AI Newsletter

What's new is the dramatic shift: Google climbs to 86% probability for the best model by end of October, while Anthropic plummets to 12% – a 78.5% surge within one month for Google. In parallel, Anthropic's IPO probability rises to 84%, with expected market capitalization between $2.0 and $2.25 trillion. This combination of Google's technological dominance and Anthropic's capital mobilization through going public marks a new power shift in the frontier AI market.

Anthropic has agreed to launch a European consulting business – a strategically significant move shortly before the expected IPO. This brings Anthropic directly into the core business of traditional IT consulting and positions it as an enterprise partner in a market where regulation and sovereignty are central purchase decisions. The timing – concurrent with the Mistral-NVIDIA deal for industrial AI – suggests intensified competition for European enterprise customers.

According to analysts, the strongest investment thesis for AI in 2026 is not the model itself, but the rollup: Thrive Holdings raised $2 billion at a $12 billion valuation to acquire service companies and embed OpenAI teams. Long Lake acquired American Express GBT privately with a similar approach. This strategy circumvents the monetization problem of pure model providers and uses the existing customer base and margins of established service providers as carriers for AI value creation.

Capital flows are clearly shifting in 2026 away from universal AI models toward vertical AI agents that solve highly specialized domain problems – a trend with 442 likes and broad resonance. OpenAI's acquisition of Orchestra Bio (AI agents for pharmaceutical trials) and Y Combinator startups like EVO HQ, which advise companies on model selection, illustrate this shift concretely. For startups, this means: generic productivity tools lose financing appeal, while niche solutions with clear ROI proof are preferred.

The AI boom is entering a harder phase: investors increasingly question whether billions invested in data centers generate sustainable profits – while inference margins could drop to 30–50% according to analysts instead of the typical 72% for software. Meanwhile, the Stanford AI Index 2026 shows that 88% of companies use AI, but only 17.3% of new AI apps actually reach $1,000 monthly revenue. This gap between adoption and monetization is the sector's central structural risk for 2027.