NVIDIA stands at a critical juncture: while most analysts maintain a positive outlook on its fundamentals, its stock performance has been lackluster year-to-date, with market concerns yet to dissipate. The annual GTC conference, commencing this week, offers the AI chip giant a prime opportunity to reignite investor confidence.

Investors are primarily focused on the company's strategic positioning in inference chips, the latest updates to its chip roadmap, and the supply outlook for wafers, memory, and optical components. These details will significantly influence market perceptions of NVIDIA's ability to maintain its leadership amid the ongoing wave of AI expenditure.

Bank of America suggests that concerns over the sustainability of capital expenditure from hyperscale cloud providers, coupled with geopolitical uncertainties, are core reasons for the persistent pressure on NVIDIA's valuation. Even if the sales outlook presented at GTC aligns broadly with consensus expectations, providing specific details on supply and demand visibility could be sufficient to support a near-term stock price recovery.

However, UBS, while bullish on NVIDIA's fundamentals, notes that it is "difficult to foresee" commentary from the conference triggering a stock surge significant enough to "alter the investment thesis." Nevertheless, the bank acknowledges that the disparity between NVIDIA's robust earnings expectations and its discounted valuation appears "unsustainable."

Behind the subdued stock price lies a gap between its "normalized" valuation and growth expectations. Following a remarkable 22,000% cumulative gain over the past decade, NVIDIA's stock has stalled this year, leading some investors to question whether its next growth phase can match previous momentum.

From a valuation perspective, NVIDIA currently trades at 17 times next fiscal year's estimated earnings, below the valuation level of the S&P 500 index. According to FactSet data, among the 70 analysts covering NVIDIA, 93% recommend a "Buy" rating, with an average price target of approximately $267.

Free cash flow projections indicate that NVIDIA's free cash flow for the current fiscal year (ending January 2027) is expected to grow by 85% year-over-year to over $178 billion, a significant acceleration compared to the previous fiscal year. If realized, this would set a new historical record for corporate free cash flow globally. Projections for the following fiscal year suggest this figure could rise further to $233 billion. However, the wide range between the high and low ends of current forecasts—a difference of $98 billion—reflects substantial remaining uncertainty regarding NVIDIA's future profitability.

A core focus of GTC is NVIDIA's push into the AI inference market. The most anticipated aspect of the GTC conference is NVIDIA's strategic shift from focusing primarily on training to inference, alongside adjustments to its supply chain布局. The company may announce a major entry into the AI inference market by integrating Groq's technology; on the manufacturing side, it might introduce Samsung as a foundry partner for the first time, breaking TSMC's monopoly, with OpenAI potentially being among the first major customers.

Bank of America Securities anticipates that NVIDIA will showcase a new generation of products during its GTC keynote. Key highlights are expected to include rack systems integrating Groq's Language Processing Unit (LPU) technology, a new generation of high-speed switches, Co-Packaged Optics (CPO) solutions, and the potential for a custom x86 processor developed in collaboration with Intel.

In recent years, AI chip spending has been concentrated on training scenarios, a demand heavily reliant on the massive parallel computing capabilities of GPUs. This explains how NVIDIA, leveraging GPU technology originally developed for gaming graphics rendering, achieved dominance in the AI infrastructure market. However, decoding scenarios within the inference process create demand for non-GPU specialized chips. To address this gap, NVIDIA acquired private company Groq's LPU technology and related talent last year for approximately $20 billion. Groq specializes in designing dedicated chips for inference decoding tasks.

It is widely expected that Jensen Huang will elaborate at GTC on how NVIDIA plans to integrate LPUs into its chip portfolio to comprehensively cover the three major application scenarios: training, pre-fill, and decoding, while also countering competitive pressure from hyperscale cloud providers developing their own chips.

NVIDIA's earnings prospects depend heavily on continued increases in AI infrastructure investment from hyperscale cloud providers like Amazon and Microsoft. Taking Amazon as an example, its capital expenditure remained in the $50-$60 billion range annually from 2021 to 2023, but is projected to jump to $190 billion this year, with the increase almost entirely directed towards AI infrastructure build-out, including significant purchases of NVIDIA chips.

Barclays predicted last week that overall industry capital expenditure in AI will peak at around $1 trillion in 2028 before experiencing a "moderate decline." The bank also believes that the consensus estimate for hyperscale cloud provider capex in 2028 is underestimated by approximately $300 billion.

Wall Street remains divided on how the market will react to GTC. Bank of America attributes the ongoing pressure on NVIDIA's valuation primarily to doubts about the sustainability of hyperscale cloud capex and geopolitical uncertainty. It suggests that even a sales outlook matching consensus, if accompanied by concrete supply/demand details, could fuel a short-term recovery. Conversely, UBS finds it hard to envision GTC commentary triggering a paradigm-shifting stock surge, though it acknowledges the seemingly unsustainable tension between strong earnings forecasts and a discounted valuation.

In other words, the market does not necessarily distrust NVIDIA's fundamentals, but is awaiting a clear signal capable of dispelling doubts and reshaping expectations. Jensen Huang's speech this week represents the most significant near-term opportunity to provide such a signal.