Under the torrent of token reasoning, GPU powerhouses are vying to control the "AI model entrance"! NVIDIA Corporation (NVDA.US) plans to swallow Hugging Face for $14 billion this week.

Last Thursday, NVIDIA Corporation (NVDA.US), the "super giant" of AI chips, announced a spectacular earnings report and is in deep negotiations to acquire Hugging Face, with a total deal size potentially reaching approximately $14 billion. This includes an estimated acquisition price of about $12.9 billion, alongside a possible $1 billion employee retention plan. This transaction will enable NVIDIA Corporation to control a crucial global platform for sharing open AI models, directly connecting developers, models, and enterprise clients, while reducing its heavy reliance on a few technology giants that contribute significant revenue data but are actively developing their own data center AI chips. What NVIDIA Corporation truly seeks to acquire is not just a typical AI application software company, but rather the "model distribution layer" and global entry point for AI developers in the era of artificial intelligence. Hugging Face functions like the "GitHub of AI models," connecting models, datasets, development tools, enterprise clients, and developers. After the acquisition, NVIDIA Corporation can enhance the stages of "model discoveryevaluationoptimizationactual deployment" on top of its GPU, NVLink network, CUDA software stack, and AI systems, allowing open models to fit more seamlessly into its massive AI training and inference platform, while identifying ahead of time which model architectures and workloads are generating the next wave of demand for computing power. This also serves as a structural hedge for NVIDIA Corporation against the proprietary model giants and the self-developed chips of ultra-large cloud vendors: as long as the ecosystem of open models remains vibrant, the demand for AI applications will not be completely locked by a few closed platforms and their proprietary accelerators. The rationale behind Stripes acquisition of OpenRouter is similarOpenRouter controls model selection, request routing, cost, and billing entry points, while Hugging Face governs the discovery, sharing, development, and deployment access of models; Stripe is competing for "model invocation traffic," while NVIDIA Corporation is contesting how "model adoption translates into computing power demand." NVIDIA Corporation's 70% growth outlook shatters the so-called "peak AI theory," and the global AI computing power infrastructure remains as vibrant as ever. Global investors, after experiencing a sharp decline in July due to AI deleveraging and congestion, still exhibit a strong risk appetite for the AI computing power supply chain. The Philadelphia Semiconductor Index had doubled from its early-year levels at Junes peak, before retreating as much as 29% in July; as of September 1, it closed at 11,288.61 points, remaining up more than 50% year-to-date. South Koreas KOSPI index fell from approximately 9,385 points in June to 5,593.56 points on July 30, a maximum retracement of over 40%, but then rebounded by more than 20%, indicating a return to a technical bull market; it closed at 6,835.80 points on September 1, with an approximately 62% increase for the year. The recent developments of these two major "AI computing power barometer" indexes suggest that the July drop was more akin to a stress test driven by high leverage, crowded positioning, and valuation compression, rather than a fundamental reversal in AI computing power demand. The demand for AI computing infrastructure is transitioning from budget intentions of major cloud computing and AI application development companies to multiyear capacity commitments. It was reported that Anthropic signed a $35 billion agreement, supported by NVIDIA Corporation, for approximately 350 megawatts of AI cloud computing capacity; not long ago, it secured about $45 billion for a six-year deal worth approximately 460 megawatts with Nscale, which will deploy NVIDIA Corporations next-generation AI computing clustersthe Vera Rubin platformon a large scale. South Korea's exports grew by 68.7% year-on-year to $98.26 billion in August, with semiconductor product exports surging 209% to a record $46.65 billion, accounting for 47.5% of total exports. AI GPU clusters and storage chips remain the most prominent bottlenecks in global AI computing systems under the wave of Token inference. TrendForce predicts a quarter-on-quarter increase of approximately 13%-18% for DRAM and 10%-15% for NAND Flash contract prices in the third quarter; by 2026, the cumulative price increase for server DRAM and enterprise-grade solid-state disks could reach approximately 270% and 235%, respectively, while HBM contract prices may rise by an additional 70%-140% in 2027. DRAM and NAND are expected to account for 47% of capital expenditures from cloud service providers, and this will increase to 68% by 2027. The AI competition has entered a "battle for the entrance to large models": NVIDIA Corporation aims to acquire Hugging Face. NVIDIA Corporation is expanding its competitive boundaries from chip performance to model distribution and developer relations, intending to grasp the source of future computing power demand. Successfully acquiring Hugging Face would represent NVIDIA Corporation's upgrade from a computing power supplier to a comprehensive foundational platform that encompasses model development, deployment, and commercialization in AI. According to informed sources, NVIDIA Corporation is in in-depth negotiations to acquire the AI startup Hugging Face, with a potential deal size of approximately $14 billion. One anonymous insider, due to the details not being public, indicated that NVIDIA Corporation might reach an agreement to acquire Hugging Face for $12.9 billion as soon as this week. The person noted that the deal could also include a retention plan worth up to $1 billion for Hugging Face employees. Sources say that no final agreement has been reached yet, and the timing or specifics of the deal may still change. Representatives from NVIDIA Corporation and Hugging Face declined to comment. The acquisition of Hugging Face will be the largest move taken by CEO Jensen Huang to expand the reach of AI applications and broaden the customer base. This transaction will allow NVIDIA Corporation to control a key platform where developers can showcase and share AI models. Jensen Huang is committed to nurturing what are