AI PC Ambitions Face an Unforgiving Reality of Memory Constraints and Budget Pressure

For the PC industry, 2025 was the year that the end of Windows 10 support would drive a massive PC refresh cycle. As part of this refresh, AI PCs, devices with neural processing units (NPUs) designed to execute AI and machine learning tasks, were expected to rise in popularity as businesses prepared their workforce for a new era of AI-infused productivity. The PC refresh cycle of 2025 did materialize to some extent, with the market growing about 5.3% year-to-year according to TBR’s estimates, but sellers ran into a snag in the commercial PC market: Many buyers were unwilling to pay 10% to 20% more for an AI PC that could future-proof their PC fleet but offer minimal added value today. This sentiment has driven ecosystem players such as Intel and Microsoft to invest in making the AI PC more compelling to and affordable for the commercial PC market. The beginning of 2026 ushers in the latest generation of AI PC processors from Intel, AMD and Qualcomm that is poised to attract customers with performance, efficiency and security enhancements. In many ways, 2026 is a critical year for the success of the AI PC due to the significant investment in the platforms that have been years in the making.
 
When vendors made these long-term investments in next-generation AI PCs, no one knew that these new models would launch just as insatiable demand in the AI server market caused an unprecedented shortage of memory. The shortage, which in TBR’s opinion could easily last for two to three years, has already caused significant price hikes in the PC market — with more on the way — and will result in longer lead times and lower availability. As such, PC market leaders’ desire to usher in a new era of more expensive AI computing now sits in direct odds with unavoidable component price increases and tight customer budgets.
 

Pressure to refresh drove 2025 growth, with the intrinsic value of AI PCs yet to be defined

In 2025 AI PCs capable of meeting Microsoft’s Copilot+ requirements began to hit the market in earnest, following their initial debut in the summer of 2024. On paper, the timing looked favorable: A Windows 11 refresh cycle was underway, and AI PCs were positioned to ride that momentum and steadily grow share within the broader commercial PC market. In practice, however, the refresh materialized unevenly and ultimately fell short of expectations. Although the market experienced solid growth, expanding an estimated 5.3% year-to-year in 2025, multiple structural and behavioral hurdles slowed adoption of AI PCs, preventing vendors from translating technical readiness into meaningful volume growth.
 
Pricing pressures emerged as a central inhibitor. Tariffs complicated OEM pricing strategies, and the incremental cost of NPU-enabled processors introduced a noticeable price increase that many commercial buyers were unprepared to absorb. This sticker shock was exacerbated by the lack of a compelling near-term value proposition for AI PCs. Vendors leaned heavily on the concept of future-proofing device fleets, a message that resonated poorly with budget-constrained IT buyers prioritizing immediate, tangible ROI. At the same time, commercial purchasing behavior has yet to normalize post-COVID. Some organizations are still digesting inventory acquired during earlier buying surges while others have permanently shifted from predictable refresh cycles to more opportunistic, as-needed purchases.
 
In response, OEMs and key ecosystem partners — notably Microsoft, Intel and AMD — adjusted their go-to-market strategies to reduce friction and keep AI PCs moving into commercial environments. OEMs refined system configurations where possible to lower bill-of-materials costs without undermining Copilot+ eligibility. Silicon providers increased front-end discounting, rebates and comarketing investments to narrow the price gap between traditional PCs and AI-capable systems. Together, these efforts drove down pricing of AI PCs in the commercial space and helped some customers make the transition to NPU-enabled silicon during their device refreshes.

Ecosystem investments years in the making underpin the future success of AI PCs

Despite increasing competition from AMD and Qualcomm, Intel remains the PC silicon market share leader due in large part to the robustness of the company’s partner and incentive programs. PC OEMs and the broader industry supplying PCs rely heavily on Intel to influence market direction and demand. Understanding that the AI PC market is somewhat nebulous due to ill-defined use cases, Intel has invested significantly to make its latest chips — Core Ultra Series 3 — a success, leaning into its heritage as an integrated device manufacturer and shifting manufacturing of the compute tile on the new system on a chip (SoC) in-house rather than outsourcing to Taiwan Semiconductor Manufacturing Co. as it did with the Core Ultra Series 2. Although the investment to bring 18A capacity online has been massive, in the long run Intel could benefit from a cost of sales perspective with the manufacturing based in the U.S. and improved yields as the process matures. In contrast, as fabless companies, AMD and Qualcomm rely solely on third-party manufacturing, which presents fewer risks but could prove more costly in the long run.
 
