The Road To HPC And AI Profits Is Paved With Good Intentions

With a profitable PC business that has 25 percent of global shipments (thanks in large part to its acquisition of IBM’s PC business two decades ago) plus a respectable smartphone business (by virtue of its Motorola acquisition), the client device business at Lenovo is finally back to where it was during the peak of the coronavirus pandemic and is consistently delivering what are decent profits for this cut-throat part of the IT sector.

We don’t normally care about such things in The Next Platform, except when a client device business puts a datacenter business under pressure or, conversely, when it relieves some pressure that is on a datacenter business. At Lenovo, the latter, not the former, is true.

In the quarter ended in September, which is the second quarter of Lenovo’s fiscal 2026 year, the company’s Intelligent Devices Group (IDG) posted sales of $15.11 billion, up 11.8 percent, and operating profits were $1.1 billion, up 11.2 percent and comprising 7.3 percent of revenues. That may not seem like a lot of profits, but it is a heck of a lot more profitable than the Infrastructure Solutions Group (ISG), which over the past ten and a half years has amassed over $84 billion in sales and only has a combined $1.95 billion in losses to show for it from sales of server, storage, and switching gear.

To be fair, the company’s Solutions & Services Group (SSG) is adjacent to the ISG and IDG slices of Lenovo and has been growing pretty fast and delivering much healthier profits due to services engagements, product tech support, and on-premises equipment rentals through cloud-like offerings. SSG had $2.56 billion in sales, up 18.1 percent year on year and delivering $571 million in operating income. That is 22.3 percent of revenues.

ISG should be so lucky. In the second quarter, ISG had $4.09 billion in revenues, up a very respectable 23.7 percent compared to Q2 F2025, but the datacenter group at Lenovo had a $32 million operating loss.

The cloud service providers – what we call hyperscalers, cloud builders, neoclouds, and large service providers – are clearly to blame. Lenovo’s top brass said sales of ISG gear to CSPs rose by 21 percent in the quarter – our model says 20.8 percent, to $2.76 billion, and we think that these CSP deals had a net loss of $349 million. (Lenovo gave out charts that let you figure out the CSP and ESMB revenue split precisely before Q4 F2023, and we have been using hints from Lenovo’s Wall Street calls to extend our model since that time.)

Sales of gear to enterprises and small and medium businesses (ESMB in the Lenovo lingo) were up 30 percent, according to the company, and that works out to $1.33 billion in our model and, more importantly, we think this part of the business had $317 million in operating income, almost pulling the whole of ISG out of the red ink.

Add it all up, and Lenovo had $20.45 billion in sales, up 14.6 percent year on year, with operating income up 17.4 percent to $1.64 billion. After paying taxes and other costs, Lenovo booked net income of $340 million, down 5.2 percent.

We think that AI is a double-edged sword for Lenovo when it comes to profits, as it is for all original equipment manufacturers (OEMs) as well as rival original design manufacturers (ODMs) who also have trouble turning a buck on AI systems, despite their high costs and incredible demand.

On a call with Wall Street analysts, Yang Yuanqing, chief executive officer at Lenovo, said that AI drove 30 percent of the company’s revenues in the quarter, up from 17 percent of revenues in Q2 F2025. This is the first time we can remember that Lenovo gave out such an explicit number, and if you do the math, this means AI-related sales for Lenovo more than doubled, from $3.04 billion in Q2 F2025 to $6.14 billion in Q2 F2026.

It is hard to know what share of this AI revenues was for services, support, and software, but we had a model for AI-accelerated server sales already (again, because Lenovo used to talk about it more in the past), so we have been able to extend that with a reasonable amount of confidence. We do not know, however, what the pipeline for AI system sales looks like, and Lenovo has not given a hard figure for its AI pipeline for a year and a half. Our guess is that Lenovo has maybe $15 billion in its pipeline, which is pretty good considering that even with “high double digit” year on year growth, this time last year was not a particularly booming quarter for AI server sales for Lenovo and it only accounted for around $1.34 billion in Q2 F2026, by our estimates, which is 82 percent growth but which is also a 39.2 percent decline from the $2.20 billion in AI system sales in Q1 F2026 and even smaller still compared to the $2.51 billion in AI server sales in Q4 F2025.

AI is the new HPC, and it is just as choppy of a business as the old HPC business ever was.

You have to want to be in any HPC business, and you have to be a hopeful sort that can delay profit gratification – perhaps indefinitely. And it is a good thing that companies like Lenovo can do that, because someone has to build and support AI and HPC machinery despite the low to non-existent profits.

One last thing. As part of its Q2 F2026 financial report, Lenovo showed the following historical market sizing for the CSP and ESMB markets that it participates in:

This TAM data was put together by the market researchers at IDC and Lenovo plotted out its own ISC revenues and compound annual growth rate against that TAM. In that chart above, the $87 billion, $125 billion, and $248 billion are not endpoints for the calendar-adjusted Lenovo ISG revenues – not even close – but are just the summation of the stacked bars for the 2016, 2020, and 2024 calendar years.

What this looks like to our eyes is that the market and Lenovo have both just gone exponential in terms of revenue growth, and thanks in large part to the GenAI opportunity in both the enterprise and among governments, HPC centers, and the largest AI model builders.

Our forecast for Lenovo’s annual ISG revenues, shown below, certainly also shows that:

But it also shows that for all of the hard work over the past decade, ISG has not been a profitable endeavor. And you can’t blame all of that on HPC and AI.