NVIDIA is pushing its GPUs deeper into scientific computing with Daviator, a new software stack aimed at removing a costly bottleneck in plane-wave density functional theory, a workhorse method in materials science and chemistry.
NVIDIA Daviator Speeds Plane-Wave DFT Workflows

That matters because the company is not just selling faster chips anymore; it is trying to own more of the workflow that creates the next generation of AI materials, batteries and advanced semiconductors. Plane-wave DFT is one of the most compute-hungry steps in training machine learning interatomic potentials, and NVIDIA said the new tools can slash execution times by more than 85% on mid-size systems and deliver more than 2x speedups on simple cases, with some higher-accuracy calculations running up to 22 times faster.

The market’s bigger takeaway is that NVIDIA is attacking the underutilization problem that has long limited GPU adoption in traditional high-performance computing. DFT codes have been built around double-precision workloads and sequential processing, while modern NVIDIA hardware is increasingly optimized for mixed-precision throughput and batched execution. Daviator changes that by grouping k-points, cutting kernel launches and keeping data resident on the GPU, which is exactly the kind of software-level optimization that can turn idle silicon into billable compute.
For investors, this is the sort of quiet infrastructure expansion that extends the AI capex cycle beyond model training and into scientific discovery. If NVIDIA can make expensive research workloads run materially faster on its stack, it strengthens the case for GPUs not just in hyperscale data centers but in labs, national research centers and industrial R&D pipelines where simulation cost and turnaround time determine competitiveness. The opportunity is especially attractive because these workloads can create recurring demand for premium hardware, software and networking, while also reinforcing NVIDIA’s ecosystem lock-in.

The timing also fits the broader AI hardware trade. NVIDIA stock has been under pressure, with the shares down to 149.46 on Nov. 19 from 221.97 on Oct. 21, and the 50-day moving average now sitting far above the price at 208.93, a sign the market has been punishing the name hard. Adalytica’s NVIDIA earnings sentiment gauge shows fear at 18, even as awareness remains extreme at 89, a combination that often marks a stock the crowd is watching but not yet comfortable owning. That disconnect is where asymmetric setups tend to appear.
The strategic narrative is simple: NVIDIA is moving from a pure GPU supplier to the operating system of accelerated computing. Daviator, alongside its mixed-precision execution layer, shows how the company can keep extracting value from its installed base even as chip cycles mature. If it can make previously inefficient scientific workloads economically attractive on its hardware, the upside extends far beyond gaming and AI training.
Investors should treat this as a picks-and-shovels expansion play on the next wave of compute demand. The beneficiary is NVIDIA, but the ripple effect could also lift the broader AI infrastructure stack, from advanced packaging and memory to networking and high-performance storage. The losers are legacy CPU-centric workflows and any rival platform that cannot match NVIDIA’s software advantage in scientific computing.
| Entity | Gains | Losses |
|---|---|---|
| NVIDIA | ▲Higher GPU utilization; wider HPC market | ▼Legacy CPU-centric compute vendors |
| AI/Materials R&D labs | ▲Faster DFT and MLIP training | ▼Slow, expensive simulation workflows |
| Semiconductor ecosystem | ▲More demand for advanced GPUs and networking | ▼Platforms lacking mixed-precision support |
| Short sellers in NVDA | ▲Volatility to trade | ▼Software-led moat expanding |