Silicon vendors’ investments in driving AI PC adoption have looked similar from company to company, with AMD, Intel and Qualcomm providing developer resources such as samples, guidance and reference applications to drive NPU-enabled application development. However, although AMD’s and Qualcomm’s resources are limited to software developer kits and other software stacks directly targeting the NPU, Intel’s play also leans into its reputation as a market maker with its AI PC Acceleration Program, incentivizing collaboration between ISVs and independent hardware vendors (IHVs) to create features that leverage the NPU. Additionally, with the engineering staff available through Intel Open Labs serving as a differentiator, Intel arguably provides more design resources, sales support and go-to-market funding than AMD and Qualcomm, echoing the company’s long-running strategy that has positioned it atop the Windows PC silicon market. Similar to what silicon vendors are providing, Microsoft has launched tools that will help ISVs refine their Windows-based apps to make the best use of CPU, GPU and NPU.
 
Qualcomm, more so than Intel and AMD, needs to prove itself in the commercial market and demonstrate not only that Arm can outperform x86 in commercial uses but also that the platform is a trustworthy investment. However, much of this weight falls on Microsoft’s shoulders as Windows on Arm is inherently its own initiative. Perhaps more important than its advances in performance in TOPS (trillions of operations per second), Qualcomm’s newly launched Snapdragon X2 Plus platform offers a competitive answer to Intel’s vPro platform with Snapdragon Guardian, which allows for advanced remote management and security features. However, Intel vPro’s Active Management Technology, which allows for out-of-band remote management, remains core to the company’s value proposition, is still unmatched by Qualcomm and AMD, and is a key reason noncloud-first organizations tend to choose Intel over its competition.

2025 Windows AI PC SoC Revenue by Covered Vendor [Intel, AMD, Qualcomm] (Source: TBR)

Massive scale AI infrastructure deployments are driving unprecedented fluctuations in memory costs, negatively impacting supply for the PC market

The PC market has weathered many component shortages that have had short-term impacts on product availability and pricing. In recent history, shortages have been caused by factors ranging from pandemic-induced buying changes and natural disasters to manufacturing industry consolidation and delays in next-generation technology rollouts. However, the current memory component shortage is unique due to its ties to structural AI demand versus a temporary shock to the market. Over the past two years, demand for AI infrastructure has shifted the slow-growing server market into hypergrowth, with TBR estimating that server spend nearly doubled from 2023 to 2025. Not only has the demand risen dramatically, but the AI server silicon, whether GPUs or AI ASICs, also requires significant amounts of high-bandwidth memory (HBM), unlike their traditional server counterparts.
 
As such, hyperscalers and neoclouds are absorbing a disproportionate share of DRAM and HBM as they scale AI training and inference infrastructure, and the same suppliers sell both HBM and conventional DRAM. As these memory suppliers prioritize higher-margin HBM, commodity DRAM output growth tightens as investments in new capacity are allocated more heavily toward supporting AI infrastructure rather than client devices.
 
This trend of elevated server demand combined with disproportionate consumption of high-performance memory will not abate anytime soon. OEMs’ total server order backlogs remain in the billions exiting 2025, and during AMD’s 4Q25 earnings call CEO Lisa Su reported the company expanded manufacturing capacity for server CPUs, confirming that component suppliers continue to shift their manufacturing lines in favor of servers rather than PCs.
 
As a result, PC DRAM supply has not grown enough to support increasing demand. The DRAM found in thin-and-light PCs is the same DRAM found in most smartphones, and the replaceable DRAM typically found in desktops and workstations is also used in traditional enterprise servers. Due to this supply-and-demand imbalance, prices for PC DRAM have skyrocketed and are likely to remain elevated through at least 2026 and into 2027. Already, contract prices for the DRAM found in desktops and workstations have increased between 10% and 30% year-to-year going into 2026 with the DRAM found in thin-and-light PCs increasing at an estimated rate of between 20% and 40%.
 

