We are entering an historic transformation in computing architecture. Artificial intelligence – especially the rise of large-scale tokenized content generation (generative AI) – is reshaping everything from silicon to software. This shift is as profound as any in tech history, impacting the entire stack: cutting-edge chips and AI factories at the infrastructure layer, new middleware and development frameworks, and a wave of AI-infused applications and services. What’s unfolding now echoes prior computing revolutions (mainframe, client-server, cloud) but is larger in volume, greater in value, and faster in velocity than these previous eras. Organizations that recognize the economic and strategic implications of this inflection point will be positioned to lead in the coming era of AI-centric IT.
An Historic Architectural Shift is Underway
Today’s enterprise technology stack is being radically reinvented to accommodate AI at scale. At the silicon level, traditional CPU-driven computing is giving way to specialized accelerators – GPUs, AI chips and advanced networking – optimized for parallel processing. NVIDIA’s newest Blackwell GPUs, for example, deliver 1 exaflop in a single rack with innovative liquid-cooling, a “wholesale reinvention of AI computing
At the infrastructure level, cloud data centers are transforming into AI supercomputing clusters, or “AI factories,” built to churn through petabytes of data and trillions of model parameters. Middleware and tools are rising to abstract the complexity of AI – from new AI operating systems like NVIDIA’s open-source Dynamo (which optimizes large-scale inference) to advanced frameworks for machine learning. Meanwhile, applications and services are being reimagined with generative AI capabilities (from coding assistants to synthetic media generators), creating new value streams and business models.
Critically, this shift is framed by economics and strategy: value is concentrating on fewer platforms that can offer end-to-end AI solutions. Just as past eras saw vertical integration (e.g. IBM in mainframes, Wintel in PCs), today we see a similar pattern with AI. Companies like NVIDIA now span chips-to-software, offering a full-stack approach that locks in customers. Cloud providers are pouring capital into AI offerings to attract enterprise workloads, even as those enterprises themselves reconsider IT architecture to harness AI for competitive advantage. In short, the balance of power and profit in tech is tilting toward whoever controls the new AI-centric stack – a dynamic that every tech executive must factor into strategic planning.
Bigger and Possibly Faster than Past Waves: The 3Vs Framework
This revolution in computing resembles prior paradigm shifts, but on a far grander scale and accelerated timeline. We characterize its magnitude using the 3Vs framework – Volume, Value, and Velocity – to compare it with earlier waves:
- Volume: The sheer scale of data, model size, and compute demand in the AI era dwarfs anything before. Today’s AI models ingest unprecedented volumes of data and require massive parallel compute. For example, NVIDIA’s Jensen Huang noted that emerging agentic AI workloads are driving a 100x increase in required computation compared to a year ago. The global infrastructure to support AI is expanding accordingly – the total data center market is projected to surpass $1 trillion by 2032, on its way to $1.7T by 2035. This volume of investment and capacity far exceeds the build-out during the early cloud era.
- Value: Each successive tech wave has created new business value, but AI’s promise is singular in its breadth. Generative AI and advanced automation stand to unlock trillions in economic value by augmenting human work and enabling new products and services. Already, internet giants have proven substantial ROI from AI in areas like ad targeting and search.
Enterprises across industries see AI as a path to revenue growth and efficiency – from richer customer experiences to autonomous operations. Crucially, despite macroeconomic headwinds, business leaders continue funding AI initiatives because the expected value (in productivity gains, innovation, and competitive edge) is so high. In surveys, 40% of organizations are reallocating funds from other IT budgets to feed AI projects reflecting a strategic bet that AI will deliver superior returns on investment. - Velocity: The pace of adoption and innovation in the AI age is unprecedented. Prior shifts (like client-server or cloud computing) unfolded over a decade or more. In contrast, generative AI went from research to mainstream in a matter of months – e.g., ChatGPT reached 100 million users in ~2 months, a speed unheard of in tech adoption. Enterprise adoption of AI is similarly rapid. According to recent data, what was “year of experimentation” in 2023 is turning into a demand for AI at scale in 2024, with companies pushing for real ROI on accelerated timelines. From a hardware perspective, vendors are releasing new AI-optimized chips on an annual cadence (NVIDIA’s roadmap promises yearly leaps in performance), compressing innovation cycles. The net effect is a compressed timeframe for this computing transition – what might have taken 20 years in the past may now play out in 5–10 years.
