NVIDIA Unveils AI Future Amid Stock Puzzle and Quantum Leap

As of last week, NVIDIA (NVDA) experienced a dip of 0.70%, marking a weekly decline of 3.26% and a year-to-date drop of 12.35%. The company's market capitalization stands at approximately $2,871.88 billion.

During NVIDIA's annual GTC conference in San Jose, CEO Jensen Huang addressed concerns regarding competition from DeepSeek's AI models in China, increasing pressure from chip manufacturers like Broadcom, and potential U.S. tariffs on chip imports. Huang argued that DeepSeek's advanced models have heightened demand for NVIDIA's GPUs due to their hefty computational needs. He also cast doubt on Broadcom's AI-specific chip market feasibility, noting that many such projects falter before reaching the market. Regarding tariffs, Huang emphasized NVIDIA's adaptable global supply chain and plans for boosting U.S.-based manufacturing.

The conference showcased NVIDIA's robust AI vision and promising product roadmaps, aiming to outperform rivals in the coming years. Attendees actively explored innovations in AI, robotics, and data center technologies amid an enthusiastic atmosphere. Continuous GPU demand in data centers was highlighted, with the number of GPUs in large "superclusters" climbing from 16,000 to over 100,000 within a year. By 2027, clusters with a million GPUs are anticipated. NVIDIA sees substantial potential in AI robotics, predicting widespread industrial applications for humanoid robots within five years.

The highlight of the event was the unveiling of NVIDIA's next-gen AI servers. The Blackwell Ultra, set for a 2024 launch, promises 50% faster speeds compared to current models. The Rubin AI server (2026) and Rubin Ultra (2027) will enhance performance over threefold and fourteenfold, respectively. Despite these advancements, NVIDIA's stock showed minimal movement, puzzling many. A recent report suggests the tech industry recognizes these impacts, with Wall Street potentially needing more time to catch up.

NVIDIA plans to launch a quantum computing research facility in Boston later this year, combining leading quantum hardware with AI supercomputers for accelerated quantum computing. Entities like Quantinuum and Quantum Machines will collaborate with academic researchers using NVIDIA's CUDA-Q platform to tackle integration challenges between GPUs and quantum processing units.

Additionally, NVIDIA recently acquired Gretel, a synthetic data company, as part of its ambition for an "AI full stack" vision. This strategic move enhances NVIDIA's capabilities by supplying high-quality training data for generative AI models and applications. Gretel's technology will become part of NVIDIA's cloud services, advancing its comprehensive ecosystem from hardware to AI application development.

Morgan Stanley's semiconductor industry report remains optimistic about NVIDIA, highlighting improvements in inventory levels for end customers and continued expectations for progress in inventory management across the sector. They favor NVIDIA due to its promising product cycles and AI-driven growth.