OEM Server Revenue in $USD Billions for 2020 through 2025 (Source: TBR)

Memory costs and longer lead times put AI PC momentum at risk

TBR expects the memory shortage will impact commercial PC buyers in three ways: pricing, availability and product assortment. Channel partners in contact with TBR have reported PC OEM prices have increased twice since the beginning of 2026, with more price hikes planned. Although price increases to date have been estimated between 10% and 20%, a partner noted a customer whose specific PC configuration price had increased by 40% since December 2025. Channel partners have also noted that delivery times have already become longer, meaning customers will have to accept longer lead times going forward.
 
These dynamics put AI PC models with higher-priced silicon in a precarious position. Customers with finite spend are already facing significant price increases on memory that offers no additional value. Many customers will face a choice of further delaying PC refresh, reallocating budget from other IT projects, or seeking watered-down configurations to reduce costs.
 
OEMs will need to maintain tight cost controls as much as possible through supply chain management and focus on product assortment. For example, OEMs will likely push more models with 16GB of RAM, which is the minimum requirement for Copilot+ PCs, over those with 32GB or 64GB. Ultimately, with longer lead times and a tight product assortment a likely outcome, TBR expects OEMs will focus on configurations and segments where they can maintain margins and best serve the customers that are willing to pay.
 
The cost burden will not fall solely on the OEMs. With all three AI PC silicon vendors launching new chips at CES in January, these vendors are motivated to make the new platforms a success. OEMs and silicon vendors already had to adjust pricing on NPU-enabled silicon in 2025 to reach a point where customers were willing to upgrade. TBR expects OEMs will pressure Intel, AMD and Qualcomm to provide additional discounting to ensure the success of the latest generation of chips.

AI PC Revenue by Big Three Windows OEMs [Dell Technologies, HP Inc., Lenovo] for 2025 and 2026 (Source: TBR)

2026 outlook: Increasing prices will stifle overall market growth and slow AI PC mix shift

Despite significant pricing impacts, TBR expects the PC market will grow in 2026. In fact, the early part of the year may perform better than expected if customers pull in buying to avoid additional price increases coming down the road.
 
Although the PC ecosystem has been investing in making NPU-enabled silicon more useful, the use cases for most AI PC buyers remain thin and focused on future innovation. The AI PC mix will continue to increase, but TBR believes the growth rate will be lower than we predicted in 2025. The increase in mix will be based more on the availability of chips and manufacturers’ emphasis on AI PCs than on organic customer demand.
 
Although TBR estimates the overall value of the PC market will grow 1.7% year-to-year in 2026, the increase will be driven by rising selling prices, not shipments, which are expected to decline as many buyers delay PC refreshes unless absolutely necessary. Meanwhile, TBR forecasts the overall value of the AI PC market will expand 33.4% in 2026 due in large part to a relatively small compare and PC OEMs’ (excluding Apple) growing AI PC revenue mix, which TBR predicts will increase from 23.9% in 2025 to 37.9% in 2026, a difference of 1,400 basis points. Although this mix shift is pronounced, TBR firmly believes that had PC DRAM prices remained the same, the increase would be more robust.
 

TBR’s AI PC Market Forecast for 2025 through 2030 (Source: TBR)

Conclusion

Rising demand for the types of DRAM found in PCs, in combination with memory suppliers focusing production capacity build-out on supporting massive-scale AI infrastructure deployments, has led to a supply-and-demand imbalance in the market, causing memory manufacturers and suppliers to rapidly increase PC DRAM prices. With margins already thin, PC OEMs are passing their rising build costs on to buyers in an attempt to protect profitability. AI PC prices were already higher than traditional PC prices, and with AI PCs typically configured with more memory, on average, AI PC prices are poised to continue increasing more than traditional PC prices, making the future-proofing-led sales proposition less compelling. Not only will these rising prices result in slower-than-anticipated adoption of AI PCs, but they will also cause more organizations to delay their PC refreshes. However, once PC DRAM prices stabilize and begin to fall, pent-up demand from delayed refreshes could yield a prosperous cycle for PC OEMs and silicon vendors alike.
 