Together, these “3Vs” illustrate why this AI-driven architectural shift is not just another cycle, but a historic inflection point. The volume of resources being marshaled, the value at stake for businesses and economies, and the velocity of change are all at record levels. Technology leaders must recognize that playing wait-and-see is not an option – the train is leaving the station faster than ever.
Economic Crosscurrents: IT Spending Pullback vs. AI Priorities
Despite the excitement around AI, today’s macroeconomic climate has put overall IT spending under pressure. Tariffs, inflation fears, and cautious corporate budgets have led to tighter IT purse strings in many areas. CIO surveys from ETR show that in early 2024, global tech budgets were expected to grow only about 4.3% – a modest uptick from ~3.5% growth in 2023. Expectations in ETRs April survey have softened to 3.4% annual spending expectations.
This is a far cry from the double-digit surges seen during the boom years, and it reflects a persistent deceleration from the peaks of the pandemic-era digital spree. Many large enterprises (including members of the Global 2000) plan to undershoot even these average growth rates, essentially flat or low-single-digit budget increases, as they grapple with economic uncertainty.
Yet, within this broad belt-tightening, AI remains a red-hot priority. Technology executives are not materially cutting, and in many cases are increasing, funding for AI initiatives even as they trim in other areas. Recent drill-down data of IT decision-makers underscores this divergence: 40% of organizations are funding new AI projects by reallocating spend from other existing budgets. In other words, dollars are being pulled from elsewhere in IT – such as legacy software, routine upgrades, or even headcount – and redirected into AI development and infrastructure. Nearly half of AI spend is still “fresh” (incremental) budget, but the willingness to cannibalize other projects highlights how strategic AI has become.
IT leaders report delaying non-critical projects, consolidating vendors, and even reducing staff in traditional areas, partly to free up resources for AI and automation. Cloud cost optimization – a big theme of the past few years – has taken a back seat, as organizations accept higher cloud bills if it means accelerating AI gains. The overall IT spending outlook, therefore, masks a two-speed dynamic: a general cooling of expenditures offset by intense investment in AI. It’s telling that analysts view 2024 as “the year of AI ROI” – the year when those hefty AI investments must start paying off to justify themselves. If they do, we could see a resurgence of tech spending in late 2024 and beyond; if not, even AI could face more scrutiny. But for now, the momentum behind AI is undiminished – enterprises clearly fear missing the AI wave more than they fear overspending in a soft economy.
Cloud AI: Massive Scale in the First Vector
The first of Jensen Huang’s three AI vectors is the Cloud – and indeed, the cloud is where AI has scaled fastest and furthest to date. The world’s major cloud service providers (Amazon Web Services, Microsoft Azure, Google Cloud, and others) are investing enormous capital to stand up AI supercomputers available as services. These hyperscalers are racing to offer the most advanced AI platforms – from rentable GPU clusters and specialized AI chips, to managed ML frameworks and pre-trained language models accessible via API. The result is an explosion of AI capacity in the cloud that individual enterprises can tap on-demand, vastly lowering the barrier to entry for advanced AI work.
