There is massive innovation and focus on GPU-based computing right now. It is fascinating—and frankly crazy impressive to watch the computing discussions and spending shift so quickly to AI. (Remember modern AI only hit the public stage in late 2022). But as someone who has lived through decades of “normal” enterprise compute cycles and now participates in this AI market actively, I regularly think about the innovation pendulum swinging so quickly towards something new that it overshadows the rest of the foundational “real work” that still needs to be planned and executed. These pendulum swings happened with Cloud, Wireless, Virtualization, the Internet, and even PC-LANs.
A quick survey of recent RFPs, press releases, and column inches confirms that many of the ‘buyers’ in the industry have formed an “all-in” mental pivot towards GPU-centric systems. Every discussion seems to be centered or driven by AI. The concern is that much of today’s existing business apps simply live on x86 or equiv. Take SAP, Oracle, Microsoft, Sage for example. In fact x86 chips alone represented more than $75Billion in 2025 so clearly the non-GPU core of IT is still alive and well.
Don’t get me wrong. The frenzy associated with AI buildouts and the new opportunity unveiled through the use of AI is humbling, exciting, and worthy. It is truly inspiring to think of how far and fast we have moved in such a short period of time. AI has changed the way we think about problem solving, and it allows IT leaders to dream like never before. But its only part of the IT story, not the replacement for most of what we do today- its actually the enabler for what many of us have only dreamed of in the past.
While a relatively small number of hyperscalers, neoclouds, and first mover enterprises are driving the bulk of GPU spending, general-purpose commercial x86 systems still run the vast majority of mission-critical workloads for hundreds of thousands of organizations worldwide—ERP, databases, transactional systems, middleware, and boutique enterprise applications. And while AI is perfectly suited for ingesting tons of data and deriving actionable results in near-real-time, it becomes clear that AI is an additive to the mix, not a replacement for most existing functions. Workhorse x86 and HPC architectures deliver answers to today’s challenges with ease and precision. When a financial institute is setting policy rates or sending out billing for customers, Enterprises rely on those x86-based systems. That said, those x86-based systems are perfect candidates to be *augmented* using AI. AI does “new” work rather than replacing “old” work.
GPUs Are Winning the Mind-Share Game… for now
There’s no debate that GPUs deserve attention. It’s incredible to see what 1000watts of silicon can accomplish. Core technologies from Nvidia, AMD and others have redefined what a ‘processing core’ is, and companies like HPE and DELL incarnate those technologies into AI Factories which can be easily deployed at scale. Make no mistake, the GPU innovation already seen is re-defining the systems that fill data centers from the ground up—servers, racks, power delivery, cooling, and interconnection. Software stacks are evolving at breakneck speed, Entirely new operational processes are being written and ecosystems of vendors are aligning roadmaps around AI training, inference, automation and workflows. And nearly every existing SaaS provider has introduced their own AI-driven add-ons (and most call them “Co-Pilots”) to the data they already have access to within their own traditional suites of software solutions. Suites like Oracle and Salesforce have tons of data, but for decades have struggled to allow natural language access to it. This is an ideal use of the GPU.
In contrast, traditional x86 ecosystems anchored by Intel and AMD are increasingly framed as “legacy” platforms- even though they deliver the majority of work processing that we do today and tomorrow. And while they are still the workhorse for business to operate and grow, the world of x86 is no longer the “exciting” topic at the CIO’s strategic table. Yet, Global 2000 business runs on x86, and science and research runs on HPC.
The most strategic IT professionals understand that the GPU opens a whole new world of outcomes which have been sought for decades, but the core foundation of business work will still be the x86 (or equiv CPU running LINUX the majority of the time).
That framing matters. Because “excitement” is where engineering resources, capital investment, and executive focus go. Look at the numbers of investors that are throwing A TRILLION DOLLAR into anything that sounds like AI. That influences or distracts the Global 2000 CIOs and their teams.
Why This Matters More Than AI Headlines Suggest
AI workloads are transformative—but they are not replacing the enterprise application stack anytime soon. Most organizations are not ready to rip-and-replace their data center and all of its applications with AI- nor could they! AI is a great new solution to a set of ‘new’ problems that were always dreamed about but could never be spoken. Traditional x86 answers those problems that have been “asked and answered” for decades. Today’s most innovative buyers will quickly realize that their destinies are hybrid environments where high-profile center-stage GPU-based systems sit alongside vast fields of x86 servers doing the unglamorous traditional work that keeps businesses running.
If innovation in x86 deployments stalls due to focus on GPUs, enterprises will be constrained, and they will suffer. They increase their technical debt, operational risk, and forced tradeoffs between stability and modernization. Costs rise, and core capabilities stagnate.
A Call for Balance
The acceleration revolution is real and necessary. But it is an argument against starving general-purpose compute in the process. x86, HPC and AI Factories must all co-exist and each are a strategic part of the IT landscape as far out as the eye can see.
Just as with networking, the industry needs to remember that “boring” infrastructure is often the most critical. Enterprises need x86 platforms and the core software that runs upin it—even as GPUs dominate the headlines.
Because when the spotlight moves on from today’s AI gold rush, it will still be x86 systems quietly carrying the weight of global enterprise IT.
Leave a Reply