Phasecraft has assembled what the company describes as the largest existing collection of variational quantum eigensolver outputs, comprising more than 3,000 emulations across 13 distinct molecular systems. The computational work took place on Nvidia Hopper processors at the University of Nottingham, leveraging Nvidia's cuQuantum toolkit. According to Phasecraft's account, the effort achieved a 15-fold acceleration relative to previous approaches for simulating many-body systems.
The simulations encompassed quantum circuits ranging from four to 32 qubits, with concentration in the 24-to-28 qubit band. This range matters significantly: it keeps the entire undertaking within the bounds where conventional computers remain capable of mimicking quantum behaviour. Phasecraft conducted the work exclusively on classical hardware, producing a computational picture of what a quantum machine would generate without deploying any actual quantum processors—currently the only viable approach for generating datasets of this magnitude.
The strategy reflects practical necessity rather than deception. Quantum chemistry represents the most frequently cited application for quantum computing, yet functional quantum hardware suitable for such work does not yet exist. Assembling training datasets now, positioning quantum-enhanced density functional theory to function when suitable machines become available, constitutes a sound long-term approach rather than merely a temporary substitute.
Phasecraft has not cited any published paper or preprint substantiating the 15-fold speedup claim, nor has the company identified the specific benchmark it surpassed beyond referring to "prior results" in many-body modelling. This contrasts with the company's March announcement last year regarding the THRIFT algorithm, which achieved a 10-fold improvement and appeared in Nature Communications on the day of disclosure. The current announcement arrives as a company blog post co-authored by the chip supplier whose processors the work employed.
Ashley Montanaro, Phasecraft's chief executive, remarked that fulfilling quantum computing's potential demands "pushing the limits of today's most capable hardware." Sam Stanwyck, representing Nvidia, observed that "the work shows what happens when leading researchers get access to accelerated computing"—language characteristic of chipmaker communications.
Montanaro has previously occupied a different vantage point on these matters. In September 2024, he told The Next Web that Britain possessed an opportunity to establish itself as the home of the next quantum Nvidia, advocating for sustained funding and relaxed export restrictions. Two years forward, his firm's most significant achievement emerges through a joint announcement with Nvidia itself, executed on Nvidia processors, utilising Nvidia software. This arrangement involves no contradiction, yet it accurately reflects where value concentration is occurring within quantum computing during the hardware development phase.
The database strategy aligns with Phasecraft's established pattern. In January 2024, the company released a materials complexity database encompassing more than 40 materials alongside a novel algorithm. The approach remains consistent: establish the reference data infrastructure first, then position the organisation to capitalise when hardware becomes available.
Financial backing originates from Wellcome Leap's Q4Bio programme, which investigates whether emerging algorithms can deliver quantum advantage in healthcare applications. The ultimate objective centres on drug discovery, though the announcement contains no description of any discovered molecule, screened compound or outcome actionable by a pharmaceutical organisation.
Phasecraft characterises the dataset as providing quantum-enhanced molecular modelling with training material. For perspective, Britain's second commercial quantum computer—a Rigetti system activated in April 2024—operates at 32 qubits, matching the maximum qubit count in Phasecraft's classical emulations. The meaningful distinction lies not in qubit quantity but in error rates, and addressing that gap remains someone else's responsibility.
Source: The Next Web