Accelerated computing in the cloud is thus surging. Our research indicates that the portion of data center spend going to specialized AI/parallel computing infrastructure is growing ~23% CAGR, far outpacing traditional IT. In 2020, only about 8% of data center capital went toward AI-specific systems; by 2030, more than 50% is projected. Much of that shift is driven by cloud providers: rather than adding general-purpose servers, they are pouring money into GPU farms, AI accelerators, and high-speed networks to connect them. This means traditional enterprise workloads are being rapidly “cloudified” and accelerated – e.g. instead of expanding an on-premises data center, a firm might rent NVIDIA HGX instances on Azure, or use AWS’s Trainium/Inferentia chips for AI inference tasks. The economics also favor this trend: cloud AI services offer virtually unlimited capacity without upfront capex, making it attractive for enterprises to experiment and scale as needed.
The value proposition of cloud AI is compelling. It provides access to leading-edge AI innovation (the latest chips, models, and tools) on a pay-as-you-go basis. It also benefits from immense network effects – every new AI service (like OpenAI’s or Google’s models) that gets deployed in the cloud draws more customers, which funds further investment by cloud providers into even larger clusters. This virtuous cycle has resulted in what NVIDIA describes as “AI factories” run by the likes of AWS and Google – effectively next-generation data centers tuned for AI workloads rather than general computing. In Huang’s words, these AI factories will replace traditional data centers over time, analogous to how cloud infrastructure replaced many corporate server rooms in the last decade.
From a business strategy view, cloud vendors see AI services as a huge growth vector (to justify their rising capex) and a stickiness factor for customers. If a company builds its AI on a particular cloud’s proprietary services or hardware, that cloud gains a long-term customer. It’s a play reminiscent of past platform battles: establishing de facto standards (e.g. Amazon’s AI stack, Microsoft’s Azure AI studio) that developers and enterprises rally around. In the process, cloud providers are vying to become the default AI platform, capturing a significant share of the value created by the AI boom. We are already seeing early signs of market consolidation around a few AI cloud ecosystems, with NVIDIA as an essential partner powering many of them. For tech decision-makers, the implication is clear: even if you plan to run AI on-prem, you will likely adopt cloud-like architectures and partnerships to remain competitive with the scale and pace of innovation set by the cloud giants.
Enterprise AI: Re-architecting the Modern Data Center
The second major vector is Enterprise AI – bringing AI’s power to the on-premises and hybrid data centers of mainstream businesses. While cloud providers currently lead in AI capacity, large enterprises and industry-specific firms are now aggressively building or upgrading their own AI infrastructure to meet needs that the cloud alone cannot. Reasons include data sovereignty, latency requirements, customization, and cost predictability for steady AI workloads. Enterprise AI is about enabling any organization – banks, manufacturers, hospitals, government agencies – to wield AI on their own terms, often by transforming their traditional IT environments into something resembling a mini “AI factory.”
This trend portends a significant data center transformation trajectory for enterprises. In essence, the classic enterprise data center (dominated by x86 servers, storage arrays, and monolithic apps) is evolving into a highly distributed, accelerator-rich environment. Projections show that by the mid-2030s, 80–90% of data center processor investments could be in accelerated architectures (GPUs, FPGAs, AI ASICs) rather than standard CPUs. In other words, the old general-purpose server is in decline, set to become the minority of infrastructure. Even today, we estimate NVIDIA alone accounts for roughly a quarter of all data center silicon spend – a stunning shift that speaks to the fast rise of AI workloads. Enterprise tech leaders must plan for this changing mix of compute: fewer racks of traditional servers, more high-density AI systems, and the networking and power infrastructure to support them.
This on-prem AI push opens both opportunities and risks for incumbent enterprise IT vendors. Companies like Dell Technologies, Hewlett Packard Enterprise (HPE), IBM, Cisco, and others have long supplied the hardware and services for corporate data centers. Now they face an inflection point: their customers want AI solutions, not just standard IT boxes. Vendors that rapidly align with the AI trend can capture new growth, for example by offering integrated AI systems (as Dell is doing with Project Helix for generative AI, in partnership with NVIDIA) or as-a-service AI cloud offerings on-prem (as HPE does via GreenLake). These firms have deep enterprise relationships and support capabilities, positioning them to build “enterprise AI factories” behind customer firewalls. IBM, for instance, is infusing AI across its hybrid cloud and mainframe offerings, aiming to leverage its AI research (Watsonx) and industry expertise to stay relevant.