Fortunately, although the rate at which memory prices are increasing is unprecedented, PC OEMs have weathered component shortages in the past. While increasing PC prices will slow the market’s transition to AI PCs in the near term, the proliferation of the AI PC over the next five years is almost guaranteed due to the alignment of ecosystem players, including PC silicon vendors, PC OEMs and Microsoft, which owns the operating system. As such, increasing memory costs and PC prices are less of a roadblock to AI PC adoption and more a speed bump impacting the momentum of adoption.

A strategic fit that can test innovation and outcome delivery readiness

KPMG’s alliance with Salesforce has moved from a high-growth implementation practice into a relationship increasingly defined by enterprise trust, measurable outcomes and the ability to operationalize agentic AI. Since entering Salesforce’s ecosystem in late 2019, KPMG has scaled the alliance to over 1,300 practitioners across more than 30 countries and is now repositioning the relationship around Agentforce-led transformation, AI-ready data foundations, and run/optimize operating models that sustain adoption. This shift mirrors a broader ecosystem trend: Leading platforms are prioritizing depth with a smaller set of preferred partners and evaluating alliances on their ability to drive usage, value realization and governance — not simply project throughput.
 
On Jan. 20, TBR spoke with KPMG’s Matthew Fidler, Global Salesforce Platform leader, U.S. Salesforce Platform & Alliance leader; and Andrew Dunn, Global Salesforce Alliance director, about the evolving relationship between KPMG and Salesforce. The following analysis reflects on this discussion and TBR’s ongoing research around KPMG and the Big Four, including our semiannual Management Consulting Benchmark and ecosystem intelligence reports.
 

Client stickiness will depend on KPMG’s ability to demonstrate depth around Salesforce capabilities

KPMG entered the Salesforce partner ecosystem at the end of 2019 and quickly scaled the relationship through a combination of organic capability build-outs and targeted inorganic investments. The alliance has been characterized by aggressive certification growth, expansion into core CRM clouds and industry clouds, and an increasing focus on AI-related capabilities. KPMG also extended its footprint in EMEA through the acquisition of iCom in France to accelerate regional delivery capacity.
 
From an ecosystem standpoint, KPMG’s positioning emphasizes depth over breadth. Rather than pursuing every Salesforce product adjacency, the firm prioritizes sectors where its brand and domain expertise offer an advantage (as confirmed by Salesforce feedback), especially in financial services, healthcare and life sciences, and the public sector. In addition, KPMG frames Salesforce programs as enterprise transformations, not stand-alone technology implementations, a setup that is not unique to KPMG but gives the firm an opportunity to act as more of a business integrator.
 
According to TBR’s October 2025 Adobe and Salesforce Ecosystem report, partner success is increasingly anchored in trust and clarity as platforms are narrowing their focus on fewer strategic partners while expecting those partners to lead with a unified narrative and deliver measurable outcomes. Two implications of this trend matter most for the KPMG-Salesforce partnership.
 
First, generative AI (GenAI) and agentic delivery are creating a managed services battleground. Buyers are less willing to treat AI as a phase two add-on; they expect continuous optimization, governance and data foundation work to be integrated into the program from day one. This behavior favors partners that can industrialize repeatable assets (agent templates, reference architectures and data-readiness checklists) and run them operationally after the go-live.
 
Second, misalignment between vendor and partner messaging is becoming a risk. Inconsistent positioning across the field confuses clients and erodes confidence in the ecosystem. For KPMG and Salesforce, this elevates the importance of two-way enablement: shared account planning, consistent value narratives, and clear role delineation for strategy, implementation and post-go-live operations.
 
In short, the partnership’s differentiation will be judged less by breadth of services and more by whether KPMG can act as an outcome operator for Salesforce’s Agentforce and data platform agenda — improving adoption, consumption and governance.

