AI Oracle Patent Application

Following its announcement of 24 October 2024 regarding Method C AI Oracle, QBT, the AIM-listed investment company focused on an R&D and investment programme within blockchain technology, is pleased to confirm the filing of a new patent application covering Method C’s disruptive AI Oracle (“AI Oracle”): “IMPLEMENTATION OF BINARY DECISION TREES”.

The patent application describes the details of an extremely efficient hardware implementation of the AI Oracle on an ASIC chip.

A field-programmable gate array (“FPGA”) version of the AI Oracle implementation has been developed by the Company and the corresponding PPA (power performance area) results offer an insight as regards the relevant overheads of the solution when implemented on custom silicon used for Bitcoin mining.

The Company has determined that the overall area required by the AI Oracle implementation is between ~1% - 4% of a double SHA-256 lane, the basic computational block of almost all ASIC chips for bitcoin mining, depending on the logic gate manufacturing technology used. The same percentage applies also to energy consumption, i.e., the energy costs to run the AI Oracle represents an additional ~1% - 4% of the energy consumed by the miner performed by an ASIC without the AI Oracle.

The above figures represent an almost irrelevant overhead when compared to the double SHA-256 percentage of avoided computations, determined by the AI Oracle.

As QBT owns the Intellectual Property (“IP”) for the AI Oracle implementation, i.e., the logic gates architecture for the ASIC design, the Company will now look to implement a relevant commercial strategy, using its new IP, to monetise this proprietary asset.

Francesco Gardin, CEO and Executive Chairman of QBT, commented, “The key requirements included in the specification of the on-chip implementation of the AI Oracle, are the limited area to be used, the low energy consumption and the processing speed. We have determined that the AI Oracle will not affect the performance of any given SHA-256 architecture of an ASIC chip. All three requirements have been met by our FPGA/ASIC design team. This has been an outstanding result for the team and QBT. Moreover, the AI Oracle implementation overhead between approximately 1% and 4% of a double SHA-256 lane, is an extremely impressive result.

“While the patent application is meant to protect these innovative applications of the AI Oracle implementation, the core technology for the generation of the Oracle, i.e., parameters to build the AI Oracle generated by the Machine Learning Model C have been kept separate as these represent an asset for the Company which it intends to protect as an industrial secret.”