Nvidia posted nine items to its newsroom at 16:00 UTC on 16 November, the day before it reports results. They were led by a CUDA-Q quantum decoding algorithm it says is up to 13,000x faster, and a Vera CPU rack it says delivered 2.4x the performance per watt of x86 systems in one test. The slate also added Nemotron for Science and a Rubin HPC GPU, which NVIDIA’s blog says was announced at SC26.
Our number: NVIDIA’s own worked example, one computation cut from 12 weeks to 45 minutes, is a 2,688x speedup, by our arithmetic. The 13,000x figure is the release’s peak, about 4.8 times that example.
Business Pill · GPUs explained
An evergreen Business Engineer explainer on GPUs. It teaches the general idea only and does not discuss Nvidia’s 16 November announcements.
The key insight: As we read the releases, Nvidia used the day before results to set out a pipeline that runs from software available today to a GPU in 2028, and its own worked quantum example, 12 weeks to 45 minutes, is a 2,688x speedup by our arithmetic, against the up-to-13,000x headline figure.
What Nvidia Announced on Vera
The Vera CPU release says the rack-scale platform combines 192 Vera CPUs with NVIDIA Quantum-X800 InfiniBand networking, about 300TB of memory and 1 petabyte of storage in a liquid-cooled design. Each Vera CPU has 88 custom NVIDIA Olympus cores, so one rack holds 16,896 cores, by our arithmetic.
In NVIDIA’s tests, Vera delivered 1.9x the performance and 2.4x the performance per watt of x86 systems on OpenFOAM’s motorBike Large airflow simulation, the release says. That pairing implies the Vera system drew about 79% of the x86 systems’ power in that test, or about 21% less, by our arithmetic.
NVIDIA says Vera CPU racks are expected from Dell Technologies, HPE, Lenovo and Supermicro beginning in early 2027. It also says DKRZ, the German Climate Computing Center, has selected Vera CPUs for its HLRE-5 supercomputer, which is built by Bull and expected to be available in early 2028. We covered the agent side of the chip in our July analysis of Vera as an agentic CPU.

The Quantum Number, Checked
The quantum release says NVIDIA’s new TCBP-JOBS decoding algorithm, available in the CUDA-Q QEC 0.9.0 library today, “accelerates the computationally intensive parts of quantum error-correction research by up to 13,000x while delivering high accuracy”. It also says the algorithm cuts “the computation time for researching new quantum computing designs from 12 weeks to just 45 minutes”.
Twelve weeks is 120,960 minutes, so that example is a 2,688x speedup, by our arithmetic. The release gives both figures and names the comparison as “existing leading algorithms”; it does not say which workload produced the 13,000x peak.
The same release announces a modular quantum-classical codesign platform with IBM and Fermilab. IBM and NVIDIA are colocating NVIDIA DGX B200 systems with IBM quantum computers, including those powered by IBM Quantum Nighthawk, which the release calls “the fastest quantum processor IBM has ever delivered”. NVIDIA is also expanding the open source Slurm workload manager so centers can schedule quantum processors alongside CPUs and GPUs, with IBM contributing its Quantum Resource Management Interface.
IBM and RIKEN modeled protein-ligand complexes containing up to 12,635 atoms, more than 40x the previous 303-atom quantum-classical benchmark, the release says; the ratio is 41.7x, by our arithmetic. The work is a finalist for the 2026 ACM Gordon Bell Prize.

Nemotron for Science and Rubin HPC
The Nemotron for Science release says the package combines the Nemotron 3.5 Super VL open model, LitmusBench v0.2, an open chemistry dataset with nearly 50,000 questions, and post-training recipes. Both the model and the dataset are available now on Hugging Face, the release says.
Lila Sciences, Terray Therapeutics and Therna are using and customizing the model, per the release. It says that at Lila, post-training with laboratory feedback took the model’s accuracy for making quantum dots to 61%, against GPT-6 Astra’s 54%, “at 10-20x lower inference cost”.
Separately, an NVIDIA blog post says Rubin HPC is a new GPU in the Vera Rubin platform with a record number of CUDA cores, “more than 50,000 on a single GPU”, and that it is expected to be available in 2028.
The Oracle Expansion Behind It
NVIDIA Japan’s X account highlighted on 16 November the Oracle expansion that NVIDIA and Oracle announced on 27 October. That release sets out plans to take OCI Superclusters with Vera Rubin NVL72 systems to multi-gigawatt scale, and says Vera Rubin NVL72 and Vera CPUs are generally available on OCI.
