"NVIDIA is essentially a platform-computing company!"
On the evening of October 28, Jensen Huang spent two hours at the GTC conference painting a detailed blueprint toward an AI-driven new industrial revolution. Huang is a believer in "First Principles," mentioning in multiple public settings how he uses "First Principles" to think and manage. Next, we will analyze and decode NVIDIA's strategic blueprint based on "First Principles."
We strongly recommend you read Jensen Huang's speech transcript before reading this article.
Jensen Huang's Latest October GTC Speech (Part 1): AI Factory Opens a New Era — Moore's Law may have ended, but computing has not stopped.
Jensen Huang's Latest October GTC Speech (Part 2): AI Factory Opens a New Era — A look ahead at the next generation of the AI ecosystem.
01 Business Essence: What Kind of Company Is NVIDIA, Really? Clarifying "Why It Exists"
When NVIDIA is mentioned, the industry holds two typical views of its positioning: some see it as a hardware company focused on high-performance GPU R&D, occupying a core position in the global market on the strength of its superior hardware compute power; others emphasize its software attributes, arguing that the programming platform and ecosystem centered on CUDA is the key to its competitive moat.
Yet Jensen Huang has stated clearly that NVIDIA's core positioning is neither a pure hardware nor a pure software company, but a "platform-computing" company that exists to solve computing problems. This positioning runs through NVIDIA's entire development: from the very beginning, all its strategic moves have revolved around this core essence. Its early GPU investment arose from the insight that general-purpose CPUs hit significant performance bottlenecks in compute tasks such as graphics rendering and complex scenario simulation, creating an urgent need for a dedicated computing architecture to break the compute ceiling. The original intent of GPU R&D was precisely to efficiently solve the core computing problem of "insufficient compute supply."
And its heavy investment in the CUDA platform was based on the deep understanding that "hardware capability must be unlocked through a software ecosystem." CUDA is not an isolated software product but the critical bridge connecting hardware performance with developer applications; its core value lies in solving the pain point that "hardware compute is hard to translate effectively into real application performance." In the end, both GPU and CUDA are platform-computing vehicles through which NVIDIA achieves its goal of "solving computing problems." Its strategy has always stayed tightly anchored to the business essence, forming a complete system where software and hardware empower each other.
This core positioning was further concretized at Jensen Huang's GTC speech on October 28, 2025, into a strategic closed loop of "new direction — new product — new ecosystem." These three are not isolated strategic modules but a deepening extension of the "solving computing problems" platform-computing essence, forming an organic whole of "goal — means — safeguard" around "breaking through computing bottlenecks more efficiently and broadly, and expanding the boundaries of computing."
02 New Direction: Anchor the Next Battlefield of "Computing Problems," Clarifying "What to Solve"
The "computing paradigm + industry form" dual-wheel new direction proposed by Huang essentially and precisely anticipates and locks in the three most core computing pain points of the next 5-10 years, drawing the "problem-solving scope" for NVIDIA's strategy:
Cracking the "compute ceiling" problem: as Moore's Law fails, general-purpose CPUs can no longer support the exponentially growing compute demand in scenarios such as AI training and quantum simulation (e.g., large-model training requires trillions of operations). To this end, NVIDIA turned to "GPU + AI accelerated computing," replacing "general architecture" with "dedicated computing architecture"—essentially breaking the traditional computing performance bottleneck through hardware-form innovation;
Solving the "computing-industry disconnect" problem: traditional computing only served "tool-type needs" (such as Excel office work, web browsing) and cannot meet "autonomous-execution needs" such as factory automation and self-driving—these scenarios require computing to actively process tasks and optimize decisions, rather than passively wait for human operation. Hence the "AI Factory" concept, upgrading computing from "auxiliary tool" to "industry executor," essentially using "scenario-based computing forms" to deeply embed computing into industrial processes;
Breaking the "single-technology computing limitation" problem: in complex scenarios, a single technology cannot independently complete computing tasks—for example, a 6G base station must simultaneously process "wireless signal transmission" and "AI real-time inference," and quantum computing needs "classical compute to support error correction and data processing." To this end, NVIDIA promotes cross-technology fusion, essentially integrating the strengths of multiple technologies through a "collaborative computing model" to solve complex computing scenarios that a single technology struggles to cover.
It is clear that the "new direction" is not an invented track out of thin air, but, based on a deep insight into the computing industry's pain points, an early lock on the core computing problems to be tackled in the future, anchoring the direction for subsequent technology and ecosystem layout.
03 New Products: Build the "Full-Stack Problem-Solving Toolkit," Clarifying "What to Solve With"
The "full-stack hardware-software product matrix" unveiled in the speech—each product is a "customized solution" for a computing pain point in the "new direction," none detached from the core goal of "solving computing problems"—forms a complete tool chain covering "compute production — compute adaptation — compute application":
Hardware products: break computing's "physical bottleneck." The Blackwell platform (GB200 NVL72) achieves a 10x performance gain through "extreme co-design" while lowering token-generation cost, chiefly solving the problem of "insufficient energy efficiency for large-scale AI computing"; NVLink interconnect technology connects QPU and GPU with 4-microsecond low latency, solving the problem of "excessively high latency in quantum-classical computing coordination"—both break, through hardware innovation, the physical limits of computing in performance and latency.
