Ed Zitron. Thousands of miles away in the
• 0:18
  Speaker 1: strange world of Australia, a pale horse has emerged in
• 0:21
  Speaker 1: the form of Firmus, an NVIDIA-backed AI data center company
• 0:24
  Speaker 1: with a whole $ 50 million in revenue that is, or
• 0:27
  Speaker 1: perhaps was, trying to raise about $ 5. 5 billion in its IPO,
• 0:31
  Speaker 1: which would have been, I think, the second largest in
• 0:33
  Speaker 1: the history of the country. At first, Firmus seemed like
• 0:37
  Speaker 1: it would be another Australian core weave, another vehicle to
• 0:40
  Speaker 1: raise debt to buy NVIDIA GPUs, backed by a $ 10
• 0:43
  Speaker 1: billion loan from Blackstone and Coichu, and a plan with
• 0:47
  Speaker 1: the data center Magnet called CDC to build $ 73 billion
• 0:51
  Speaker 1: worth of capacity in what was once called Project Southgate.
• 0:54
  Speaker 1: with plans to build the country's largest data center by
• 0:56
  Speaker 1: April of this year. Just any guesses, you know, put
• 1:00
  Speaker 1: down some money. How likely do you think that was?
• 1:02
  Speaker 1: Do you think it happened? Do you think it did? Anyway,
• 1:06
  Speaker 1: Firmus has every backup that you'd expect from a dodgy
• 1:09
  Speaker 1: data center company, funding from NVIDIA, Jane Street, and Blackstone,
• 1:12
  Speaker 1: all of whom also back CoreWeave and multiple other AI
• 1:15
  Speaker 1: data center projects, and an initial commitment from Meta to
• 1:18
  Speaker 1: artificially boost its potential earnings by saying yes, Once you
• 1:21
  Speaker 1: build one of these data centers, which you have never
• 1:23
  Speaker 1: done before, by the way, we'll absolutely pay you for it,
• 1:26
  Speaker 1: under the direct assumption that very little will get done
• 1:29
  Speaker 1: for many, many years, and I will say the correct one, too.
• 1:33
  Speaker 1: It's unclear what happened to the data centers, other than
• 1:35
  Speaker 1: the fact they haven't been built, but you've probably worked
• 1:38
  Speaker 1: out by now why I keep referring to Project Southgate
• 1:41
  Speaker 1: in the past tense, as CDC pulled out of the
• 1:43
  Speaker 1: project a few days ago with around 2.5% of the
• 1:46
  Speaker 1: capacity built, and that's less than a month from when
• 1:49
  Speaker 1: this company was meant to go public. Not a great sign.
• 1:52
  Speaker 1: It's unclear who else Firmus will build with, but it's
• 1:55
  Speaker 1: got bigger problems to deal with. You see, its initial
• 1:58
  Speaker 1: IPO valuation was meant to be around $ 50 billion. Australian,
• 2:02
  Speaker 1: I'm just saying dollars, just get over it. Except when
• 2:05
  Speaker 1: banks started offering shares, it priced around $ 11 a piece,
• 2:09
  Speaker 1: or a valuation of $ 43. 7 billion, followed by Australian pension
• 2:14
  Speaker 1: fund Unisuper saying that, and I quote, it all comes
• 2:17
  Speaker 1: down to valuation... and that it will not have any
• 2:20
  Speaker 1: meaningful exposure, citing that Firmus may have a compelling story,
• 2:26
  Speaker 1: but not a compelling valuation. And for those of you
• 2:30
  Speaker 1: at home who've been playing, is that good? That is
• 2:33
  Speaker 1: not good. Per the Australian Financial Review, Firmus has initially
• 2:37
  Speaker 1: claimed in its original term sheet that it had indicative
• 2:40
  Speaker 1: orders that priced above $ 11 a share. Only for the
• 2:44
  Speaker 1: pricing of the IPO to be cut to $ 8. 25 a share,
• 2:48
  Speaker 1: with the AFR now reporting that the company may end
• 2:50
  Speaker 1: up taking anywhere between $ 6 and $ 7 apiece to get
• 2:53
  Speaker 1: the thing out the door. As noted by hedge fund CEO,
• 2:57
  Speaker 1: CEO of Plato, David Allen, speaking to the AFR, if
• 3:01
  Speaker 1: the IPO was so well covered at $ 11 a share,
• 3:03
  Speaker 1: why was it cut? Well, I think the answer is obvious.
• 3:07
  Speaker 1: Firmus is a fucking stinker. 90% of its revenue backlog
• 3:10
  Speaker 1: comes from two customers, Meta, as mentioned, and an unnamed
• 3:14
  Speaker 1: major global technology company, as well as, though this wasn't
• 3:17
  Speaker 1: in the prospectus, a multi-year deal to provide compute-to-open AI
• 3:20
  Speaker 1: out of data centers in Malaysia that do not exist and,
• 3:24
  Speaker 1: let's be honest, will never exist. Indeed, I don't know
• 3:27
  Speaker 1: why nobody pointed at that within these stories. I've read
• 3:30
  Speaker 1: a lot of Firmus stories and I haven't really found
• 3:32
  Speaker 1: much discussion of that. It's almost like everyone just kind
• 3:35
  Speaker 1: of assumes that that was a bit. I think it's
• 3:37
  Speaker 1: a joke. A little funny one, except a multi-billion dollar joke.
