The immutability, decentralization, and scalability potential of sharded blockchain make it an ideal solution for addressing trust and high-processing demands in financial data sharing. Existing sharding architectures, however, struggle to balance high throughput with low latency under heterogeneous transmission conditions and dynamic financial transaction loads in financial networks. To address this challenge, we propose a heterogeneous-aware adaptive load-balancing sharding blockchain for financial data sharing, featuring an innovative dual-dimensional architecture design. First, we partition nodes into heterogeneous shards while establishing an adaptive matching mechanism between shard performance and load intensity, thereby breaking through the throughput bottleneck of shards. Second, we develop a parameter adjustment method based on spatiotemporal distribution characteristics of financial transactions, continuously optimizing shard operational parameters within epoch cycles to achieve lower average transaction latency. Simulation experiments verified that our solution significantly outperforms mainstream sharding blockchain schemes in both throughput and latency metrics.