Giza

Thesis

Giza builds ZK-ML infrastructure on StarkNet, which gives it a natural advantage from StarkNet's native STARK proof system. Purpose-built for zero-knowledge computations. While EZKL and Modulus Labs build more generalized ZK-ML libraries, Giza's deep integration with StarkNet means it can use the chain's proof infrastructure directly, potentially offering better performance for on-chain AI verification. The seed opportunities analysis groups Giza with EZKL and notes both have 'working codebases with ONNX model support,' meaning standard ML models can be verified without custom reimplementation. The Giza pre-seed funding of approximately $3M is the smallest among the three ZK-ML leaders, but the StarkNet system provides distribution advantages that raw capital cannot buy. Like its ZK-ML peers, Giza has no token and no announced tokenization timeline. Capital deployment requires either angel investment in the company or buying STRK (StarkNet token) as system exposure. The 1000x Opportunity Map suggests this latter approach as a proxy strategy. The key risk is that Giza's StarkNet-specific approach could limit its reach to a single system, while EZKL's chain-agnostic approach could win broader adoption. The thesis hinges on whether ZK-ML becomes essential infrastructure for on-chain AI or remains a technical curiosity. If the autonomous AI agent economy scales to even a fraction of Gartner's projected $30T by 2030, verifiable inference becomes non-negotiable for high-value transactions. Giza's position within StarkNet. Itself backed by some of the strongest cryptographic talent in blockchain. Gives it a technical credibility advantage, but commercial traction remains unproven.

Catalysts

• +StarkNet system growth provides built-in distribution for ZK-ML tooling
• +EU AI Act enforcement (eIDAS 2.0 June 2026) creating demand for verifiable AI outputs
• +Potential token launch or StarkNet system airdrop could provide first investable moment

Risks

• -StarkNet-specific approach may limit reach compared to chain-agnostic competitors (EZKL)
• -Smallest funding (~$3M pre-seed) among ZK-ML leaders; highest execution risk
• -No token, no clear tokenization timeline; only accessible through private investment or STRK system proxy

Common questions