• 3:42
  Speaker 1: You know, much like Firmus. As mentioned at the jump,
• 3:45
  Speaker 1: Firmus has a whole $ 50 million in revenue, annually speaking,
• 3:49
  Speaker 1: but plans to build 10 gigawatts of capacity in the
• 3:52
  Speaker 1: next five years, costing it hundreds of billions of dollars
• 3:54
  Speaker 1: that it just doesn't have and nobody's going to give it.
• 3:58
  Speaker 1: Every promise underlying Firmus' valuation comes down to, to paraphrase
• 4:02
  Speaker 1: Plato's David Allen, flaws the execution of some of the
• 4:05
  Speaker 1: most ambitious infrastructure projects in history, and as Allen noted,
• 4:09
  Speaker 1: valuing Firmus based on its EBITDA multiple, as in working
• 4:12
  Speaker 1: out how to value, How valuable it is based on
• 4:14
  Speaker 1: the cash that it might bring in is kind of
• 4:16
  Speaker 1: like valuing an airline based on planes being free and
• 4:19
  Speaker 1: lasting forever. In any case, as I read this, it's
• 4:23
  Speaker 1: unclear whether Firmus lists at all, or if it does,
• 4:26
  Speaker 1: how much it'll actually list for. The Australian financial media
• 4:30
  Speaker 1: has been beating the absolute shit out of it for
• 4:32
  Speaker 1: months in a way that should embarrass the vast majority
• 4:35
  Speaker 1: of journalists that have covered data centres in America and England.
• 4:39
  Speaker 1: pushing back against basically every ridiculous claim and talking to
• 4:42
  Speaker 1: skeptic after skeptic, mocking the company for its dodgy metrics
• 4:46
  Speaker 1: and scrutinizing the banks backing it. A great example is
• 4:49
  Speaker 1: Mark DiStefano of the AFR, who made the point yesterday
• 4:52
  Speaker 1: morning that banks had, as mentioned, said the IPO was
• 4:54
  Speaker 1: effectively filed at $ 44 billion, despite the fact that it
• 4:58
  Speaker 1: was anything but, and that there'd be potential regulatory problems
• 5:01
  Speaker 1: as a result. Can you imagine any of this kind
• 5:03
  Speaker 1: of pushback against Corweave, Nebius, Iron, or any of the
• 5:06
  Speaker 1: other dogshit merchants raising billions to maybe build a data
• 5:09
  Speaker 1: center someday? No, no, no, no. We absolutely couldn't here
• 5:12
  Speaker 1: in America. It'd be considered uncouth to point out their
• 5:15
  Speaker 1: financial fragility. In fact, if anything, we should help pump
• 5:19
  Speaker 1: these stocks so they can raise more money to send
• 5:22
  Speaker 1: to our glorious god, Mr. Jensen Huang. Firmus is set
• 5:27
  Speaker 1: to go public in the next few weeks on October 23rd,
• 5:30
  Speaker 1: but the rumblings are that it might not even make
• 5:32
  Speaker 1: it out the door, and if it does, it'll be
• 5:34
  Speaker 1: at a vast discount. It's a sign that investors have
• 5:37
  Speaker 1: finally started to see the AI bubble for what it is,
• 5:40
  Speaker 1: a bunch of grifters trying to dump questionably valued assets
• 5:43
  Speaker 1: onto an unsuspecting public under the auspices of innovation, kind
• 5:47
  Speaker 1: of gussied up by the press. I'll also add that
• 5:50
  Speaker 1: basically every single problem and criticism that people have with Firmus,
• 5:55
  Speaker 1: revenue concentration, questionable delivery timelines, unbelievable debt load today and
• 6:00
  Speaker 1: in the future can be levied against every single American
• 6:04
  Speaker 1: AI data center company without exception. And that the only
• 6:07
  Speaker 1: real difference is that some of those data center companies
• 6:10
  Speaker 1: already had some, but not much capacity built as a
• 6:13
  Speaker 1: result of their cryptocurrency mining operations. CoreWeaver was smart to
• 6:17
  Speaker 1: go public in 2025. And I think Lambda, another Nvidia
• 6:20
  Speaker 1: backed Neo cloud planning to do so in 2027, is
• 6:23
  Speaker 1: going to have a few more problems. The Australian coverage
• 6:27
  Speaker 1: of Firmus should be an embarrassment to the American finance
• 6:29
  Speaker 1: and tech press because it shows what actual journalism looks like.
