AI & data development
AI and data companies approaching public markets
Translate AI capabilities into paying demand, delivery economics, cash collection and durable competitive strength.
01Locate the business in the value chain
Separate chips, computing infrastructure, models, applications and data engineering. Determine what the customer buys and which capabilities are owned rather than sourced. Hardware resale, project delivery and recurring software have different costs and reinvestment needs. Analyse mixed businesses by segment before assigning a valuation framework.
02Distinguish demonstrations from demand
A demonstration is not recurring willingness to pay. Identify the customer's problem, budget owner and measurable benefit. Integration, permissions, training and procurement can delay deployment. Track trial, production use, expansion and renewal separately. Customer counts should distinguish free use, paid relationships, affiliates and distribution reach; missing disclosure is not evidence of retention.
03Measure complete delivery economics
Include inference, storage, networking, review, support and failed attempts. The useful unit is a billable outcome, not necessarily an API call. Scale may improve utilisation while also increasing isolation, reliability and specialist-support requirements. Test whether each additional customer improves contribution economics rather than merely enlarging an uneconomic activity.
04Examine usable data rights
Data volume is not equivalent to unrestricted, durable access. Review provenance, permitted uses, renewal, quality controls and contractual limits. Record count says little about relevance, accuracy or coverage. Concentration in a few data suppliers creates renewal and repricing exposure. Distinguish owned tools and accumulated expertise from resources available only under temporary agreements.
05Connect research spending with renewal
Large R&D spending alone does not prove leadership. Link it to product milestones, commercial adoption and the useful life of existing products. Review accounting policies before comparing earnings. External models and chips can create switching costs and supplier exposure; flexibility and dependable delivery matter alongside technical benchmarks.
06Treat governance as operating capability
Errors, bias, outages and data incidents can become commercial costs. Examine testing, monitoring, escalation and customer responsibilities. NIST's voluntary framework is a reference for discussing risk, not proof of certification. Sensitive applications need use-case-specific review, including human oversight and audit trails. Separate documented commitments from verifiable implementation.
07Follow cash and dilution
Assess the funding required by hiring, equipment, delivery and development schedules. Do not extrapolate one quarter's cash burn mechanically. Review share-based compensation, potential dilution and acquisition funding. Technology adoption, recognised revenue, collected cash and shareholder value are connected but distinct. Industry opportunity does not remove offering-price or financing risk.
08Compare different AI revenues
Recurring software and one-off data projects require different tests. Subscription cancellation and promotional pricing matter; project firms may still improve efficiency through reusable tools. Supplier price cuts affect each model through customer repricing, staffing and contracts, not through the technology label alone.
09Maintain operating checkpoints
Monitor paid customer mix, renewal, acceptance periods, service margins and cash conversion using consistent definitions. More usage with lower pricing may not strengthen economics. Ask who captures savings from technical progress: suppliers, application companies or customers. SilverCape's focus is the connection between capability, demand, cost and governance, supported by evidence rather than a universal technology narrative.
