# After Seven CEOs in 10 Years, Imagination Is Sticking to Its Strategy > Source: > Published: 2026-08-07 22:00:00+00:00 KINGS LANGLEY, U.K. — As Imagination Technologies’ seventh CEO in a decade takes the helm at the British IP firm, the company has abandoned its CPU and NPU ambitions to strengthen its core GPU business, but this definitely isn’t [another](https://www.eetimes.com/imaginations-new-ceo-plots-turnaround-strategy/) [reboot](https://www.eetimes.com/rebooting-imagination-a-heterogeneous-compute-strategy/), Imagination chief revenue officer Jake Kochnowicz told EE Times in a recent interview. Since EE Times last visited Imagination headquarters in Kings Langley [two years ago](https://www.eetimes.com/imagination-raises-100-million-investment-to-take-on-edge-ai/), the company has lost its CEO, CTO, and VP of marketing. Despite changes to the executive team, senior engineering leaders are still in place, many with decades of service, Kochnowicz said. “We still have senior leaders in position—VPs, senior directors, key architects, and so on—[who] have been here for many, many years,” he said. Imagination’s CFO, general counsel, and HR leader are still in their positions, but some flattening of the leadership hierarchy is a good thing, Kochnowicz argued. [View All](https://www.eetimes.com/category/sponsored-content/) “We’re trying to improve [leadership structure] so that we can go faster and leverage the fact that we are a relatively small company still, so we should be agile,” he said. “I don’t think it’s anything radical… what we’re building, the direction we’re going, is largely the same as what we set out previously.” Kochnowicz is Imagination’s chief revenue officer; he runs product (including marketing), operations, research, sales, and customer support. New Imagination CTO Ed Plowman’s remit is focused on R&D. New CEO Markus Mosen, who joined Imagination in February 2026, replaced Simon Beresford-Wylie, who retired. Mosen has made some changes to Imagination’s operations, but overall, is continuing along the path set by Beresford-Wylie, Kochnowicz said. **Roadmap focus** Imagination’s roadmap has gained strategic focus in the last couple of years after abandoning its RISC-V CPU effort and shelving a standalone AI accelerator. Imagination had launched a RISC-V CPU in 2021 but discontinued it at the start of 2025. “The time frame we’d have had to continue to invest before we’d get a return was very long,” Kochnowicz said. “The market was, and is still, ripe for consolidation, and some RISC-V CPU companies are starting to struggle… rather than split our investment between GPU and CPU, where the GPU was generating revenue, it was odd to starve that of resources by having a CPU—we’re a GPU company, let’s focus on that.” Imagination’s NPU was shelved in 2024; the company retains the IP, but as it was developed pre-transformer, there is limited scope for it in the world of [ChatGPT](https://www.embedded.com/7-expert-tips-for-writing-embedded-software-with-chatgpt/), Kochnowicz said. Imagination’s GPU portfolio will address AI, but the company remains a GPU company that does AI rather than the other way around, he added. “We are a graphics company and will remain that way,” he said. “What we are doing is making AI a first-class citizen alongside graphics; it’s a core pillar of what we’re doing, and we treat them both equally.” In general, it makes sense to avoid software fragmentation and programming model fragmentation by keeping graphics and AI GPUs on the same architecture, Kochnowicz said, noting that some customers want to do both—perhaps cloud gaming workloads during the day when demand is high, and AI at night when it is quiet. Imagination has two big customers for its existing E-series, one developing an AI PC/graphics application in China and the other a global consumer electronics company. **GPU architecture** Without the distractions of CPUs and NPUs, Imagination is betting its future on a [next-generation GPU architecture](https://www.edn.com/gpus-a-high-throughput-architecture-confronting-a-workload-shift/) that can scale from automotive to the data center and from graphics to AI workloads. The company took out a $100 million loan two years ago, which has been used to develop this next-generation architecture, Kochnowicz said. The forthcoming F-series will be a “significant refresh” of Imagination’s GPU architecture and will enable a “big leap” in the size and scale of GPUs it can power. “This is a new platform we’ll be able to invest more and more into over the next decade,” he said. “We also fixed