Algorithms over atoms: why quantum software is having its moment

• Phasecraft raised $34m to push quantum algorithms into real-world materials and optimisation use-cases; UK/Europe angles are getting sharper.

• Algorithmic gains (e.g., Gidney’s 20× qubit cut for Shor) keep lowering the hardware bar, but fault-tolerant million-qubit machines still aren’t here.

What happened

On 2 September 2025, UK quantum-algorithms start-up Phasecraft announced a $34m Series B co-led by Plural, Playground Global and Novo Holdings. The pitch: software advances that make today’s and next-gen hardware useful sooner, starting with hybrid quantum/classical workflows in materials, energy and networks.

Context & data

• Funding & partners: Phasecraft’s round brings total funding (incl. grants) to >$50m, with partnerships across Google, IBM, Quantinuum and QuEra; end-users include Johnson Matthey, Oxford PV, NESO and BT. (2 September 2025).

• UK policy tailwind: the National Quantum Strategy commits £2.5bn over 10 years from 2024; ambition includes accessible UK-based quantum systems by 2035. (20 March 2023).

• Market sizing: McKinsey pegs quantum-tech revenue potential at ~$97bn by 2035, with computing up to ~$72bn and biggest impact in chemicals, life sciences, finance and mobility. (23 June 2025).

• Algorithmic step-change: Google’s Craig Gidney shows RSA-2048 factoring could require <1m noisy qubits (vs ~20m in 2019) with <1 week runtime under stated assumptions—highlighting how software can re-shape hardware thresholds. (5 June 2025).

My take

I like the shift in spend from cryostats to code. Hardware roadmaps are lumpy; software has faster iteration cycles and, critically, expands the useful qubit budget via error-aware algorithms and hybrid pipelines. Phasecraft’s focus on materials (DFT-adjacent hybrids) fits where early “scientific advantage” is most plausible: small-system chemistry and solid-state models where even modest accuracy gains improve R&D ROI. The enterprise wedge is clear: co-develop workflows with marquee industrials, rent access via cloud hardware partners, and monetise as tools + services (ARR) before full platform plays.

But I stay valuation-disciplined. Until error-corrected hardware arrives, revenues look like pilots, co-funded research and optimisation modules—good logos, limited ARR depth. The defensibility question is whether Phasecraft’s IP (compilers, resource-frugal ansätze, error-aware transforms) compounds into a moat across hardware stacks. Cross-platform partnerships help here; so does publishing state-of-the-art (to attract talent) while productising the non-obvious bits.

Risks & watch-list

• Hardware timeline risk: Million-qubit, fault-tolerant systems remain uncertain; if timelines slip, “scientific advantage” may not translate to commercial advantage quickly.

• Hype/expectations gap: Bold timelines invite scrutiny; I’d track concrete KPIs (paid pilots to ARR conversion, gross margin on software modules, payback vs classical baselines).

• Competitive set: Global algorithm shops (and hyperscalers’ in-house teams) can fast-follow; need evidence of superior resource estimates translating to cheaper or more accurate results on real instances.

• Policy & funding cyclicality: UK public funding is a tailwind but not guaranteed through cycles; watch procurement routes and any consolidation moves.