Amidst the flash and fanfare of CES 2025, Jensen Huang, the charismatic CEO of NVIDIA, made a grand entrance clad in a dazzling black crocodile leather jacket that immediately captured attentionOn the evening of January 6, he took center stage for an impressive 90-minute keynote presentation, during which he unveiled a lineup of innovative technologies that spanned graphics cards, world models, autonomous driving, and roboticsThe atmosphere was electric, fueled by high expectations from the market, especially with NVIDIA's stock price hitting an all-time high just before his talk, reaching a market capitalization of $3.66 trillion.

However, the sparkle of Huang's jacket was overshadowed by the stark reality of the company's post-presentation stock performanceOn January 7, NVIDIA saw its stock price plummet by 6.22%, closing at $140.14. This drop not only erased Monday’s impressive gain of 3.4% but also marked the largest single-day decline over the past four monthsAnalysts quickly began to dissect Huang's presentation, suggesting that his remarks lacked sufficient detail on NVIDIA's most profitable business—chips used for training AI models.

During the presentation, Huang did mention the company’s Blackwell AI processors being fully in production, yet he notably withheld detailed updates on the next-generation GPU platform, RubinThis omission contributed to market sentiments cooling off significantlyHuang’s remarks were ambitious, as he projected transformative applications for AI within businesses, but without specific metrics to back up the potential growth, investor confidence wavered.

Embodying a playful spirit, Huang began his address with a light-hearted exchange, asking attendees, “Do you like my jacket?” As he became more animated, he humorously wielded a NVLink 72 display piece to mimic Captain America’s shield, capturing the audience's imagination

Advertisements

This theatrical flair set the tone for a presentation that would showcase the latest in gaming technology, notably the eagerly awaited RTX 50 series graphics cards, specifically the RTX 5090.

The RTX 5090, designed primarily for gaming—historically a significant revenue stream for NVIDIA—was introduced with great fanfareHuang highlighted the card's impressive specifications, boasting 92 billion transistors and offering around 4000 TOPS of computational power, which nearly doubles the performance compared to its predecessor, the RTX 4090. The pricing for this high-end gaming card started at a hefty $1,999, with availability slated for January, while laptops featuring the RTX 5090 were set to hit the market at $2,899 beginning in March.

Yet, the gaming graphics card launch was merely a teaserIn the wake of a generative AI explosion, NVIDIA's data center division had come to dominate the company's revenue contributions, now accounting for an impressive 90%. Unsurprisingly, market observers were primarily focused on the progress of NVIDIA's AI chip developmentsHuang responded to this trend by introducing NVIDIA's version of AI Agents, dubbed "Blueprint for AI Agents," which allows developers to create and analyze video content utilizing visual perception capabilities, operating at speeds 30 times faster than traditional monitoring methods.

Huang expressed optimism that every corporate IT department will soon transform into a human resources department for AI Agents, possibly unlocking over a trillion dollars worth of opportunitiesAmid ongoing speculation about NVIDIA branching into the PC market with AI-driven processors, Huang confirmed that the company is actively exploring ways to integrate AI technology into personal computers, specifically starting with devices running Windows 11 through the Windows Subsystem for Linux.

As generative AI applications surged, Huang emphasized the significance of Physical AI as a cornerstone of NVIDIA's vision

Advertisements

He unveiled “NVIDIA Cosmos,” a foundational model specifically designed for robotics and autonomous driving technologiesThis efficient training tool could generate highly realistic and physics-compliant videos, fundamentally serving as the bedrock technology for both self-driving vehicles and robots.

Much like large language models power applications such as ChatGPT, Cosmos stands poised to provide foundational capabilities for smart driving and robotics, facilitating the creation of synthetic data for training and testing without reliance on costly real-world data collectionHuang illustrated the transformative potential of Cosmos, envisioning a similar impact on the robotics and industrial AI sectors as seen with Meta's Llama3 model in the enterprise AI landscapeHowever, he acknowledged the current necessity for further data and optimization to elevate Cosmos to a level of maturity comparable to more established models like ChatGPT.

Highlighting the practical applications of Cosmos, Huang pointed to both the burgeoning fields of autonomous vehicles and AI-driven robotics, emphasizing their need for systems that comprehend the physical world and translate data into machine-readable formats that produce visual outputsNVIDIA foresees substantial growth in the humanoid robotics market over the next two decades, estimating it could reach $38 billion, while collaborating with industry leaders, including some prominent companies in China.

Concluding with a dramatic "One More Thing" moment, Huang revealed Project Digits—a handheld AI supercomputer priced at $3,000. Slightly larger than Apple's Mac Mini, this device boasts computational power comparable to that of a data center, supporting models with 200 billion parameters, effectively turning personal computers into mini data centersHuang articulately stated, "This is an AI supercomputer, running the complete NVIDIA AI technology stack," suggesting its significant potential for education and research.

With flagship graphics cards, super PCs, and foundational models encapsulated within the diverse technological landscape Huang covered, one might wonder why NVIDIA's stock took such a downturn post-lecture

Advertisements

Some analysts speculated that broader market trends, such as the decline of major indexes like the NASDAQ and S&P 500, contributed to this shift where tech giants saw collective lossesHowever, many agree the primary cause of NVIDIA's decline lay more within the company's specificsInvestors were left wanting substantial insights on the progress of Blackwell expansion and the development of the Rubin GPU platform.

Chris Murphy, co-head of derivatives strategy at Susquehanna International Group, indicated that options trading was unusually active surrounding NVIDIA, with around 200,000 call options set to expire, hinting at market caution amidst uncertaintyWhile optimism regarding NVIDIA's long-term potential was prevalent following the keynotes, short-term gains appeared inadequate to satisfy investors’ expectations.

Huang's ambitious forecasts painted a picture of future applications across one billion humanoid robots, ten million automated factories, and over one-and-a-half billion autonomous vehiclesHe confidently declared partnerships with leading automakers, like Toyota, utilizing NVIDIA's Drive chips and software, projecting the automotive sector's AI business would expand to $5 billion by the fiscal year 2026.

Despite the long-term aspirations laid out, immediate returns on investments in robotics and automotive sectors did not align with NVIDIA's lofty valuationAs the industry anticipated a slowdown in important scaling laws affecting data and power, a reassessment of NVIDIA's true worth looms on the horizonIndustry voices, including OpenAI co-founder Ilya Sutskever, have echoed concerns regarding the diminishing returns of investing in large-scale language models and the growing complexities faced by AI labs.

In this whirlwind of predictions and projections, NVIDIA’s lofty market cap has surged over the past two years due, in no small part, to sky-high growth expectations for the AI sector

Advertisements

Advertisements