Google vs IBM: Two Frontiers, Two Futures
This week, my MIT assignment asked us to revisit three landmark quantum‑simulation papers from a decade ago and decide which “horse” we would back today (link to download my assignment in pdf). The debate among my cohort quickly crystallised into a familiar polarity: Google vs IBM. I chose IBM, not because Google’s Scalable Quantum Simulation of Molecular Energies is anything less than elegant, but because the two companies are optimising for fundamentally different frontiers.
Google’s approach is algorithmically purist: chemically structured ansätze, physically faithful mappings and circuits that mirror the intrinsic symmetries of molecular systems. For large‑scale quantum chemistry, this is a powerful long‑term strategy. But having seen IBM’s Network Two 156‑qubit machines at VivaTech Paris a couple of days ago and photographed standing next to them, it is clear that IBM is pursuing a different kind of ambition – one that is architectural rather than purely algorithmic. Network Two pushes coherence, connectivity, calibration stability, and system‑level orchestration in ways that directly benefit variational, noise‑aware workloads such as VQE.
As a former quant fund manager who pushed hundreds of millions through electronic markets every month, this distinction matters. Finance is not chemistry. Our problems are high‑dimensional, noisy, path‑dependent, and structurally irregular. The hybrid philosophy demonstrated in the 2016 VQE paper — short circuits, hardware‑efficient ansätze, real‑device sampling – maps naturally onto the quantum feature‑generation pipeline behind the HSBC experiment. Network Two’s lower error rates, faster sampling, and more stable calibration cycles make it a better fit for these near‑term, data‑driven tasks.
What impressed me most was not the qubit count but the system engineering: cryogenic routing, control electronics, crosstalk suppression, and the uniformity of qubit performance across the chip. These are the unglamorous but essential ingredients that determine whether a quantum computer can run deep variational circuits without collapsing into noise. Google’s approach is algorithmically beautiful; IBM’s is industrially scalable.
So would I invest? I would not frame it as “Google vs IBM.” I would invest in the modality aligned with the use case. For chemistry, Google’s structured ansatz approach may scale further. For finance, optimisation and quantum‑enhanced machine learning, IBM’s Network Two architecture is the more commercially relevant path — not because it “beats” Google, but because it is built for the noisy, hybrid, real‑time workloads where quantum advantage will first emerge. Yes, I would back this horse because I work in finance.