However, the window for these incumbents is narrow – likely 18–24 months – to firmly establish their value in the AI era. If they fail to deliver robust AI infrastructure options quickly, enterprises may bypass traditional vendors, either by shifting more workloads to public cloud or by adopting newer players (including pure-play AI infrastructure startups). The risk is a repeat of the early cloud disruption: some on-prem suppliers were caught flat-footed as customers migrated to AWS/Azure. Now the stakes are higher – entire data center architectures are at stake. On-prem vendors must convince clients that they can provide comparable AI performance, ease-of-use, and cost-effectiveness as the hyperscalers, while also addressing data governance and customization needs better than cloud can. Those who succeed will secure a role in the next decade’s enterprise tech landscape; those who don’t could see their core server and storage businesses erode rapidly as AI workloads render older systems obsolete.
From the enterprise buyer’s perspective, the strategy likely involves hybrid architectures: a mix of on-prem AI capacity for sensitive or high-throughput tasks, plus cloud bursting for elastic scale. Ensuring interoperability (between cloud AI services and on-prem tools) will be crucial – a fact not lost on vendors, which are collaborating across traditional boundaries. NVIDIA’s dominance in AI silicon means it has partnerships with practically every enterprise IT provider, enabling solutions like NVIDIA-certified systems from Dell/HPE and software stacks that mirror those used in the cloud. The message to CIOs and CTOs is clear: begin redesigning your IT environments now for an AI-centric future. That means investing in accelerated hardware, retraining staff on AI frameworks, and updating procurement strategies to evaluate offerings by their AI capabilities. The enterprise data center in five years will look dramatically different, and planning for that transformation is an urgent priority.
AI in the Real World: Robotics and the Next Automation Wave
The third vector – the Real World – refers to AI’s expansion beyond data centers into physical environments: robots, autonomous vehicles, and embedded AI systems operating in factories, warehouses, cities, and homes. If cloud AI is about scale and enterprise AI about adoption, AI in the real world is about physical impact – literally bringing intelligent behavior to machines that move through and manipulate our environment. Jensen Huang often speaks of “AI factories” not only as data centers but also as literal factories, where AI-driven robots and digital twin simulations reinvent manufacturing and logistics. This is a nascent but hyper-growth area, poised to redefine entire industries through automation.
In recent remarks, Huang predicted that humanoid robots will be operating in manufacturing plants within five years. This highlights how quickly the line is blurring between the digital and physical domains thanks to AI. Several forces are aligning to accelerate robotics and real-world AI deployment. First, advancements in computer vision, sensors, and edge AI hardware (often derived from technologies built for autonomous cars) are making it feasible to deploy robots that can perceive and navigate complex, unstructured environments. Robots are learning not just to perform repetitive motions, but to understand context – what Huang called “physical AI” able to grasp concepts like friction, force, and object permanence.
Second, the costs of robotics are coming down while capabilities improve. Cloud-connected robots (“cloud robotics”) and AI model sharing mean even smaller companies can leverage collective improvements in robotic intelligence, rather than each having to solve AI problems in isolation.
Moreover, geopolitical and labor market factors are catalyzing automation in the real world. An aging workforce in many countries, along with chronic labor shortages in sectors like logistics and healthcare, is increasing demand for AI-driven machines. Geopolitical frictions are also playing a role: companies are rethinking global supply chains due to trade wars and are looking to onshore more production. Automation becomes a key enabler of this shift – for example, a company might relocate a factory closer to home (to avoid tariffs or export controls) and invest heavily in robotics to keep that factory efficient with higher local labor costs. Trade policies such as U.S. export controls on advanced chips are further prompting nations to strive for self-sufficiency in AI technology. China, restricted from importing the top Nvidia GPUs, is doubling down on its domestic AI chips and industrial automation to remain competitive. In sum, global tensions and national strategies are incentivizing investments in robotics and AI-driven automation, as both corporations and governments see it as critical for economic resilience.
