Nvidia is expanding beyond hardware sales by directly financing AI infrastructure projects to accelerate adoption of its own silicon.
Nvidia has expanded far beyond its origins as a graphics processing unit (GPU) manufacturer, now encompassing networking, software, computing systems, and strategic investments in artificial intelligence companies. Crucially, it is also helping to finance the very infrastructure that purchases its products. With annual revenue projected to approach $400 billion this year, investors are increasingly questioning how long this extraordinary growth trajectory can sustain itself.
A week and a half prior to publication, Nvidia released its second-quarter results for fiscal 2027. The company follows a fiscal year that concludes in the last week of January rather than aligning with the calendar year. For the quarter, revenue totaled $96.2 billion, representing a 106% year-over-year increase, with guidance pointing to approximately $108 billion for the next quarter. These figures reinforced market expectations that rapid expansion will continue through calendar 2027. Following the earnings report, Nvidia’s shares surged, pushing its market capitalization to roughly $5.5 trillion on August 27. The following day saw a modest sell-off, bringing the valuation down to $5.24 trillion.
As the entity with the most profound impact on the AI revolution, Nvidia’s processors, computing systems, networking equipment, and software underpin much of the infrastructure used to build and run the world’s most advanced AI models. In recent years, the company has rapidly broadened its portfolio—from GPU chips to full-stack communications systems, storage, software, and integrated computing platforms—aiming to become the central provider across nearly every layer of data center infrastructure required for the AI era. Several facets of this expansion have been accelerated through strategic acquisitions.
Nvidia’s meteoric rise prompts a broader industry question: Is this merely an unusual investment cycle, or does it signal the start of a long-term structural shift in the semiconductor sector? Traditionally, semiconductors have been highly cyclical, characterized by alternating phases of surging demand and investment followed by overcapacity, inventory gluts, and slowdowns. While the typical cycle spans three to five years, duration varies significantly across segments. This pattern is especially visible in memory chips but also applies to CPUs and other processor categories. However, the current surge in demand for GPUs and AI systems operates on a fundamentally different scale. Massive capital expenditures by cloud providers and data centers are rapidly accelerating computing demand, potentially extending the positive investment cycle well beyond historical norms. Consequently, semiconductor investors remain vigilant for any signs of deceleration.
Capital market performance underscores the uniqueness of the current cycle. The SOXX semiconductor index fund appreciated approximately tenfold between June 2019 and June 2026, outpacing the SPY S&P 500 ETF, which rose 2.5 times over the same period. Nvidia’s own stock climbed roughly 48-fold during that timeframe. A portion of this exceptional appreciation stems from the 2020 acquisition of Mellanox, which substantially scaled Nvidia’s data center networking and communications capabilities. Since early July, the SOXX fund has retreated by approximately 20%, whereas Nvidia’s stock rebounded to near its earlier-year peaks following its strong earnings release. This divergence highlights that investors do not price all chipmakers equally, with Nvidia continuing to command premium valuations driven by exceptionally high growth expectations.
Nvidia employs approximately 42,000 people globally, including roughly 6,000 in Israel. Its Israeli footprint has grown substantially since the Mellanox acquisition, encompassing significant operations in communications, chip design, software, and systems engineering. Given its scale and scope, Nvidia exerts an increasing influence on the Israeli economy, impacting exports, employment, high-tech activity, and state tax revenues. Notably, the company accounts for 90% of Israel’s foreign production, actively reshaping national GDP metrics.
Over the past decade, Nvidia has continuously upgraded its GPU lineup. Products once dedicated primarily to graphics rendering have evolved into essential parallel processors for AI applications, simulations, and data centers. In 2022, the company introduced the Hopper architecture, headlined by the H100 processor, engineered specifically for training large AI models, high-performance computing, and data center deployments. By 2024, Nvidia launched the Blackwell architecture, shifting its strategy from single-chip solutions to comprehensive systems integrating CPUs, GPUs, memory, inter-accelerator communication, and software.
In December 2025, Nvidia acquired a substantial portion of Groq’s technology, notably its LPU architecture optimized for high-speed inference—the execution of pre-trained models to generate responses. The transaction carried a $13 billion upfront payment plus an additional $4 billion in deferred payments. Another major advancement is the Vera Rubin architecture, which deepens integration across processors, AI accelerators, memory, and communications to enable faster, more efficient large-scale operations. Designed for both model training and inference optimization, Vera Rubin aims to deliver more responses in less time with reduced power consumption, particularly for large models. During the post-earnings analyst call, Nvidia confirmed that commercial deliveries of Vera Rubin-based systems began in August 2026.
Over the weekend, Nvidia announced an agreement to acquire Hugging Face for $11.9 billion in cash and an additional $1 billion in options. The deal brings Nvidia access to open-source models, a vast developer community, and runtime capabilities. Combined with the Groq acquisition, these moves extend Nvidia’s reach further along the AI value chain as it seeks to maximize revenue generation per data center. Market expectations indicate that data center workloads will gradually pivot from training—dominated by developers building large language models—toward inference, which powers live applications and task execution. Nvidia is positioning for this transition by combining Groq’s low-latency inference expertise with the Vera Rubin platform, which can integrate Groq’s LPX inference accelerators alongside its native GPUs and CPUs.
Investor caution largely stems from anticipated intensifying competition in inference-focused chips, particularly from Google. A sustained shift toward inference workloads could therefore present a steeper competitive landscape for Nvidia. Nevertheless, the AI boom has driven unprecedented revenue growth for the company. Fiscal 2023 revenue (ending January 2023) stood at $26.97 billion, jumping to $60.92 billion in fiscal 2024, doubling again to $130.5 billion in fiscal 2025, and reaching $215.9 billion in fiscal 2026—a 65% growth rate. This year, revenue is projected to surge to approximately $400 billion, an 85% increase. On the earnings call, Nvidia guided for another 70% revenue jump in the upcoming fiscal year, which concludes in January 2028. Management’s confidence rests heavily on sustained demand and the assessment that supply constraints, rather than weak orders, will be the primary bottleneck. Ninety-two percent of Nvidia’s revenue currently derives from data center infrastructure. Nvidia’s gross mar