newsroom
Frontier AI Models · Inference compute demand will inflect as test-time scaling shows no wall yet
now playing · Frontier AI Models
SaaS Business Modelsmixedscore 9/10ed
Public SaaS stocks now filtered by clear good/bad attributes: usage-based, AI tailwind, share gain vs. replicability, incumbent risk, product lameness
The market has moved from indiscriminate SaaS selloff to a structured framework: companies with usage-based revenue, clear AI tailwinds, and share-gain potential (DataDog, CrowdStrike) trad…
US government gating AI model access based on capabilities is a Rubicon moment
The Anthropic Fable ban marks the first US regulation of an AI model purely on capability grounds; if extended to OpenAI/Google models in 3-6 months, it could lead to government controlling…
SaaS Business Modelstailwindscore 8/10ed
SaaS must shift from seat-based to outcome-based pricing to survive AI disruption
Intercom's pivot to per-resolution pricing demonstrates the winning model: align revenue with customer outcomes, forcing vendors into the resolution path and enabling measurable value deliv…
Any liquidity for pre-AI SaaS companies is top-decile performance; transformation is mandatory
Pre-AI SaaS companies growing at single digits with no AI story are 'zombie' assets; the Intercom/Finn outcome proves that burning the boats to adopt outcome-based AI pricing is the only pa…
Robotics & Physical AItailwindscore 8/10ev
Robotics is a trillion-dollar opportunity by 2060 but adoption timeline is decades long
Physical automation faces harder scaling than software due to hardware supply chains, real-world edge cases, and labor cost parity requirements; however, AI-driven flexibility (LLMs on robo…
Gamma squeezes and forced buying distort price discovery in low-float IPOs like SpaceX
With only 4% float trading, options-driven gamma squeezes and index inclusion forced buying can inflate prices far above fundamental value until lockup expiry; shorting is prohibitively exp…
Frontier AI Modelstailwindscore 7/10ev
Inference compute demand will inflect as test-time scaling shows no wall yet
Frontier models continue to improve with more test-time compute (chain-of-thought, agents, repeated sampling); this could create another kink in the token demand curve, benefiting inference…
Sovereign AItailwindscore 7/10ev
Sovereign AI demand will create national model markets even for second-best models
Countries and enterprises will pay for sovereign models they control rather than risk US export restrictions cutting off best-in-class models, creating a viable market for players like Mist…
Open Source AIheadwindscore 7/10ev
Outside China, no good open source models exist; talent to build frontier models is extremely scarce
Pre-training, mid-training, and post-training expertise is concentrated in a few frontier labs; sovereign efforts will likely fine-tune Chinese open weights rather than build from scratch,…
AI model layer will consolidate to a 2-3 player oligopoly in US with European affiliations
The extreme cost of frontier model development makes a small oligopoly inevitable; sovereign efforts in Europe will affiliate with US leaders rather than build full stacks, similar to cloud…
Memory & Storagetailwindscore 6/10ed
Memory stocks trade at very cheap multiples amid AI-driven demand
Memory semiconductors are a neglected beneficiary of AI infrastructure buildout, offering low multiples with high leverage to data center capex.