The 27 October release quotes OpenAI’s Uday Ruddarraju, chief technology officer of compute: “We are bringing Vera into production on OCI to support these workloads at scale”. Physical builds for Vera Rubin are already drawing suppliers, as in Daikin’s 60 MW cooling design.
What the Slate Adds Before Tuesday
Nvidia reports third-quarter fiscal 2027 results after the US close on Tuesday 17 November, with a consensus of $2.49 a share, as our earnings preview set out; our valuation comparison puts that against Broadcom and AMD. The announcements came four days after the board named CFO Colette Kress interim CEO in NVIDIA’s 12 November release.
The NVIDIA executives quoted in the three 16 November press releases are Ian Buck, vice president of hyperscale and high-performance computing, and Krysta Svore, vice president of applied research for quantum computing. By our count, the three releases name 15 outside organisations as partners, buyers or users, from Dell and HPE to IBM, Fermilab, Argonne and RIKEN. More on the company is in our Nvidia hub.
The Structural Read
As we read the Vera release, the CPU is offered as a rack rather than a chip: 192 CPUs, InfiniBand networking and reused NVL72 rack components, from four server makers.
As we read the quantum release, NVIDIA positions its GPUs and CUDA-Q software as the classical side of quantum computing, with IBM supplying the quantum computers and Fermilab the error-correction and physics research.
Ian Buck, NVIDIA
“Scientific simulations need CPUs that can keep large datasets moving and sustain performance as applications scale.”
Three Implications
FOR INVESTORS The releases give no prices, unit volumes or order values; NVIDIA says the Vera racks come from early 2027 and Rubin HPC in 2028.
FOR RESEARCH BUYERS NVIDIA says CUDA-Q QEC 0.9.0, Nemotron 3.5 Super VL, LitmusBench v0.2 and HPC SDK 26.11 are available now.
RELATED Our earnings preview sets out the $2.49 consensus for Tuesday’s results.
The Business Engineer Lens
This story maps onto the Business Engineer framework The AI Supercycle.
The framework argues: “Three competitive categories within one layer: generalist substrate (NVIDIA), hyperscaler custom (TPU, Trainium, MTIA), merchant custom challengers (Cerebras, Groq, SambaNova).”
As we read it, the 16 November releases stretch NVIDIA’s generalist position past the GPU, into CPU racks, quantum decoding software and open science models. The framework predates the releases and does not discuss them.
What Is Not Established
The 13,000x, 2.4x and 1.9x figures are NVIDIA’s own test results. The Vera figures come from a single OpenFOAM workload, and NVIDIA describes its Weather Research and Forecasting gains of 1.6x in performance and 2x in energy efficiency as projected.
The releases give no prices, unit volumes or order values. The Vera racks and Rubin HPC are not yet available, by the dates NVIDIA gives. We did not contact NVIDIA, Oracle or IBM for this piece.
The Bottom Line
Nvidia posted nine newsroom items the day before results, led by a CUDA-Q decoder it says is up to 13,000x faster; its own 12-weeks-to-45-minutes example works out to 2,688x, by our arithmetic. The rest runs from Vera CPU racks from four server makers in early 2027 to a Rubin HPC GPU in 2028.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. The Vera CPU, quantum and Nemotron for Science details are from NVIDIA’s three press releases of 16 November 2026, and the Rubin HPC details from NVIDIA’s blog post of the same day. The Oracle expansion is from the NVIDIA and Oracle release of 27 October 2026, which NVIDIA Japan’s X account highlighted on 16 November. The leadership change is from NVIDIA’s 12 November release.
All performance figures are NVIDIA’s own test claims. The 2,688x ratio, the 4.8 multiple, the core count, the implied power share, the 41.7x ratio and the count of items and organisations are our arithmetic. NVIDIA, Oracle and IBM were not contacted for this piece.
Sources: NVIDIA, Vera CPU platform release, 16 Nov 2026 · NVIDIA, IBM and Fermilab quantum release, 16 Nov 2026 · NVIDIA, Nemotron for Science release, 16 Nov 2026 · NVIDIA blog, Rubin HPC, 16 Nov 2026 · NVIDIA and Oracle release, 27 Oct 2026 · NVIDIA, CEO transition release, 12 Nov 2026