Platform products: crack computing's "industry-adaptation bottleneck." NVIDIA Arc (6G base-station computer) integrates Grace CPU, Blackwell GPU, and ConnectX networking to realize integrated computing of "communication signals + AI inference," solving the problem of "hard multi-task computing adaptation in 6G scenarios"; the Drive Hyperion autonomous-driving platform supports real-time environment perception and decision-making for driverless cars through multi-sensor data-fusion computing, solving the problem of "adapting computing to vehicle needs in autonomous-driving scenarios"—both use scenario-based platforms as the vehicle to precisely match computing capability to industry needs.
Software tools: lower computing's "application-threshold bottleneck." The CUDA-Q open-source platform unifies the programming logic of quantum and classical computing, letting developers build applications without mastering complex quantum technology, solving the problem of "high quantum-computing programming threshold"; the Isaac GR00T-Dreams framework generates virtual training data through "dream generation" technology, reducing robots' reliance on real data for training, solving the problem of "shortage of scenario computing data for robots"—both lower, through software optimization, the application threshold of computing technology, letting more scenarios access advanced computing capability.
These products are not isolated "technology exhibits" but a "hard-software collaborative toolbox" formed around computing pain points, providing all-around support for solving computing problems from hardware performance to software adaptation.
04 New Ecosystem: Build the "Scaled Landing Network," Clarifying "How to Get More People Using It"
The core value of the "new ecosystem" is to move NVIDIA's "computing solutions" from "single-point technology breakthroughs" to "industry-wide scaled application," essentially solving the landing problem of "computing technology struggling to penetrate every industry," amplifying the value of computing solutions through coordination:
Cross-industry cooperation: solve the "hard scenario landing of computing solutions" problem. Partnering with Nokia to upgrade millions of base stations worldwide to 6G+AI puts the "6G multi-task computing solution" into the telecom industry; partnering with Uber to deploy 100,000 Robotaxis in 2027 puts the "autonomous-driving computing solution" into the mobility industry; co-building 7 AI supercomputers with the U.S. Department of Energy puts the "quantum-classical hybrid computing solution" into scientific research—by binding to vertical industry scenarios, computing solutions find concrete application carriers, avoiding "technology floating in the air."
Open-source and developer ecosystem: solve the "too few computing-technology participants" problem. Open-sourcing AI models such as Nemotron and Cosmos, and co-developing the Newton physics engine with Disney, lowers the development cost of computing technology; through the Inception program it supports 18,000 startups with CUDA technical certification and hardware resources—essentially expanding the developer base through "open source + empowerment" so that practitioners across more industries can use NVIDIA's tools to solve computing problems in their own fields (such as medical imaging analysis, supply-chain optimization), driving computing capability into specialized domains.
Ethics and security ecosystem: solve the "sustainable application of computing technology" problem. Adopting Google DeepMind's SynthID watermarking technology embeds digital identifiers into AI-generated content, solving the "IP traceability of AI computing output" problem; co-building a "cloud + edge" AI cybersecurity agent with CrowdStrike solves the "cybersecurity risk during computing" problem—by establishing ethics and security rules, computing technology lands within a compliance framework, avoiding scaling blockage from risk issues.
In short, the "new ecosystem" turns "solving computing problems" from "NVIDIA's business" into "the whole industry's business," truly bringing computing solutions into every sector through cooperation, open source, and security, maximizing value.
The NVIDIA strategic blueprint painted by Jensen Huang is essentially a "computing-centric" strategic deepening: "new direction" clarifies "which new computing problems to solve"—the "bull's-eye" of the strategy; "new products" provide "what tools to solve these problems with"—the strategy's "weapons"; "new ecosystem" safeguards "that these solutions can land across the industry"—the strategy's "safety net." The three link tightly, always revolving around the core positioning that "NVIDIA is the company that solves computing problems." Without deviating from the essence, and through the extension of "from compute breakthrough to scenario adaptation, from technology innovation to ecosystem coordination," it continuously expands the boundary of "solving computing problems," thereby laying the logical foundation for NVIDIA's long-term leadership in the computing industry.