a lot of technical debt, overhauled a lot of our verification and design methodologies and infrastructure to allow us to go faster and to upscale our team more easily.” The F-series will target automotive and data center use cases. It will also scale to a bigger silicon area than the E-series, which currently tops out around 300 mm2 in TSMC N5. The F-series will not reach 1 PFLOPS, but the generation after it could, Kochnowicz said—this would be huge for an IP-driven design. Most customers for the F-series will not sell chips. In general, they are more verticalized, perhaps selling PCIe cards for existing server infrastructure or even pre-assembled servers, but some will verticalize entirely, running their own data centers based on their own custom GPUs. A classic example is cloud gaming, Kochnowicz said, where customers want to add video processing or other interface technologies to their GPU silicon to customize it for their needs. Licensing IP enables them to create a GPU with video codecs, scaling, and compression all on one product, he said. The F-series is due for its official launch around the end of 2026, but two big customers have already been signed, according to Kochnowicz. **Chinese market** Some of Imagination’s thorniest problems to date have involved China. The company was accused of stealthily transferring its technology to China by a former CEO after it was [bought by Chinese-backed private equity firm Canyon Bridge Capital Partners](https://www.eetimes.com/imagination-technologies-life-after-apple/) in 2017. New CEO Markus Mosen was previously CEO of WeEn Semiconductors, a European bipolar power technology startup that now has Chinese ownership, like Imagination. EE Times reported in 2019 that China represented around one-third of the company’s business. That figure varies today, Kochnowicz said, since IP licensing is relatively intermittent. Imagination did a lot of business in China between 2019 and 2022, but many deals had subsequent challenges, either because companies folded or because of regulatory restrictions. These challenges have settled down, and Imagination’s Chinese business is growing, he said. One Chinese customer is a large public EV company using Imagination’s GPU in an ADAS system. The current, more manageable regulatory environment means Imagination can also license GPU IP for Chinese “big GPUs,” provided potential customers are not on the U.S. Bureau of Industry and Security (BIS)’s Entity List. In practice, Chinese customers are blocked from building large GPUs on advanced nodes due to a lack of access to leading-edge foundry technology, Kochnowicz said. Imagination GPUs are not intended to compete with the state-of-the-art supercomputing GPUs from the likes of Nvidia, in any case, Kochnowicz said. Imagination’s IP generally goes into feature-rich GPUs for Chinese desktop-class graphics processors as well as Chinese cloud gaming and data center AI applications. Imagination also prohibits military applications for small GPUs such as drones in its licensing terms. “Increasingly, our biggest opportunities are with customers that are vertically integrated, not just making a chip and selling it, perhaps building cards for a desktop computer or edge server, so they are product companies that make their own chip,” he said. “Many are extremely sensitive to the fact that they need EDA tools, they need IP, they need foundry access, and getting caught red-handed [selling chips to a banned application] would be a death sentence for the company [because they’d end up on the Entity List], so they are nervous about that.” Imagination’s team in China is an integrated part of Imagination, not a subsidiary. The team is mostly customer support and sales personnel, and they report to Kochnowicz in the U.K. Far from “technology transfer,” Imagination’s activities in the region are within normal practices for IP licensing, Kochnowicz said. This naturally includes some scope to modify the IP for specific use cases, but all patents and tools are kept within Imagination, he added. “We don’t develop IP in the U.S. or China, because if you develop in one place, you can’t sell to the other,” Kochnowicz said. “Both are very important markets for us.” ##### Read also: [AI Chip Startup Taalas Acquired by AMD](https://www.eetimes.com/ai-chip-startup-taalas-acquired-by-amd/) [Why Qualcomm Bought an Open AI Software Stack](https://www.eetimes.com/why-qualcomm-bought-an-open-ai-software-stack/) [Microchip Acquires Edge AI Chip Startup Hailo](https://www.eetimes.com/microchip-acquires-edge-ai-chip-startup-hailo/)