The Secret CFO

How did one VP of Finance bring Macy's to its knees with a $151M accounting scandal? Here’s how I think it all went down… (THREAD)

@27XVII As somebody from outside the world of software I always found it utterly bewildering

Anyone that mentions Rule of 40 should be heavily discounted in this new world. The rule of 40 was created to compare hypergrowth, cash-burning companies to mature, cash-flowing companies. Those days are are over and done - showing you still view the world through this lens indicates a lack of understanding of what is going on.

@OnlyCFO He didn’t know enough CEO

Mindbody CEO was ordered to pay $48M in damages to common stockholders for a breach of fiduciary duties. A lot of folks are put in a "fiduciary" role and don't even know what that means. It carries significant responsibility (and potential liability) onlycfo.io/p/fiduciary-du…

@OnlyCFO @malhotra_monal @TheSaaSCFO I hope there’s enough CFO to go round

@malhotra_monal @TheSaaSCFO "knows more CFO than you" ok

The ROI I get from Claude has broken my brain on all vendor renewals “You want me to pay 2x what we pay for Claude for your little workflow tool?” I have never pushed harder on pricing and threatening to churn in my life…

@lucas__crespo Soon to be surpassed : )

this is so so visually delightful!

@TJ09299872 you are right about that. but i do know a piece of coal doesn't take 3 years to burn...

@SecretCFO What you're not familiar with is the economics and underlying architecture of neural networks.

OpenAI has a real problem imo. The public capital markets are the only place they’ll be able to raise the kind of capital it says it needs. The quantum is just so huge. But at the same time they are so far from being ready for the scrutiny of that world. It doesn’t have any of the business model clarity, financial history, or management maturity to make it work today. Throw on top the opaque round trip partnership deals, and it’s just so far off. ‘Size’ alone (however you want to measure that) in no way dictates public company suitability. Good to hear the CFO is finally standing her ground a little…

@TJ09299872 - on a long enough time horizon yes. - on a 3 year cycle (if that's right) not that I'm aware of no. - as big in $ terms as we are talking about here - again no, I very much doubt it.

@SecretCFO Is there a comparable business where maintenance capex = 70% of spend?

@TJ09299872 dude i am very familiar with this concept. it's called maintenance capex. it's not a unique concept to data centers (but the size of it is) read this 3 quarters of the way down the page cfosecrets.io/p/allocating-c…

@SecretCFO GPUs don't have maintenance like a forklift to extend life. They are not comparable to any normal piece of equipment. The money used to purchase them is burned with nothing left to show in 3 years.

@TJ09299872 no i'm not at all. depreciation period is totally irrelevant. what you are talking about is called maintenance capex, and i'm pretty familiar with how it works. it's not a novel concept. what is novel here is how big it is in the case of data centers build out.

@SecretCFO @CollettKevin You're hung up on the accounting depreciation. This isn't that. You can depreciate an asset over 3 years and use it for 10+. This is retirement. The thing you bought must be thrown away and repurchased every 3 years.

@TJ09299872 you're right, its not a perfect analogy. but infinitely better than coal... it's literally nothing like that at all.

@SecretCFO @CollettKevin No, its not like building a factory. The vast majority of the money goes to the chips, not the building, the land, etc. The chips are $40k each, and you need thousands of them. I'm not aware of any factories where every single piece of equipment is thrown out every 3 years

It’s not a consumable if the time to zero is measured in years. So the coal analogy doesn’t work It’s more like building a factory… land, infrastructure, buildings, fittings, equipment, etc. In case of data centers big part of it is GPU spend, but not all. And that spend is just like depreciating equipment in a factory it wears out, goes obsolete, gets superseded, etc. And eventually the economics of a unit of compute will have to reflect all of that…

@SecretCFO @CollettKevin GPUs are a consumable. The value goes to 0. At some point the GPU becomes so expensive to run compared to newer generations, that you throw it out. No one will want/use it. And the time to 0 is a handful of years

@TJ09299872 @CollettKevin ok so coal that takes years to burn? ... not exactly the same is it. but yes there is an ongoing maintenance capex load with the data center infrastructure for sure. and no corporate user has the first clue what those long term economics look like yet...

@SecretCFO @CollettKevin Uhh, the companies buying them are writing them off on a 5-6 year schedule, e.g. they'll be thrown in the trash. And considering advancements and drops in power usage, their useful life is lower, perhaps 3 - 3.5 years.

@TJ09299872 @CollettKevin You’re right, I don’t. Explain it to me?

@SecretCFO @CollettKevin Then you don't understand that buying GPU's (where all the money goes) is like buying coal, not building railroads

Anthropic just announced that in February 500 companies were spending $20k pw on Claude ($1m annualized run rate). Just two months later that number has doubled to 1,000. Pretty insane

I haven’t spent much time here recently, so probably catching up , but the ai replies issue has sure got a lot worse over the last few weeks

@CollettKevin I was referring to the demand for computational power. I think that is increasing by an order of magnitude whatever happens. Creating silly AI generated videos (for example) uses a huge level of compute.

@SecretCFO I am unconvinced by the certainty of the expected “demand explosion” for AI. I don’t see companies vibe coding replacements for their SaaS and software tools. Several companies also appear, anecdotally, to be re-hiring staff that they laid off due to AI-based efficiencies too.

@MBGilroy If the can access the capital they need privately there is not one single good reason for OpenAI to IPO

@SecretCFO I completely agree with all of the sentiment here other than access to capital. They are staying in the exact right place (private capital) to access all of the capital they need for quite a long time. There are 10 different examples of this in the private markets rn

This is going to blow minds on LinkedIn when they hear about it next week

I actually think investing in massive compute capacity is not terribly risky in the long run. They are hard assets that will provide capacity among a demand explosion for what today is clearly a scarce resource. But where that sits in the value chain and how that gets funded, commercialized and collateralized is a different question. That’s down to the deal structure … and no one understands those deals right now

@SecretCFO All these data centre projects look super risky too.

@jwsktweets Uber raised something like $8b at IPO. It’s not the same at all.

@SecretCFO Uber had the same case, turns out fine

@KevinEspiritu @nikpil06 @tbpn 😂

@nikpil06 @tbpn thought about it it's not

openAI acquiring @tbpn is basically like google acquiring youtube think about it