Sidetrade : Chiffre d’affaires du premier semestre 2026

This is a better signal for vertical software than for “AI” broadly. The key mechanism is that customers are paying for workflow-embedded, domain-specific agents with predictable economics, which supports pricing power and longer contract duration; that should widen the valuation gap versus generic copilots and horizontal SaaS vendors that lack proprietary data moats. The cleanest public-market read-through is NVDA: private inference stacks still require premium GPUs, and this model monetizes hardware without relying on hyperscaler usage-based cloud attach.

The less obvious loser is the public-cloud monetization narrative at AMZN and GOOGL. If enterprises increasingly want sovereign, private, cost-capped AI execution, the spend can migrate away from hyperscaler AI services toward on-prem/private infrastructure, compressing the take-rate on enterprise AI workloads even if total AI demand rises. HCKT gets a credibility halo from the third-party validation angle, but the financial impact is likely more reputational than material unless it converts into higher-margin advisory or benchmarking spend.

Near term, the market may overreact to bookings momentum because the real test is conversion: Q3/Q4 revenue recognition, deployment timing, and whether the new-agent mix sustains after the first wave of early adopters. The 1-3 month catalyst path is proof of repeatability; the 6-18 month path is whether this becomes a durable operating model or just an unusually strong product cycle. Falsifiers are simple: decelerating new ARR, shorter initial contract durations, or weak commentary on customer rollout timing at the next update.