For technology providers and enterprises, the rise of real-world AI means bridging the gap between IT and operations (OT). Companies like NVIDIA are extending their platform to edge devices – e.g., Nvidia’s IGX and Jetson platforms for robots – and providing simulation tools (like Omniverse) to design and train robots in virtual worlds. This end-to-end capability from simulation to deployment is hastening the velocity of robotics innovation. Businesses implementing automation must integrate AI models with reliable hardware and ensure safety, all while calculating ROI. Early use cases (autonomous forklifts, inspection drones, collaborative robotic arms, autonomous vehicles) are proving out, and their successes will lead to broader adoption. Analysts forecast the global robotics market to grow exponentially in this decade, reaching hundreds of billions in value, as AI enables robots to take on more varied tasks.
For decision-makers, the real-world AI vector offers perhaps the longest-term payoff but is essential to monitor now. Leaders in industries like manufacturing, retail, supply chain, energy, and transportation should be piloting AI-driven automation and upskilling their operational teams to work alongside intelligent machines. Just as cloud and enterprise AI will transform data centers, robotics will transform physical workflows. Those who start integrating AI into their operations stand to gain in efficiency and agility; those who ignore it may find themselves outpaced by rivals with smarter, faster production and delivery capabilities.
NVIDIA’s End-to-End Play: Dynamo and the New Wintel?
Underpinning all three AI vectors (Cloud, Enterprise, Real World) is an emerging end-to-end strategy exemplified by NVIDIA. In the PC era, Wintel – the Microsoft-Intel duopoly – defined the dominant computing platform by tightly coupling a software OS (Windows) with ubiquitous hardware (x86 CPUs). Today, NVIDIA is orchestrating a modern equivalent: a full-stack AI computing platform spanning silicon, systems, software, and services. However, unlike the two-headed Wintel, NVIDIA’s approach is more vertically integrated (one company providing most pieces) and targeted at the demands of AI rather than general-purpose computing.
Consider NVIDIA’s portfolio now: it designs the leading AI chips (GPUs like Blackwell, and even DPUs and upcoming CPUs), builds the core systems (DGX and HGX servers, reference designs for OEMs), provides the key software frameworks (CUDA for compute, TensorRT for inference, cuDNN for deep learning, Omniverse for simulation, etc.), and even curates an ecosystem of pretrained models and services (through its AI Enterprise suite and partnerships). The recent introduction of NVIDIA Dynamo is especially noteworthy. Dynamo has been called an AI “operating system” for large-scale inference – essentially a software layer that manages how massive AI models run across many GPUs to serve users efficiently. By offering Dynamo as open-source, NVIDIA aims to make it the standard way to deploy big generative models in production, much as Windows became the standard for PC applications. In Huang’s GTC keynote, he demonstrated Dynamo enabling Blackwell GPU systems to achieve a dramatic 40X performance improvement for certain AI workloads. Dynamo intelligently orchestrates the “prefill” and “decode” phases of inference, optimizing memory usage and compute across a cluster for maximum throughput and responsiveness. This kind of system software is complex and critical – by providing it, NVIDIA further embeds itself into its customers’ AI operations.
The strategic significance of NVIDIA’s end-to-end play cannot be overstated. If successful, it creates a virtuous cycle similar to Wintel’s dominance decades ago: developers target NVIDIA’s platform first (because that’s where the best performance and largest user base is), customers default to NVIDIA-based solutions (because that’s where the software and skills ecosystem is richest), and NVIDIA captures a disproportionate share of value (by selling not just chips but also software subscriptions, services, and perhaps taking a cut of cloud AI consumption through partnerships). There are, of course, key differences from Wintel. One is that AI workloads are often run in the cloud or at the edge, so the platform needs to span diverse environments – NVIDIA’s strategy reflects this by working closely with cloud providers and embedding in edge devices. Another difference is the role of open-source: whereas Windows was closed-source, much of the AI software stack (PyTorch, TensorFlow, etc.) is open. NVIDIA walks a line here by open-sourcing certain tools like Dynamo, but keeping others proprietary (CUDA). This mix of open and closed components is designed to cement NVIDIA’s platform advantages while avoiding alienation of developers who expect open ecosystems. It’s somewhat analogous to Android in mobile – open-source core, but tightly influenced by one company.
