산업 분석에 따르면 오픈AI가 자체 설계한 AI 칩을 개발함에 따라 엔비디아의 하드웨어 부족 현상은 5년 이내에 해소될 것으로 보이며, 이는 현재 가속기 공급 계층 구조를 교란시킬 가능성이 있다.

Two days after reporting the largest quarter in semiconductor history, macro investor Jordi Visser told listeners on The Pomp Podcast that the scarcity sustaining NVIDIA (NASDAQ:NVDA) carries an expiration date. His argument distinguishes near-term performance from terminal valuation, anchored by a single observation: OpenAI’s newly developed Jalapeno silicon—described as “a chip designed by OpenAI, by AI”—will not disrupt NVIDIA’s financials this year or next. However, Visser cautioned that “Nvidia right now has scarcity, but at some point, five years from now, six years from now, it won’t have scarcity.” With NVIDIA’s market capitalization standing at $5.31 trillion, which reflects discounted future cash flows, any market reassessment suggesting GPUs become optional within a decade would trigger a contraction in that valuation.

NVIDIA’s second-quarter fiscal 2027 results underscored its current dominance. Revenue reached $96.22 billion, a 105.8% year-over-year increase, driven by $89.02 billion in Data Center revenue and non-GAAP earnings per share of $2.22. Supply commitments surged to $279 billion, predominantly tied to memory requirements for the upcoming Vera Rubin architecture. Full breakdowns are available in the Q2 FY2027 8-K exhibit. Looking ahead, CEO Jensen Huang projected that fiscal 2028 revenue should grow approximately 70% year over year. “At this moment, we have supply for 70%. We have more supply than 70%, but about 70%. Our demand is much higher than that,” Huang stated. Following the report, shares declined 4.57% on August 28, 2026, falling from $227.98 to $217.55, though they remain up 16.79% year to date and trade at a forward price-to-earnings ratio of 26.

The competitive landscape is shifting as custom silicon gains traction. Chip analyst and 247 Wall Street contributor Eric Bleeker highlighted a SemiAnalysis headline declaring “OpenAI Jalapeno better than Nvidia Blackwell,” noting that early benchmarks indicate superior total cost of ownership compared to Blackwell. On CNBC, Jim Cramer observed that “Nvidia invested 30 billion in OpenAI” and noted OpenAI is “downright gleeful about inventing this new chip, Jalapeno, that can compete with Nvidia’s.” Structural shifts were further emphasized by Chamath Palihapitiya on the All-In podcast: “You’re going to look at these big companies in five years, they’re all going to have their own cloud, they’re all going to have their own models, they’re all going to have their own silicon, they’re all going to have their own data centers.”

Huang countered these long-term projections by pointing to existing and planned commitments, which represent approximately 12 gigawatts of NVIDIA compute through 2030. He added that AI laboratory demand should account for roughly a quarter of NVIDIA’s business next year. Visser concurred on the near-term outlook, describing the current infrastructure cycle as “the sweet spot of the infrastructure build out” and forecasting “another three to five years of needing a lot.” The true divergence lies in terminal value. If custom silicon captures even a quarter of hyperscaler workloads by 2031, the discounted cash flow model supporting today’s equity price would need downward revision. Conversely, if Vera Rubin’s economics of $40 billion per gigawatt continue to expand, NVIDIA’s competitive moat remains intact.

Beyond NVIDIA, broader market movements highlight divergent AI-driven outcomes. Eli Lilly (NYSE:LLY) cited a co-innovation AI lab with NVIDIA for drug discovery, reporting $22.97 billion in second-quarter revenue and EPS of $8.38. The company raised its full-year 2026 revenue guidance to between $85.0 billion and $87.0 billion, and its shares have climbed 61.49% over the past year. In contrast, Coinbase (NASDAQ:COIN) reported an 18.51% year-over-year revenue decline to $1.22 billion in the second quarter. Despite generating past $100 million in annualized prediction-markets revenue, the stock has fallen 21% year to date.

For NVIDIA investors, the critical forward-looking question is straightforward. If Jalapeno and subsequent custom chips capture even a fraction of inference workloads by 2030, will the current valuation multiple compress before fiscal 2028 growth materializes? Market participants should monitor memory pricing trends, the composition of hyperscaler capital expenditure, and the deployment velocity of OpenAI’s proprietary silicon.

For questions or corrections, contact [email protected]. Joel South covers large-cap equities, dividend investing, and major market trends, specializing in earnings analysis, valuation frameworks, and translating complex data into actionable investor insights. He brings over 15 years of experience as both an investor and financial journalist, including 12 years at The Motley Fool, where he served as an investment analyst, Bureau Chief, and later led the Fool.com investing news desk. He has also co-hosted an investing podcast and appeared across television and radio discussing market developments.