Salesforce’s Agentforce provides a new path for KPMG to grow the business but only if commercials are aligned

Salesforce’s Agentforce push is reducing the time between platform adoption and measurable business outcomes, but it is also raising the bar for data readiness, integration and governance. The hard part of CRM transformation is both upstream (data harmonization, integration, identity, policy) and downstream (change management, adoption analytics, continuous optimization). For Salesforce, the upstream challenge is particularly relevant. Years of expansion into adjacent workflows, via both acquisition and internal development, along with less focus on tying the whole portfolio together, have made data harmonization a serious barrier. The acquisition of Informatica was driven in part by the need to offer a native platform that could overcome this hurdle, but Salesforce has not necessarily solved the upstream challenge yet. Given AI’s dependence on quality data and cross-workflow capability, this will be a critical obstacle Salesforce must overcome as it works to scale Agentforce adoption. In this environment, partners must treat AI as an operating model program, not a feature deployment.
 
In parallel, Salesforce’s partner model is shifting toward consumption-led measurements and AI-era enterprise license constructs, increasing the importance of adoption and sustained usage. This will create a commercial and governance inflection point. GenAI improves transparency around productivity and cycle-time benefits. At the same time, customers are increasingly questioning traditional time-and-materials billing models for agent-led work. For KPMG, alliance success will increasingly depend on pairing implementation excellence with credible activation motions (value instrumentation, managed services and consumption-aligned commercial models) and tight joint enablement so the field sellers tell one consistent story.

Partner-told use cases pave the way for future right-to-win areas

To build stakeholder trust, KPMG has begun translating Agentforce into client stories and Salesforce use cases as the firm looks to win new buyer personas. For example, a joint Agentforce-driven onboarding approach for a healthcare client compressed a process from 17 days to under five minutes on average, with the majority of cases requiring zero human touch. The use case is notable because it couples orchestration with risk and decision logic and demonstrates rapid time to value. Additionally, KPMG’s work with a global consumer brand illustrates how Agentforce can augment contact center operations, including scoring 100% of interactions against quality rubrics, automating returns workflows and reducing seasonal labor needs during peak periods. KPMG is also acting as an SI to Salesforce as the latter applies agents to its security review and controls workflows, reducing review cycle time from hours to minutes. This reinforces KPMG’s credibility in regulated workflows and creates reusable assets for clients. Last, following Salesforce’s acquisition of Informatica, KPMG reorganized internally to align Salesforce, MuleSoft and Informatica under one practice and is seeing increased field attention for master data management and data foundation projects — conditions that are increasingly prerequisites for reliable agent deployments.
 
Two of the aforementioned use cases stand out: KPMG acting as an SI for Salesforce; and KPMG’s internal reorganization to align its Salesforce, MuleSoft and Informatica capabilities. The first example reflects a customer-zero approach, which is among the most widely used approaches for consultancies to demonstrate value and trust and scale through the 360-degree relationship. Although this method is not unique to KPMG, it resonates with clients seeking reassurance that third-party providers trust each other.
 
The second example extends beyond operational efficiencies as realignment allows for a better multiparty alliance construct, which can lead to higher win rates. In our special report Informatica’s Alliance Strategy: Powering GSIs, Scaling AI and Strengthening the Data Ecosystem TBR analysts spoke with Richard Ganley, senior vice president, Global Partners, who shared the following: “We looked at basically all the opportunities that we’d had in our system, which we’d either won or we’d lost over the past two years. And we found if we didn’t work with a partner, our win rate was around 17%. If we worked with one partner, it went up to 47%, which kind of makes sense because we’ve got somebody in there speaking up for us, recommending us. But if we worked with two partners, and by two we mean one from the GSI and one from the ecosystem … the win rate goes up to 83%.” With Informatica among KPMG’s top partners, KPMG has an opportunity to build a deeper moat within the data management space, which we believe will remain the battleground for client share in the short to midterm.
 
As enterprise AI projects grow more complex and interdependent, multiparty alliances are becoming a preferred delivery mode. Salesforce is responding by forming deeper relationships with hyperscalers and global systems integrators, and by shifting from referral-based partnerships toward integrated frameworks that combine infrastructure, data and services into unified operating motions. This creates an opportunity for KPMG to build reusable components and industry-specific agent templates while also taking responsibility for trust, risk management and compliance to align its data and AI strategy with Salesforce’s and to make  multiparty alliances more natively integrated and programmatic.

