Platformonomics TGIF #83: March 7, 2025

By

Platformonomics TGIF is a weekly roll-up of links, comments on those links, and perhaps a little too much tugging on my favorite threads.

I made a third appearance on the Telco20 podcast to discuss 2024 hyperCAPEX numbers and contrast them with telco industry spending. It is a quick 19 minute review of the numbers, plus I express skepticism about the current outbreak of telco AI dreams (and look back at how their internet/cloud/edge dreams panned out).


News

Welcome to the Capricocracy

Current events don’t just echo the New Deal with broad assertions of executive power, but also capricious and perverse economic policy.

Here’s FDR’s strategic and systemic approach to setting the price of gold:

They met in his bedroom at breakfast. Roosevelt sat up in his mahogany bed. He was usually finishing his soft-boiled egg. There was a plate of fruit at the bedside. There were cigarettes. Henry Morgenthau from the Farm Board entered the room, Professor George Warren of Cornell came; he had lately been advising Roosevelt. So did Jesse Jones of the Reconstruction Finance Corporation. Together the men would talk about wheat prices, about what was going on in London, about, perhaps, what the farmers were doing.

Then, still from his bed, FDR would set the target price for gold for the United States— or even for the world. It didn’t matter what Montagu Norman at the Bank of England might say. FDR and Morgenthau had nicknamed him “Old Pink Whiskers.” It did not matter what the Federal Reserve said. Over the course of the autumn, at the breakfast meetings, Roosevelt and his new advisers experimented alone. One day he would move the price up several cents; another, a few more.

One morning, FDR told his group he was thinking of raising the gold price by twenty-one cents. Why that figure? his entourage asked. “It’s a lucky number,” Roosevelt said. “because it’s three times seven.” As Morgenthau later wrote, “If anybody knew how we really set the gold price through a combination of lucky numbers, etc., I think they would be frightened.”

I’m already frightened, but suspect understanding the tariff strategy and decision-making process would be even more frightening.

CAPEX Clues: CoreWeave S-1

The CoreWeave crypto bros are racing for IPO. They are the largest of the neoclouds NVIDIA has cultivated in hopes of reducing the hyperCAPEX companies’ buying power. CoreWeave spent $8.7 billion in CAPEX in 2024 (basically bonsai/Oracle-scale numbers).

I’ve been pointing out Microsoft’s CAPEX spend understates their GPU fleet, and the CoreWeave S-1 confirms this (CoreWeave claims over 250,000 GPUs):

For the years ended December 31, 2023 and 2024, our largest customer was Microsoft, which accounted for 35% and 62% of our revenue, respectively. Any negative changes in demand from Microsoft, in Microsoft’s ability or willingness to perform under its contracts with us, in laws or regulations applicable to Microsoft or the regions in which it operates, or in our broader strategic relationship with Microsoft would adversely affect our business, operating results, financial condition, and future prospects.

Re: those negative changes in demand:

CAPEX Clues: Softbank Needs More Cash

In a world where we prioritize entertainment value above all else, of course we should fund short-lived assets like GPUs with debt.

But they’re going to need a whole lot more than $16 billion to keep up with the hyperCAPEX companies.

Salesforce Holds “Developer” Conference

Amway Salesforce had a “developer” conference. You can tell it was a developer conference because the keynote speaker wore a t-shirt. An eyewitness described it as “cringe” and lacking in both developer attendance and content. No word whether Salesforce stalwart will.i.am delivered breakout sessions.

Not a good omen for Salesforce if their future requires winning over actual developers. But that shall not deter the marketing machine!

As the industry poster child for not being able to get off Oracle Database, I assume there were PL/SQL sessions aplenty (please see our previous SQL jokes here and here).

Software Migration Alert: Egnyte

As you know by now, when private equity comes amalgamating, it is time to start migrating.

119th Congress Update 1

The 119th Congress remains on recess.

Viewer Mail Episode 3.1 – Elon’s Tesla Strategy

Flat out begging for responses to the question of how Elon is going to extract himself from his Tesla problem elicited exactly one response:

Ultimately and hopefully his endgame involves a bunker and a handgun.

I will attempt to emphasize in the future that responses to these questions will be kept completely anonymous, so you don’t have to worry about a thin-skinned fractional CEO shutting down your Starlink.

5 responses

  1. Great read as always.

    Verizon failed with SD-WAN, NaaS, etc.

    For them to think they will be a player in AI edge compute is… bold

  2. Charles Fitzgerald Avatar

    Thanks! I think the telcos have a regulatory obligation to say they plan to do whatever the new thing is. So they announce Internet/cloud/edge/AI is their thing, and because it involves a network, they’re certain to be successful…

  3. Ah! That is interesting and makes sense. Didn’t think of that…

  4. Charles Fitzgerald Avatar

    I think Microsoft was doing Open AI training on CoreWeave and now getting out of the middle of that…

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Get Updates By Email

Discover more from Platformonomics

Subscribe now to keep reading and get access to the full archive.

Continue reading