BNP Paribas is bullish on Nvidia (NVDA.US): Strong momentum continues at Berlin GTC conference, AI factories accelerate gigawatt-level deployment

According to the Zhihui Finance APP, NVIDIA (NVDA.US) held its Berlin GTC conference earlier this week, and BNP Paribas was deeply impressed by the company's continued operational scale and speed after attending the event.

BNP Paribas analyst Karl Ackerman wrote in a client report: “We took away several key points from the Berlin GTC conference: 1) NVIDIA continues to optimize at the AI factory level to accelerate gigawatt-scale deployments; 2) NVIDIA's CPO (co-packaged optics) rollout plan for horizontal and vertical scaling networks is still scheduled to be implemented in the second half of 2026 and the second half of 2027, respectively; 3) As data center deployments spread across Europe and Asia, NVIDIA is making significant investments in the AI supply chain. We remain bullish on NVIDIA, view it as the undisputed leader in AI infrastructure, and see its expansion into turnkey AI factory solutions, secure autonomous agents, and sovereign AI construction.”

Ackerman gave NVIDIA an “outperform” rating with a target price of $345.

At the conference, NVIDIA showcased several upcoming products, including Vera Rubin Ultra NVL576, Vera CMX, Groq 3 LPX, and Spectrum-X Photonics Switches.

In addition, Ackerman said he saw the NVL72 rack in person, with eight NVL72 racks optically connected to form a 576-GPU vertical scaling cluster. “We believe each NVL72 system may be equipped with 648 3.2T optical engines, which means there are 9 optical engines per GPU or per switch ASIC,” Ackerman added. “Assuming an external laser source module provides light for every two optical engines, this means each NVL72 is equipped with 324 ELS modules—i.e., each NVL72 system has 2,592 ultra-high-power continuous-wave lasers.”

Another notable release is the DSX PODs, a Vera Rubin AI factory reference architecture that can reduce GPU installation time from six months to 14 days. NVIDIA is currently working with 17 partners and plans to deploy DSX PODs before the first quarter of 2027.

Other important announcements include: launch of OpenShell reference architecture; plans to increase substrate capacity by 2.5 times over the next three years with partners including IBIDEN, Samsung and Unimicron; agreement reached with Amkor; and installation of Google A5X instances based on Vera Rubin NVL72 in Europe.

“Google plans to deploy up to 960,000 Rubin GPUs, leveraging NVIDIA CX-9 SuperNIC, and connect to the TPUv8t supercluster via Google’s Virgo network,” Ackerman added. “We believe 960,000 GPUs translates to 13,300 NVL72 racks, or about 2.9 gigawatts of AI computing capacity (excluding horizontal scaling networks).”