Aug 27, 2026 · 27 min
Higher rates test AI infrastructure and investor discipline
Late August 2026 Random Ramblings
The episode connects interest-rate sensitivity, financing risk, and management capital allocation to the harder question of identifying durable compounders.
- 1Higher rates can compress equity valuations while weakening the economics of heavily financed AI data-center projects.
- 2UWMC and Cogent raise different warning signs about how management teams allocate capital and communicate business risk.
- 3Mark Leonard and Constellation Software provide a contrast between durable compounding and success built on one leveraged bet.
Don't miss
The episode’s sharpest contrast is between Mark Leonard’s repeatable compounding model at Constellation Software and a CEO whose apparent success may rest on one leveraged bet.
The brief
Andrew Walker opens with a broad question: what happens to equity valuations and AI data-center economics when higher rates raise financing costs and make terminal values harder to defend?
The discussion turns to Nvidia, AMD, and CoreWeave as reference points for an AI infrastructure buildout whose returns depend on financing conditions, tenant quality, and sustained demand.
UWMC and Cogent become case studies in capital allocation, with the episode asking whether management decisions reveal durable business judgment or warning signs masked by favorable narratives.
Aaron Chan of Recurve Capital joins the conversation around company analysis, sharpening the distinction between a genuine long-term compounder and a CEO whose record rests on one leveraged bet.
Mark Leonard and Constellation Software supply the episode’s clearest comparison: investors must separate repeatable compounding from apparent success that depends on a single high-stakes outcome.
What was said on this episode
14 statements · 2 positive · 9 negative · 1 mixed · 2 neutral
Higher interest rates reduce the terminal value of data-center assets.
“if interest rates go from 4% to 5%, the terminal value of that asset goes down”
Listen at 10:48
An AI-bubble burst could reduce data-center lease revenue tenfold.
“if the AI bubble burst, well, that lease that went for 100 million to AMD or whatever, if it had to go to a Bitcoin miner or the next best player, I mean, go back to 2023, it would have been going for 10 million per year”
Listen at 13:08
Higher interest rates could cause data centers to crowd out other AI investment.
“if interest rates keep creeping higher and the data centers are a material portion of the AI build, where the data centers are going to start crowding out some of this AI investment”
Listen at 13:31
Rising interest rates crowd out investment.
“as interest rates go up, investment gets crowded out”
Listen at 14:08
Cogent paid large dividends while leverage increased during Sprint integration.
“Cogent, CCOI, the company was paying a huge dividend even as their leverage was really starting to tick up as they integrated the Sprint deal.”
Listen at 16:08
Cogent continued paying large dividends after they ceased making financial sense.
“They paid a huge dividend long past the point of when paying a huge dividend makes sense.”
Listen at 16:24
Management-run capital allocation primarily benefiting executives is extremely rare.
“It is extremely rare.”
Listen at 17:08
Management misalignment occurs when company actions benefit managers over shareholders.
“Anytime someone, this is classic management misalignment, right? Management is doing something for the company that benefits them versus shareholders.”
Listen at 17:13
Self-serving management capital allocation is a major warning sign.
“It's a pretty big red flag”
Listen at 18:24
Constellation Software has been among the best-performing stocks over 20–30 years.
“Mark Leonard is the founder CEO over at Constellation Software, which is one of the best, if not the best performing stock of the past 20 to 30 years”
Listen at 19:49
Mark Leonard compounds Constellation Software over decades.
“He compounds this business over 20 to 30 years.”
Listen at 20:42
Repeated leverage can produce short-term returns but eventually causes a blowup.
“you can get great returns in one year by levering up and YOLOing something. You'll get great returns, but eventually it'll blow up if you keep levering and levering and levering.”
Listen at 21:27
Repeated major success across different industries is very rare.
“It's very rare to find somebody who does something multiple times in different industries and has big successes.”
Listen at 23:26
External environment and chance can substantially affect perceived management success.
“It's just interesting how the dice roll of the world and the environment you're in can impact everything.”
Listen at 24:57
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.
Featuring
Aaron ChanMentioned
Cogent Communications
Nvidia Corporation
Advanced Micro Devices, Inc. (AMD)
CoreWeave
Constellation SoftwareListen to the full episode and explore every guest, topic, and moment on PodLume.