• 6:34
  Speaker 1: As I read this sentence, I see a headline from
• 6:36
  Speaker 1: Bloomberg about Corweave signing a deal with Adani Connects. You know,
• 6:40
  Speaker 1: that guy who was under massive fraud allegations. But they've
• 6:43
  Speaker 1: all gone away. Thank you, our president. in India for
• 6:46
  Speaker 1: 240 megawatts of capacity that won't come online until the
• 6:48
  Speaker 1: middle of 2028, which is functionally identical to the promises
• 6:51
  Speaker 1: made by Firmus. Except, instead of saying that, the article
• 6:55
  Speaker 1: chooses instead to talk about how CoreWeave has, and I quote,
• 6:58
  Speaker 1: been expanding geographically in order to boost available capacity and
• 7:02
  Speaker 1: better serve global customers. When all of this shit finally collapses,
• 7:06
  Speaker 1: something has to fundamentally change about how the business and
• 7:09
  Speaker 1: tech press does its fucking job. The amount of times
• 7:12
  Speaker 1: a neocloud or AI company has managed to get a
• 7:14
  Speaker 1: glossy piece about some theoretical project as a means of
• 7:17
  Speaker 1: raising more debt or venture investment off the back of press,
• 7:20
  Speaker 1: unquestionably repeating their marketing collateral, is actively dangerous for investors
• 7:24
  Speaker 1: and the general public. It is very, very, very easy
• 7:27
  Speaker 1: to see the problems in this industry, but I guess
• 7:29
  Speaker 1: it's much harder to do so if you're intentionally choosing
• 7:32
  Speaker 1: to ignore them. Firmus is as bad as CoreWeave was,
• 7:36
  Speaker 1: and is as bad as Lambda. Perhaps they have a
• 7:38
  Speaker 1: little bit more infrastructure, but their problems are identical. And
• 7:42
  Speaker 1: I know that there are journalists who are listening to
• 7:43
  Speaker 1: this who might be a little bit butthurt, but it's
• 7:45
  Speaker 1: not my fucking problem or my fucking fault that nobody
• 7:49
  Speaker 1: seems to be able to criticize these companies. I challenge you,
• 7:52
  Speaker 1: if you feel like your feelings are hurt, to go
• 7:54
  Speaker 1: and read the coverage of Thermos if you cover AI
• 7:57
  Speaker 1: data centers and see what the difference is between the
• 8:00
  Speaker 1: American coverage and the Australian coverage. Because it's distinct. And
• 8:05
  Speaker 1: had anyone bothered to take CoreWave or Iron or Nebius
• 8:10
  Speaker 1: or Lambda or any of the American ones seriously like
• 8:13
  Speaker 1: the Australians have with Thermos, maybe we wouldn't be in
• 8:17
  Speaker 1: this situation we're in today. And maybe the future stock
• 8:20
  Speaker 1: crash that happens, which fucks every investor, retail and otherwise,
• 8:24
  Speaker 1: that is connected to a NeoCloud, maybe we could have
• 8:26
  Speaker 1: avoided that. We can't now. So thankfully, it looks like
• 8:30
  Speaker 1: Australian investors won't get harmed as much as American ones.
• 8:33
  Speaker 1: But when it comes down to the point when this
• 8:36
  Speaker 1: stuff falls apart, when it comes to the point when
• 8:38
  Speaker 1: people get hurt, people are going to ask who to
• 8:40
  Speaker 1: blame and the answer is partially the media. If you
• 8:44
  Speaker 1: remember of the media hearing this, maybe it's not about you.
• 8:48
  Speaker 1: Maybe you did your job. Maybe you were critical. Or
• 8:52
  Speaker 1: were you? Go and check your work. Go and check
• 8:56
  Speaker 1: whether you actually criticized and actually scrutinized and were an
• 8:59
  Speaker 1: actual skeptic of these companies. Or whether you spent far
• 9:03
  Speaker 1: more effort scrutinizing the scrutinizers. Because that's how we get bubbles.
• 9:09
  Speaker 1: That's how hype cycles grow. And that's how regular people
• 9:12
  Speaker 1: get fucked. Fucked as retail investors or those who are
• 9:16
  Speaker 1: having their insurance funds put into these companies. And guess what?
• 9:20
  Speaker 1: Guess fucking what? You know who actually backed Firmus? Blackstone's
• 9:25
  Speaker 1: credit and insurance fund. So when this dog dies, and
• 9:29
  Speaker 1: it will, it will be directly as a result of
• 9:33
  Speaker 1: nobody really pointing that out and allowing and actually greenlighting
• 9:37
  Speaker 1: and approving and celebrating companies like Blackstone sinking pension funds
• 9:42
  Speaker 1: and insurance funds into these dodgy goddamn companies. All of
• 9:46
  Speaker 1: this could have been avoided, and my anger that you
• 9:48
  Speaker 1: hear in my voice is because I've been saying it
• 9:51
  Speaker 1: for years. And I hope something changes as a result
• 9:55
  Speaker 1: of this. I will be calling for change. I've been
• 9:58
  Speaker 1: taking detailed notes. Oh, how I have. And I will,
• 10:02
  Speaker 1: of course, be back next week with a great episode
• 10:05
  Speaker 1: on Bournes. And, of course, another monologue. It's always a
• 10:08
  Speaker 1: pleasure to serve you. Thanks as ever for listening. I
• 10:11
  Speaker 1: love you all.