Doubling down on agentic AI will test KPMG’s culture and readiness to transform its operating and commercial model

As the firm looks to secure its long-term play given the rapidly evolving nature of agentic AI, in the past year KPMG has accelerated its portfolio and go-to-market transformation by embedding AI across internal operations and client services, fueled by a large amount of capex investment from within KPMG’s partner ecosystem. We believe deepening its relationship with Salesforce and integrating Agentforce into the firm’s daily operations will enable consultants to accelerate speed to value. The use of Agentforce in an era led by Microsoft Copilot and OpenAI hints at the rise of hybrid AI and how vendors are preparing for buyers to use several AI platforms rather than settling for just one.
 
We believe KPMG’s trust, risk, compliance and industry operating-model capabilities map well to these requirements, particularly in regulated environments. The firm’s cocreation mechanisms (incubator-style engagements and Ignition Center experiences) can shorten the time from pilot to production, but differentiation will depend on repeatability and operationalization: packaged agent patterns, reference architectures and managed services that demonstrate sustained KPI movement. Equally important, the partnership should treat messaging alignment as an operational discipline — shared plays, account governance and value narratives that are consistent across Salesforce sellers, KPMG teams and client stakeholders. Last, using new commercial models where KPMG can leverage Salesforce-ready solutions as a loss leader to get paid on a per-usage basis can help the firm demonstrate pricing agility and expand its addressable opportunities with more price-sensitive midsize clients.
 
According to TBR’s October 2025 Adobe and Salesforce Ecosystem report: “As both tech and services partners look to diversify sales channels, the opportunity SMB clients present will lead to new alliance partner relationships, especially with startups and/or specialized vendors, thus challenging the status quo. Services vendors have an opportunity to maintain trust by using tech vendors’ marketplaces to drive portfolio awareness. Tech vendors need to maintain commercial agility as SIs are getting smarter about building their own off-the-shelf solutions and focusing on contextualization.”
 
KPMG’s ambition to establish a stronger foothold in the GenAI space — and to use the technology to deepen client relationships — could, in time, place an additional strain on the firm’s operating model. Building, selling and managing software-like solutions introduces new requirements around talent, product management and pricing that differ from traditional consulting motions. That said, KPMG appears to be taking steps to address some of these potential challenges, which may help reduce internal friction as the shift takes hold. In parallel, the growing appeal of managed services may continue to offer a relatively straightforward path to near-term revenue, but it does not necessarily have to come at the expense of GenAI momentum. If KPMG can align incentives and capacity planning with its GenAI priorities, partners may be less likely to default to “quicker win” engagements — and the firm may be better positioned to reinforce, rather than dilute, its consulting value proposition.

Accelerating stakeholder trust remains KPMG’s top priority as the firm leans on Salesforce amid ongoing market challenges

Deepening relationships with strategic alliance partners such as Salesforce will bolster KPMG’s trust within the ecosystem, helping it increase client stickiness as clients more often look for depth than scale in the AI era. Although KPMG has a tall order in business model transformation, the firm is not alone. In the next two years, AI-ready platforms such as KPMG Velocity and KPMG Workbench, backed by common governance and an ongoing commitment to funding and positioning with clients, will accelerate KPMG’s efforts to move beyond the traditional time-and-materials operating model. Although it can be challenging to address change at scale, getting the right framework in place, including establishing common KPIs and deeper collaboration with Salesforce, will give KPMG the necessary building blocks to move away from its historically risk-averse culture rooted in audit and assurance.
 
The best way for KPMG to differentiate will be to package reusable assets, generate outcomes (adoption and consumption) and establish joint field enablement so the KPMG-Salesforce story is consistent from boardroom strategy through post-go-live operations. If KPMG executes on all of these efforts, the firm can become one of Salesforce’s most strategic partners in AI- and data-intensive sectors, helping move deals from “CRM projects” to enterprise transformations measured in data leverage, AI impact, and realized business outcomes and, in turn, driving durable platform consumption.