The Twenty Minute VCAnthropic Raises $45B but Falls Short on Compute & Thoma Bravo Hand Back Medallia Keys to Creditors
At a glance
WHAT IT’S REALLY ABOUT
AI compute economics shift as agents reshape software, exits, PE
- OpenAI’s reported growth miss is framed as a backward-looking reflection of last year’s model leadership gap versus Anthropic, rather than a definitive signal of long-term weakness.
- The panel argues AI agents will increasingly choose tools and models, shifting advantage away from human preference and toward whoever controls the agent layer and the most “agent-compatible” workflows.
- Anthropic’s $45B hyperscaler-backed financing is interpreted as a response to compute scarcity, highlighting the extreme capital intensity and forecasting risk of scaling foundation-model businesses.
- They challenge the simplification “compute equals revenue,” emphasizing that compute only monetizes when paired with a competitive model and sufficient demand, creating volatile boom/bust compute cycles.
- Thoma Bravo’s Medallia handover is presented as evidence that “overpaying” (not just over-levering) can wipe out equity, contributing to a collapse in PE as a reliable exit route and forcing VC to rely on fewer, much larger outcomes.
IDEAS WORTH REMEMBERING
5 ideasOpenAI’s miss is treated as a lagging indicator, not a regime change.
Rory argues the market reaction reflects last year’s period where OpenAI under-shipped on model quality, allowing Anthropic to gain share—yet current releases (e.g., coding performance) may have already reversed the narrative.
Agents may erase “human preference moats” between models and tools.
Jason’s claim is that as workflows become agent-run, the selection criteria shifts to what agents optimize for (API reliability, integration, cost/performance), making last year’s human-led “Claude advantage” potentially temporary.
Controlling the agent layer could become the real lock-in point.
If the dominant agent framework defaults to its own model/provider, the winner isn’t just the best LLM—it’s whoever owns the orchestration layer that makes purchasing decisions (Sam/Benioff-style “agent wars”).
Anthropic’s hyperscaler money signals compute scarcity and strategic tethering.
The $45B package is interpreted as a way to secure capacity and reduce near-term bottlenecks, but it also deepens Anthropic’s dependence on Google/Amazon—potentially making the hyperscalers the “house” that wins either way.
“Compute equals revenue” is a dangerous simplification; sequencing matters.
They reframe it as “no compute means no revenue, but compute + weak model still means no revenue,” implying scaling requires synchronizing model quality, demand creation, and capacity procurement.
WORDS WORTH SAVING
5 quotesThe dirty little secret of venture, again, is how much of your money you make in that one year in 10 when everybody buys the dream.
— Rory O’Driscoll
They just mask decay. Churn, churn that is, churn that is deferred still exists, and it is, it is where the rent-a-CEO and the mediocre hide.
— Jason Lemkin
No compute equals no revenue. But compute and a shitty model also equals no revenue. See Groq for details, right?
— Rory O’Driscoll
You can't service two billion-plus of debt on a one billion, low-growth company with a pre-AI story that has to transform to AI. You simply can't.
— Rory O’Driscoll
It's entirely plausible in a world of super big exits that 10 super big exits cover the entire nut from the LP perspective such that it's still a good business, and that literally nobody cares about the fact that the other 96 companies wither off on the vine, right?
— Rory O’Driscoll
High quality AI-generated summary created from speaker-labeled transcript.