05 Strategic Reference: How to Effectively Borrow This When Making Next-Year Planning
If NVIDIA's "new direction — new product — new ecosystem" closed-loop logic is to truly land in a company's year-end strategic plan, it must take "First Principles" as a premise: first penetrate the business surface and anchor your own essential value, otherwise the "three new" will be nothing but castles in the air detached from the core. The complete landing system unfolds around "clarify essence → solve problems → landing safeguard," with these specific steps:
Premise: Use First Principles to Decompose the Business Essence — Answer "Whose Core Problem Do We Ultimately Solve, and What Is It"
The core of First Principles is "strip the surface, return to the root"—not relying on industry conventions or competitors' practices, but focusing only on the business's most bottom-level value logic. Before year-end strategic planning, complete a 3-step essence decomposition:
Decompose core value: drop the habitual notion of "we make XX products" and ask "what is the fundamental reason users choose us" and "what irreplaceable value do we create." For example: a retail enterprise is not "a seller of goods" but essentially "efficiently connecting people with needs, lowering users' decision cost"; a manufacturing enterprise is not "a parts producer" but essentially "providing downstream with stable and reliable supply-chain assurance, raising its production efficiency";
Strip non-essential links: sort out "processes / products / services in current business unrelated to the essential value" (such as excessive marketing packaging, non-core value-added services), clarifying "which links must be kept and which can be optimized / discarded," to avoid wasting resources on non-core areas;
Validate the essence hypothesis: through customer interviews and data review (such as core-user retention rate, repurchase rate of must-have scenarios), verify "whether our judgment of the business essence holds." For example: if a retail enterprise believes its essence is "connecting people with needs," but data shows users choose it for "low price" rather than "precise matching," it must revise its essence understanding to "efficiently connecting people with needs at high cost-performance."
Only by first anchoring the business essence through First Principles will the subsequent "three new" strategy have a clear anchor, avoiding the trap of "chasing trends, blindly iterating."
Step 1: Review + Anchor the "New Direction" — Based on Business Essence, Clarify Next Year's "Core Problem to Solve"
Centered on business essence, review the year's business pain points (all pain points must correspond to problems of "deviating from essence" or "not fully realizing essential value"), and combined with industry-trend prediction, lock in 1-2 core breakthrough directions (avoid spreading across multiple lines):
Example 1: if a retail enterprise's essence is "efficiently connecting people with needs at high cost-performance," and its pain point that year is "users take long to find suitable products, low repurchase rate," it can set the new direction of "private-domain precise matching + repurchase activation";
Example 2: if a manufacturing enterprise's essence is "providing stable and reliable supply-chain assurance," and its pain point that year is "production failures cause delivery delays and many customer complaints," it can set the new direction of "smart early-warning in production + cost reduction"; core logic: the "new direction" must be "the key obstacle to better realizing the business essence," not blind innovation detached from essence.
Step 2: Focus on "New Product / Service" — Around Business Essence, Build the "Core Tool That Solves the Problem"
Based on business essence and the set "new direction," sort out the shortcomings of existing products / services (shortcomings must correspond to problems of "not supporting essential-value landing"), and clarify the core carrier to iterate or add next year, avoiding "big and comprehensive":
Prioritize building 1-2 "sharp-knife products / services": concentrate manpower and budget resources, focus on the core problem. For example: a retail enterprise, around "private-domain precise matching," develops a "user-need-tagging smart recommendation tool"; a manufacturing enterprise, around "production early-warning," builds a "real-time equipment-data monitoring and failure-prediction system";
Set small-step validation milestones: avoid one-time over-investment; set validation goals by quarter (e.g., Q1 completes prototype development, validating "whether it solves the core pain point"; Q2 runs a small pilot, validating "user / customer acceptance"), ensuring the product / service always revolves around business essence and does not deviate.
Step 3: Build the "New Ecosystem" — Support the Essence, Construct the "Resource Network for Strategic Landing"
All ecosystem-resource building must revolve around "better realizing business essence and supporting new product / service landing," avoiding "building an ecosystem for its own sake":
External: lock in 2-3 core partners. Choose partners highly aligned with business essence and the new direction, and clarify cooperation rights and duties. For example: a retail enterprise connects with a "user-behavior data-analysis SaaS vendor" (supports precise matching) and a "private-domain content service provider" (supports repurchase activation); a manufacturing enterprise connects with an "industrial-sensor supplier" (supports data collection) and an "AI algorithm company" (supports failure prediction);
Internal: optimize the support system. Adjust the organizational structure (e.g., form an "essential-value-landing task force" to avoid departmental walls), and sort out budget allocation (prioritize funding core product R&D and ecosystem-cooperation landing);
Establish a risk backstop mechanism: around business essence, set up compliance review (e.g., data security, industry-regulation requirements) and contingency plans (e.g., supply-chain disruption, core-product failure), ensuring that in the strategic-landing process, the realization of essential value is not affected by major risks.
The key to a company's year-end strategic planning is to form a strategic-thinking closed loop of "essence → direction → product → ecosystem": first anchor "business essence" with First Principles (answering "who we are, for whom we create value"), then use "new direction" to lock in "the core problems blocking essence realization," use "new product / service" to build the "problem-solving tool," and use "new ecosystem" to safeguard "tool landing, essence landing."
Under this logic, the "three new" are not isolated strategic modules but a complete closed loop from "cognition to landing" formed around business essence. It avoids strategic "idling" and ensures every move serves the core value, giving next year's strategy both direction and strong execution.