Competitors are keenly aware of NVIDIA’s moves. Just as IBM, AMD, and others tried to counter Wintel, we see efforts to forge alternative AI stacks – for instance, some cloud giants are developing their own AI chips (Google TPU, Amazon Trainium) and open-sourcing software (Meta’s PyTorch leadership, OpenAI’s efforts) to prevent a single vendor lock-in. Governments too are wary: any time one company holds the keys to a foundational technology, regulatory attention follows (for example, potential antitrust scrutiny if NVIDIA’s dominance grows). Huang’s genius has been not just in technological foresight but in ecosystem cultivation – making NVIDIA hardware indispensable while enabling others (OEMs, startups, researchers) to build on top. The coming years will determine if NVIDIA’s model truly becomes the “Wintel of AI” in ubiquity. For now, tech executives should assume that NVIDIA’s platform (or an equivalent accelerated computing stack) will be central to AI strategy. Architectural decisions – whether you’re a cloud provider choosing hardware for your AI cloud, or an enterprise choosing an AI platform – must weigh the benefits of NVIDIA’s integrated approach against potential long-term dependencies. The analogy to Wintel serves as a useful strategic lens: controlling the base of the computing stack can yield tremendous leverage in the market. NVIDIA is clearly aiming for that level of influence over the AI era, though in a more networked and open way than the Wintel days.
Geopolitical Friction and the Automation Imperative
Broader forces in geopolitics and policy are acting as both headwinds and tailwinds for the AI revolution. On one hand, trade restrictions and international tensions risk fragmenting the global tech ecosystem; on the other, they are inadvertently spurring faster adoption of AI and automation as nations and companies adapt. Tech leaders must factor these macro variables into their roadmaps, as they can impact supply chains, market access, and the urgency of deploying AI.
A prime example is the U.S.-China technology rivalry. U.S. export controls introduced in 2022 and expanded in 2023 explicitly ban the sale of high-end AI chips to China. For NVIDIA, this cut off what was likely a multi-billion dollar market of Chinese hyperscalers eager to buy its latest GPUs. In the short term, such policies constrain who can access the top-tier AI hardware, potentially slowing down AI progress in affected regions. But in the longer term, these controls have prompted China to accelerate its own semiconductor and AI development efforts. Chinese companies are investing heavily to create domestic GPU alternatives and are also turning to AI-driven automation to reduce reliance on imported technology and foreign labor. In effect, the geopolitical bifurcation is driving two parallel AI ecosystems to develop – each racing to be as self-sufficient as possible. We see, for instance, Chinese firms like Huawei and Biren announcing advanced AI chips built on older process nodes (to circumvent restrictions), and at the same time, Western companies re-shoring critical manufacturing with the help of robotics. Competition is fueling innovation: necessity is pushing both sides to advance AI deployment under new constraints.
Geopolitics also directly threatens the supply chain underpinning AI computing. Taiwan’s semiconductor industry (TSMC) is the linchpin of advanced chip fabrication for NVIDIA and many others. Any instability in the Taiwan Strait could disrupt the supply of cutting-edge AI chips overnight – a risk factor not lost on industry planners. This risk is motivating efforts to diversify chip manufacturing (e.g., TSMC and Intel building fabs in the U.S., Europe’s chip initiatives). In the meantime, companies hedge by stockpiling critical components and exploring architectures that can do more with less (for instance, optimizing software like Dynamo to get maximum performance from available hardware, or investigating analog/quantum accelerators as future alternatives). Additionally, energy security and regulations influence AI infrastructure: Europe’s strict data and AI regulations, and global pushes for energy-efficient computing, shape how and where firms deploy large AI models (for example, favoring local data centers in certain cases, or driving innovation in low-power AI chips).