referred to as open-source large AI models to drive robust long-term expansion of AI GPU demand and ecosystem, in order to avoid the technological dominance of AI by a few large companiesthese firms currently contribute the majority of NVIDIA Corporations revenue while actively advancing their own chip projects. NVIDIA Corporation is already one of Hugging Face's investors, along with other investors such as Alphabet Inc. Class C, Amazon.com, Inc., Intel Corporation, and Salesforce. Founded in 2016, Hugging Face operates a platform for sharing AI models that others can use freely. This startup was valued at $4.5 billion during a funding round three years ago. The well-known outlet Business Insider previously reported on NVIDIA Corporation's negotiations with the company. Hugging Face develops AI software and offers software hosting services to other companies. The company was recently at the center of a cybersecurity incident: a model being tested by OpenAI accidentally breached the platform. This security vulnerability has raised concerns about the safety of cutting-edge AI technology. OpenAI stated that it could have responded more proactively to prevent the attack on the Hugging Face system. In the past year, NVIDIA Corporation has secured multiple deals, including an August licensing agreement worth $6 billion with the startup Poolside, which also included job offers for many of its employees. NVIDIA Corporation also paid nearly $20 billion to acquire most of the business of the chip startup Groq. NVIDIA Corporation is the worlds most valuable publicly traded company and the absolute leader in the AI accelerator field; such chips are instrumental in training and running AI models. To scale the AI economy, the company has gradually increased its range of other technologies, including software. Based in Santa Clara, California, NVIDIA Corporation recently provided higher-than-expected sales forecasts for the fiscal year 2028, with Jensen Huang's management team projecting overall revenue growth of approximately 70% for fiscal year 2028. Global AI capital expenditure has crossed into the trillion-dollar era, and the bull market trajectory of the AI chip super giant is far from over. As global AI capital expenditure continues its strong expansion, Hugging Face can fill the software distribution gap for NVIDIA Corporation, shifting its growth logic from chip shipments to comprehensive ecosystem penetration. If the deal goes through, NVIDIA Corporation will further control the model discovery, distribution, and deployment entry points from being a GPU supplier, efficiently translating developer and application ecosystem Token traffic into CUDA and AI computing power demand. Deep collaboration between the model platform and GPU, CUDA, and the Vera Rubin system can enhance developer stickiness and demand insights, enabling NVIDIA Corporation to capture incremental opportunities for training and inference computing power more swiftly. Based on the AI computing power supply chain and NVIDIA Corporation's performance, along with positive signals related to AI computing demand released at the recent Hot Chips conference, Goldman Sachs Group, Inc. has recently translated the well-known concept of "capital expenditure for ultra-large US cloud service providers (hyperscalers)" into "global AI investment." Goldman Sachs Group, Inc. stated that the previously compiled figure of $794 billion omitted private AI enterprises, Asian companies, and non-US projects, and erroneously included traditional capital expenditure not fully dedicated to AI; Goldman Sachs Group, Inc. is now predicting approximately $1 trillion in global AI investment by 2026, with about $581 billion occurring in the US. Another Wall Street financial giant, Morgan Stanley, predicts that by 2028, nearly $3 trillion in AI-related infrastructure investment will flow through the global economy, with over 80% of the expenditure still ahead. The core incremental demand from the AI computing power side stems from AI transitioning from "answering questions" to "executing workflows": intelligent agents need to continuously plan, call tools, verify results, and retry upon failure, with the token consumption of AI inference for singular tasks potentially reaching 10, 20, or even 50 times that of traditional chat queries; world models will expand the boundaries of computing power from text to Siasun Robot & Automation, industrial simulation, and physical systems. Goldman Sachs Group, Inc. officially predicts that by 2030, global token consumption may grow 24 times to 120 quadrillion per month; meanwhile, the unit cost of AI inference tokens may decline annually by 60%-70%, creating the Jevons effect of "cost reductionapplication diffusiontotal demand growth." NVIDIA Corporation's latest performance and outlook, combined with the substantial cloud computing resource agreements recently signed by leading AI applications like Anthropic and the robust semiconductor export data from South Korea, underscore that global AI computing demand is still far from reaching its peak and is in a large-scale expansion cycle. NVIDIA Corporation reported a year-on-year revenue increase of 106% to $96.2 billion for the second quarter of fiscal year 2027, with data center business revenue surging 117% to $89 billion; guidance for third quarter revenue reached $108 billion, while the company's management forecasts around 70% revenue growth for fiscal year 2028. Such a strong and still supply-constrained outlook is the fundamental basis for Wall Street's continued upward revisions of the company's earnings forecasts and target prices. Goldman Sachs Group, Inc., Morgan Stanley, and other Wall Street financial giants have a positive outlook on the demand for NVIDIA Corporations next-generation computing architecture, Vera Rubin, the growth visibility of GPU clusters in 2027 and beyond, and the expansion of AI infrastructure; their consensus bullish judgment stems from the sustained strength of AI computing demand, increased volume of Vera Rubin, and the ongoing enhancement of NVIDIA Corporation's integrated software and hardware platform advantages. Citigroup has raised its target price from $300 to $315; Morgan Stanley has raised it from $288 to $300 while maintaining an "overweight" rating; Goldman Sachs Group, Inc. has increased its price target from $285 to $300; and Bank of America Corp continues to maintain its "buy" rating, global preferred stock status, and $350 target price.