Another geopolitical aspect is the automation response to deglobalization. As global trade flows face uncertainty, many firms are localizing production. But manufacturing in high-cost regions is only viable with heavy automation – fueling demand for AI in robotics, as noted earlier. Governments themselves are investing in AI: the U.S. CHIPS and Science Act, EU’s digital sovereignty programs, China’s AI 2030 plan, all inject funding and focus into AI R&D, effectively subsidizing faster progress. National strategies see AI leadership as a source of economic and military strength. This can translate into favorable policies (tax breaks for AI investments, expedited approvals for autonomous vehicle pilots, etc.) that encourage adoption. On the flip side, regulatory risk looms: concerns over AI ethics, bias, and job displacement are prompting discussions of new rules that could limit AI use in certain domains. Firms might soon face compliance requirements for AI systems, which could slow deployment or raise costs if not anticipated.
In sum, the geopolitical environment is a double-edged sword for AI. It injects uncertainty and complexity, but it also increases the impetus for automation as organizations hedge against labor volatility and supply chain disruptions. Tech executives should maintain a close watch on policy developments: export controls, data residency laws, AI-specific regulations, and international standards will all influence how you source technology and deploy AI globally. Navigating these waters requires agility – e.g. having multi-source strategies for critical hardware, and ensuring your AI architectures can be modular to swap components if needed. Those who plan proactively can even turn geopolitics into an advantage (for example, capturing market share in regions your competitor cannot serve due to export restrictions, or leveraging government incentives for domestic AI projects). The key is to align automation strategies with the emerging geopolitical reality: in a world of friction, the more you can automate and localize with AI, the more resilient your operations become.
The Clock is Ticking for Incumbents – Risks and Opportunities
As we have outlined, the shift to an AI-first computing model presents a mix of risk and opportunity across the industry. Nowhere is this more stark than for traditional enterprise tech vendors and others whose business models are rooted in the pre-AI architecture. Companies like Dell, HPE, IBM, Oracle, Cisco, even cloud providers that are built on conventional paradigms, all face a moment of truth. The next 18–24 months will likely determine the new winners and losers.
For on-premises hardware and infrastructure vendors, the risk is clear: if the world moves to AI-fueled, highly efficient computing, demand for older-generation gear could plummet. Why buy dozens of standard servers when a handful of AI systems can deliver better outcomes? There is a real possibility of a capital spending rotation away from traditional equipment (general-purpose servers, spinning disks, etc.) toward AI-centric gear. This could pressure revenues and margins of firms not deeply involved in AI. Moreover, hyperscale cloud providers (who are major customers of OEM hardware) might design more of their own hardware (following the lead of AWS designing Nitro, Trainium chips, etc.), eating into the TAM for merchant suppliers. The risk isn’t limited to hardware – software incumbents selling legacy business applications could also see declines if AI-native applications render some legacy functions obsolete.
However, these incumbents also have tremendous assets to leverage if they move decisively. They hold longstanding customer relationships, global support networks, and deep domain knowledge of enterprise needs – factors that upstart AI companies often lack. By pivoting to offer AI solutions (either developed in-house or via partnerships/acquisitions), incumbents can become the trusted guides for their customers’ AI journey. For instance, consultative services and integration will be in high demand as enterprises adopt AI – companies like IBM and Accenture are already positioning around AI consulting for this reason. Hardware firms can embed AI in their products (smart storage that does on-board AI processing, networking gear optimized for AI data flows) to differentiate beyond commodity features. Hybrid cloud offerings are another play: many enterprises will seek a seamless experience between their data center and public cloud, and vendors like Dell and HPE are leaning into this with on-prem cloud stacks that include AI capabilities, ensuring they remain central in a hybrid world.
One notable opportunity area is at the edge and industry-specific AI. While hyperscalers dominate central cloud AI, there is a burgeoning market for AI in verticals like telecom (e.g. AI in 5G networks), healthcare (diagnostic AI), finance (fraud detection), etc., often requiring tailored solutions. Traditional enterprise providers, with their vertical sales teams and custom solution experience, can capture these niches by packaging AI into offerings that solve specific business problems. We are also seeing incumbents invest in AI software platforms (for example, offering management tools for AI models, or marketplaces for AI solutions) which can create new recurring revenue streams beyond selling hardware.
In essence, incumbents must reinvent themselves around AI – much as IBM transitioned from mainframes to services and software in past eras, or how Microsoft successfully pivoted to cloud. Those who do not cannibalize parts of their own legacy in favor of new AI-driven offerings may find the market doing it for them. The clock is ticking: customers are already evaluating how partners and suppliers can help them with AI. In the next two years, enterprises will solidify their AI infrastructure choices, and vendor mindshare is being decided right now. Every CEO and board of a tech vendor should be asking: Do we have a bold AI strategy that matches the magnitude of this shift? Are we investing enough, acquiring talent and technology, and retraining our salesforce to tell an AI story? The opportunity is to ride the biggest IT wave of our generation; the risk is to be left behind as spending reallocates. For many incumbents, it truly is an adapt-or-decline moment.
Conclusion: Embracing the 3Vs – A Call to Action for Tech Leaders
The evidence is overwhelming that a once-in-a-generation architectural transition is upon us. AI-driven computing – accelerated, intelligent, and ubiquitous – is the new foundation on which business value will be built. This transition brings unprecedented Volume, Value, and Velocity. The volume of data and compute power is skyrocketing, the value potential for those who harness AI is immense, and the velocity of change leaves little time for hesitation. Technology leaders must act with urgency and vision.
Enterprise architects should reevaluate every layer of their stack: Can our silicon and infrastructure handle the coming volume of AI workloads? Are we positioned to capture the value from AI by reallocating resources and focusing on high-ROI use cases? And are we moving with the velocity required – fostering a culture of agile experimentation, rapid scaling of successful pilots, and swift decommissioning of outdated tech? The 3Vs provide a guiding framework in navigating this disruption. They remind us to scale up ambitiously, focus on business outcomes, and accelerate our timelines.
It’s worth recalling that previous computing revolutions rewarded those who anticipated and embraced the change. In the 1990s, companies that quickly adopted the Wintel PC model outpaced those stuck on minis and mainframes. In the 2000s and 2010s, those that moved to cloud and mobile early gained competitive advantage. Today’s wave is bigger – and moving faster – than any before. Senior technology executives, from CIOs to CTOs and CEOs, need to champion bold action. This could mean green-lighting significant AI infrastructure investments even amidst budget constraints, doubling down on upskilling your workforce in data science and AI engineering, and reimagining business processes through the lens of AI and automation.
We are at the start of what will likely be a decade-long journey of architectural change. But decisions made in the next 18–24 months will shape the trajectory of enterprises for years to come. Now is the time to map your strategy in “Jensen’s world” of AI – not as it exists today, but as it will rapidly evolve across cloud, enterprise, and robotics domains. Share your perspective, challenge your teams to think beyond incremental improvements, and be ready to disrupt your own legacy practices. The organizations that thrive in this new era will be those that internalize the 3Vs – scaling to new volumes, seizing value creation opportunities, and moving with decisive velocity. The future of IT architecture is being written now. It’s a future where AI is integral to every system and service. As a technology leader, the call to action is clear: engage, adapt, and lead your enterprise through this historic transformation. The time to prepare your AI infrastructure and strategy is today – tomorrow will be too late. Let’s embrace the disruption, and in doing so, secure our place in the next chapter of computing.
